Professional Documents
Culture Documents
Complement G: One-Dimensional Gaussian Wave Packet: Spreading of The Wave Packet
Complement G: One-Dimensional Gaussian Wave Packet: Spreading of The Wave Packet
Complement GI
One-dimensional Gaussian wave packet: spreading of the wave packet
This wave packet is obtained by superposing plane waves e with the coefficients:
2
1 2
( 0) = e 4 ( 0)
(2)
2 (2 )3 4
+
2
( + )2
( )= e d (3)
where and are complex numbers [note that, for the integral (3) to converge, we must
have Re 2 0]. The method of residues enables us to show that this integral does not
57
COMPLEMENT GI •
depend on :
( )= ( 0) (4)
Now all that remains is to evaluate (1 0), which can be done classically, through a
double integration in the plane and a change into polar coordinates:
+
2
(1 0) = e d = (6)
Thus we have:
+
2
( + )2
e d = (7)
with: 4 Arg + 4.
Let us now calculate ( 0). To do this, let us group, in the exponents of (1), the
-dependent terms into a perfect square, by writing them in the form:
2 2 2 2
2 2
( 0) + = 0 2
+ 0 2
(8)
4 4
We can then use (7), which yields:
1 4
2 2 2
( 0) = 2
e 0
e (9)
We find, as could be expected, that the Fourier transform of a Gaussian function is also
a Gaussian (cf. Appendix I).
At time = 0, the probability density of the particle is therefore given by:
2 2 2
( 0) 2 = 2
e 2
(10)
The curve representing ( 0) 2 is the familiar bell-shaped curve. The center of the
wave packet [the maximum of ( 0) 2 ] is situated at the point = 0. This is indeed
what we could have found if we had applied the general formula (C-16) of Chapter I
since, in this particular case, the function ( ) is real.
= (11)
2
58
• ONE-DIMENSIONAL GAUSSIAN WAVE PACKET: SPREADING OF THE WAVE PACKET
= (12)
2
We can proceed in the same way to calculate the width , since ( 0) 2 is also a
Gaussian function. This gives:
1
= (13-a)
or:
~
= (13-b)
Thus we obtain:
~
= (14)
2
a result which is entirely compatible with the Heisenberg relation.
3-a. Calculation of ( )
with ( ) = ~ 2 2 (dispersion relation for a free particle). Let us show that, at time ,
the wave packet still remains Gaussian. Expression (15) can be transformed by grouping,
as above, all the -dependent terms in the exponents into a perfect square. We can then
use (7), and we find:
2
~ 0
2 1 4
2 e
( )= e 0
exp (16-a)
4~2 2 1 4
2
2~
4 + +
2
59
COMPLEMENT GI •
+
Let us show that the norm of the wave packet, ( ) 2 d , is not time-
dependent (we shall see in Chapter III that this property results from the fact that the
Hamiltonian of the particle is Hermitian). We could, to this end, use (7) again in order
to integrate expression (17) from to + . It is quicker to observe from expression
(15) that the Fourier transform of ( ) is given by:
( )
( )=e ( 0) (18)
( ) therefore obviously has the same norm as ( 0). Now the Parseval-Plancherel
equality [cf. relation (45) of Appendix I] tells us that ( ) and ( ) have the same
norm, as do ( 0) and ( 0). From this we deduce that ( ) has the same norm as
( 0).
We could have expected this result, in view of the general expression (C-32) of
Chapter I, which gives the group velocity .
Let us take up formula (17) again. The width ( ) of the wave packet at time ,
from definition (11), is equal to:
4~2 2
()= 1+ 2 4
(20)
2
We see (cf. Fig. 1) that the evolution of the wave packet is not purely a simple
displacement at a velocity 0 . The wave packet also undergoes a deformation. When
increases from to 0, the width of the wave packet decreases, reaching a minimum
at = 0. Then, as continues to increase, ( ) grows without bound (spreading of the
wave packet).
It can be seen from (17) that the height of the wave packet also varies, but in
opposition to the width, so the norm of ( ) remains constant.
The properties of the function ( ) are completely different. In fact [cf. formula
(18)]:
( ) = ( 0) (21)
60
• ONE-DIMENSIONAL GAUSSIAN WAVE PACKET: SPREADING OF THE WAVE PACKET
ψ (x, t) 2
0
x
t<0 t=0 t>0
Figure 1: For negative , the Gaussian wave packet decreases in width as it propagates.
At time = 0, it is a “minimum” wave packet: the product is equal to } 2. Then,
for 0, the wave packet spreads again as it propagates.
Therefore, the average momentum of the wave packet (~ 0 ) and its momentum dispersion
(~ ) do not vary in time. We shall see later (cf. Chap. III) that this arises from the
fact that the momentum is a constant of the motion for a free particle. Physically, it
is clear that since the free particle encounters no obstacle, the momentum distribution
cannot change.
The existence of a momentum dispersion =~ = ~ means that the velocity
~
of the particle is only known to within = = . Imagine a group of classical
particles starting at time = 0 from the point = 0, with a velocity dispersion equal
~
to . At time , the dispersion of their positions will be = = ; this
dispersion increases linearly with , as shown in Figure 2. Let us draw on the same graph
the curve which gives the evolution in time of ( ); when becomes infinite, ()
practically coincides with [the branch of the hyperbola which represents ( ) has
for its asymptotes the straight lines which correspond to ]. Thus, we can say that,
when is very large, there exists a quasi-classical interpretation of the width . On
the other hand, when approaches 0, ( ) takes on values which differ more and more
from . The quantum particle must indeed constantly satisfy the Heisenberg relation
> ~ 2 which, since is fixed, imposes a lower limit on . This corresponds
to what can be seen in Figure 2.
Comments:
(i) The spreading of a packet of free waves is a general phenomenon which is not
limited to the special case studied here. It can be shown that, for an arbitrary free
wave packet, the variation in time of its width has the shape shown in Figure 2
(cf. exercise 4 of Complement LIII ).
61
COMPLEMENT GI •
δxcl
t
0
Figure 2: Variation in time of the width ∆ of the wave packet of Figure 1. For large ,
∆ approaches the dispersion of the positions of a group of classical particles which
left = 0 at time = 0 with a velocity dispersion ∆ .
2 d2
2 (22)
d 2 = 0
If, on the contrary, the phase ( ) cannot be approximated in the domain by a linear
function with an error much smaller than 2 , we find when we return to the argument
of Chapter I that the wave packet is larger than was predicted by (C-17).
In the case of the Gaussian wave packet studied in the present complement,
1 ~ 2
we have and ( ) = . Consequently, condition (22) can be written
2
1 2~
2 . Indeed, we can verify from (20) that, as long as this condition is fulfilled,
the product is approximately equal to 1.
62
• STATIONARY STATES OF A PARTICLE IN ONE-DIMENSIONAL SQUARE POTENTIALS
Complement HI
Stationary states of a particle in one-dimensional square potentials
We saw in Chapter I (cf. § D-2) the interest in studying the motion of a particle in a
“square potential” whose rapid spatial variations for certain values of introduce purely
quantum effects. The shape of the wave functions associated with the stationary states
of the particle was predicted by considering an optical analogy which enabled us to
understand very simply how these new physical effects appear.
In this complement, we outline the quantitative calculation of the stationary states
of the particle. We shall give the results of this calculation for a certain number of simple
cases, and discuss their physical implications. We limit ourselves to one-dimensional
models (cf. Complement FI ).
d2 2
( )+ 2 ( ) ( )=0 (1)
d 2 ~
We shall distinguish between several cases:
(i)
Let us introduce the positive constant , defined by
~2 2
= (2)
2
The solution of equation (1) can then be written:
( )= e + e (3)
63
COMPLEMENT HI •
( )= e + e (5)
How does the wave function behave at a point = 1 , where the potential ( )
is discontinuous? One might expect the wave function ( ) to behave strangely at this
point, becoming itself discontinuous, for example. The aim of this section is to show that
this is not the case: ( ) and d d are continuous, and it is only the second derivative
d2 d 2 that is discontinuous at = 1 .
Without giving a rigorous proof, let us try to understand this property. To do this, recall
that a square potential must be considered (cf. Chap. I, § D-2-a) as the limit, when 0, of
a potential ( ) equal to ( ) outside the interval [ 1 1 + ], and varying continuously
within this interval. Then consider the equation:
d2 2
( )+ [ ( )] ( )=0 (6)
d 2 ~2
where ( ) is assumed to be bounded, independently of , within the interval [ 1 1 + ].
Choose a solution ( ) which, for 1 , coincides with a given solution of (1). The
problem is to show that, when 0, ( ) tends towards a function ( ) which is continuous
and differentiable at = 1 . Let us grant that ( ) remains bounded1 , whatever the value of
, in the neighborhood of = 1 . Physically, this means that the probability density remains
finite. Integrating (6) between 1 and 1 + , we obtain:
1+
d d 2
( 1 + ) ( 1 )= [ ( ) ] ( )d (7)
d d ~2 1
At the limit where 0, the function to be integrated on the right-hand side of this expression
remains bounded, owing to our previous assumption. Consequently, if tends towards zero, the
integral also tends towards zero, and:
d d
( 1 + ) ( 1 ) 0 (8)
d d 0
64
• STATIONARY STATES OF A PARTICLE IN ONE-DIMENSIONAL SQUARE POTENTIALS
2
from (1), makes a jump at = 1, which is equal to ( 1) [where represents the
~2
change in ( ) at = 1 ].
Comment:
Let us now carry out the quantitative calculation of the stationary states, performed ac-
cording to the method described above, for all the forms of the potential ( ) considered
in § D-2-c of Chapter I. This will ensure that the form of the solutions is indeed the one
predicted by the optical analogy.
V(x)
V0
Figure 1: Potential step.
I II
x
0
2
= 1 (9)
~2
2 ( 0)
= 2 (10)
~2
65
COMPLEMENT HI •
The solution of (1) has the form (3) in the two regions I ( 0) and II ( 0):
I( )= e + e (11)
1 1
1 1
II ( )= e + e (12)
2 2
2 2
Since equation (1) is homogeneous, the calculation method of § 1-c can only enable
us to determine the ratios 1 1 , 2 1 and 2 1 . In fact, the two matching conditions
at = 0 do not suffice for the determination of these three ratios. This is why we shall
choose 2 = 0, which amounts to limiting ourselves to the case of an incident particle
coming from = . The matching conditions then give:
1 1 2
= (13)
1 1+ 2
2 2 1
= (14)
1 1 + 2
I ( ) is the superposition of two waves. The first one (the term in 1 ) corresponds
to an incident particle, with momentum = ~ 1 , propagating from left to right. The
second one (the term in 1 ) corresponds to a reflected particle, with momentum ~ 1 ,
propagating in the opposite direction. Since we have chosen 2 = 0, II ( ) consists of
only one wave, which is associated with a transmitted particle. We shall see in Chapter III
(cf. § D-1-c- ) how it is possible, using the concept of a probability current, to define the
transmission coefficient and the reflection coefficient of the potential step (see also
§ 2 of Complement BIII ). These coefficients give the probability for the particle, arriving
from = , to pass the potential step at = 0 or to turn back. Thus we find:
2
1
= (15)
1
and, for2 :
2
2 2
= (16)
1 1
4 1 2
=1 2
(17)
( 1 + 2)
4 1 2
= 2
(18)
( 1 + 2)
2 The physical origin of the factor 2 1, appearing in is discussed in § 2 of Complement JI .
66
• STATIONARY STATES OF A PARTICLE IN ONE-DIMENSIONAL SQUARE POTENTIALS
It is easy to verify that + = 1: it is certain that the particle will be either transmitted
or reflected. Contrary to the predictions of classical mechanics, the incident particle has
a non-zero probability of turning back. This point was explained in Chapter I, using the
optical analogy and considering the reflection of a light wave from a plane interface (with
1 2 ). Furthermore, we know that in optics, no phase delay is created by such a
reflection; equations (13) and (14) do indeed show that the ratios 1 1 and 2 1 are
real. Therefore, the quantum particle is not slowed down by its reflection or transmission
(cf. Complement JI , § 2). Finally, using (9), (10) and (18) it is easy to verify that, if
0 , we have 1: when the energy of the particle is sufficiently large compared
to the height of the potential step, the particle clears this step as if it did not exist.
2 ( 0 )
= 2 (19)
~2
II ( )= e + e (20)
2 2
2 2
2 =0 (21)
1 1 2
= (22)
1 1 + 2
2 2 1
= (23)
1 1+ 2
67
COMPLEMENT HI •
Comment:
1 1
(25)
2 0
In the 0 region, the wave, whose range decreases without bound, tends towards
zero. Since ( 1 + 1 ) 0, the wave function ( ) goes to zero at = 0, so
that it remains continuous at this point. On the other hand, its derivative, which
changes abruptly from the value 2 1 to zero, is no longer continuous. This is
due to the fact that since the potential jump is infinite at = 0, the integral of
(7) no longer tends towards zero when tends towards 0.
V(x)
V0
I II III
Figure 2: Square potential barrier.
x
0 l
I( ) = e + e (26-a)
1 1
1 1
II ( ) = e + e (26-b)
2 2
2 2
III ( )= e + e (26-c)
1 1
3 3
68
• STATIONARY STATES OF A PARTICLE IN ONE-DIMENSIONAL SQUARE POTENTIALS
T
1
4E(E V0)
4E(E V0) + V 20
0 l
/k2 2 /k2
It is then easy to verify that + = 1. Taking (9) and (10) into account, we have:
4 ( 0)
= (29)
2 2
4 ( 0) + 0 sin 2 ( 0) ~
69
COMPLEMENT HI •
2
3 4 ( 0 )
= = (30)
1 4 ( 0 )+ 0
2 sinh2 2 ( 0 ) ~
We have already seen in Chapter I why, contrary to the classical predictions, the particle
has a non-zero probability of crossing the potential barrier. The wave function in region II
is not zero, but has the behavior of an “evanescent wave” of range 1 2 . When . 1 2 ,
the particle has a considerable probability of crossing the barrier by the “tunnel effect”.
This effect has numerous physical applications: the inversion of the ammonia molecule
(cf. Complement GIV ), the tunnel diode, the Josephson effect, the -decay of certain
nuclei, etc...
1 1.96
Å (32)
2 0
where and 0 are expressed in electron-volts (this formula is easily obtained by replacing, in
formula (8) of Complement AI , = 2 by 2 2 ). Now consider an electron of energy 1 eV
which encounters a barrier for which 0 = 2 eV and = 1 Å. The range of the evanescent wave
is then 1.96 Å, that is, of the order of : the electron must then have a considerable probability
of crossing the barrier. Indeed, formula (30) gives in this case:
0.78 (33)
The quantum result is radically different from the classical result: the electron has approximately
8 chances out of 10 of crossing the barrier.
Let us now assume that the incident particle is a proton (whose mass is about 1 840 times
that of the electron). The range 1 2 then becomes:
1 1.96 4.6 2
Å 10 Å (34)
2 1 840( 0 ) 0
If we retain the same values: = 1 eV, 0 = 2 eV, = 1 Å, we find a range 1 2, much smaller
than . Formula (31) then gives:
19
4 10 (35)
Under these conditions, the probability of the proton’s crossing the potential barrier is negligible.
This is all the more true if we apply (31) to macroscopic objects, for which we find such small
probabilities that they cannot possibly play any role in physical phenomena.
70
• STATIONARY STATES OF A PARTICLE IN ONE-DIMENSIONAL SQUARE POTENTIALS
V(x)
a a
+
2 0 2 x
Figure 4: Square well potential.
I II III
V0
I( ) = 1 e + 1 e (36-a)
II ( ) = 2 e + 2 e (36-b)
III ( )= 3 e + 3 e (36-c)
with
2
= (37)
~2
2 ( + 0)
= (38)
~2
Since ( ) must be bounded in region I, we must have:
1 =0 (39)
71
COMPLEMENT HI •
3 e
= ( + )2 e ( )2 e
1 4
2 2
3 +
= sin (41)
1 2
But ( ) must also be bounded in region III. Therefore, it is necessary that 3 = 0,
that is:
2
= e2 (42)
+
Since and depend on , equation (42) can only be satisfied for certain values
of . Imposing a bound on ( ) in all regions of space thus entails the quantization of
energy. More precisely, two cases are possible:
(i) if:
= e (43)
+
we have:
= tan (44)
2
Set:
2 0 2 2
0 = = + (45)
~2
We then obtain:
2 2 2
1 + 0
= 1 + tan2 = 2
= (46)
2
cos2
2
cos = (47a)
2 0
tan 0 (47b)
2
The energy levels are determined by the intersection of a straight line, having a slope 1 0 ,
with sinusoidal arcs (long dashed lines in Figure 5). Thus we obtain a certain number
of energy levels, whose wave functions are even. This becomes clear if we substitute
(43) into (40) and (41); it is easy to verify that 3 = 1 and that 2 = 2 , so that
( ) = ( ).
72