Gaussian wave packet

Lecture notes

The Gaussian wave function is determined by its wave function in the wave-
vector space. In 1-dimensional space it is:

(k − k0 )2
 
f (k) = A exp − , (1)
4∆2k
where A is the normalization constant and ∆k is the width of the packet in the
k-space. Its graph as function of K is a bell-shaped curve centered near k0 .
The
R ∞ normalization constant should be found from the normalization condition
2
−∞
| f (k) | dk = 1. This integral is of the Gaussian type. Gaussian integral
in 1 dimension is defined reads:
Z ∞ r
2

π
I(α) = exp −αx dx = (2)
∞ α
−1/2
Employing this formula one finds A = (2π)−1/4 ∆k . To relate ∆k to the
average characteristics of the Gaussian wave packet, let us calculate the average
square of the deviation of the wave vector from its central point k = k0 (quantum
fluctuation or square of uncertainty). Its definition is:
Z ∞
2 2
h(∆k) i = (k − k0 ) | f (k) |2 dk (3)
∞

This integral is reduced to the integral of a type:

Z ∞
dI I(α)
x2 exp −αx2 dx = −


I1 (α) = = (4)
∞ dα 2α
2
Employing this equation, we find that h(∆k) i = ∆2k .
Next we find the wave function of the Gaussian packet in the coordinate
space applying the standard Fourier transformation:
Z ∞
1
ψ(x) = √ f (k)eikx dk (5)
2π ∞
This integral is reduced to a slightly generalized standard Gaussian integral (2):
Z ∞ r  2
π β
exp −αx2 + iβx dx =


I(α, β) = exp (6)
∞ α 2α

Employing the latter equation with α = 1/(4∆2q ) and β = ix eq. (1) for f (k),
one can transform eq. (5) as follows:
√
eik0 x 21/4 ∆k
exp −∆2k x2


ψ(x) = 1/4
(7)
π

1
This wave function has the shape of the Gaussian wave packet (2), but in the
coordinate space. It is centered at x = 0 and its width is ∆x = 1/(2∆k ). The
wave function
x2
 
ψ(x) = (2π)−1/4 ∆−1/2
x exp − (8)
4∆2x
is automatically normalized as it follows from Percival theorem. We also find
the uncertainty relation in the form ∆x ∆k = 1/2. We will see that at any
non-zero time t > 0 the coordinate uncertainty increases.
To find the time evolution of the Gaussian wave packet we again start with
the wave function in the k-space, since the energy is well defined for the states
2 2
k
with fixed k: E(k) = ~2m . Thus , the time evolution of the wave function (1)
has the following form:

(k − k0 )2 i~k 2 t
 
−1/2
f (k, t) = (2π)−1/4 ∆k exp − − (9)
4∆2k 2m

The modulus of this

 wave  function does not depend on time since the modulus
i~k2
of the factor exp − 2m is 1. Thus, the average square of the momentum
fluctuation is the same as at t = 0 (equal to ∆2k ). However, the uncertainty of
coordinate increases with time as it is shown below. The time dependent wave
function ψ(x, t) is the Fourier transform of the function f (k, t) (9):
Z ∞
1
ψ(x, t) = √ f (k, t)eikx dk (10)
2π ∞
The integral is again of the Gaussian type (6). The exponential integrand has
the argument
2
(k − k0 ) i~k 2 t
− 2 + ikx −
4∆k 2m
It is reasonable to subtract from
it its value at k = k0 and compensate it by a
factor exp ik0 x − i~k02 t/(2m) in front of the integral. Then the exponent in
the integrand acquires a form:
2


(k − k0 ) i~ k 2 − k02
− + i(k − k0 )x − (11)
4∆2k 2m
2
The last term can be transformed using a simple identity k 2 − k02 = (k − k0 ) +
2k0 (k − k0 ). Then the previous expression for the argument (11) can be rewrit-
ten as follows:
 
1 i~t 2
− + (k − k0 ) + i (k − k0 ) (x − v0 t) ,
4∆2k 2m

where v0 = ~k0 /m is the classical velocity of the center of the packet. It is also
1 2
reasonable to replace in the previous expression the fraction 4∆ 2 by ∆x . Then
k

2
the integral is exactly the Gaussian I(α, β) (6) with α = ∆2x + i~t/(2m) and
β = x − v0 t. Thus, we find:
 
i~k2 t 2
exp ik0 x − 2m0 − 4 (x−v 0 t)
(∆2x + 2m
i~t
)
ψ(x, t) = 1/2
(12)
1/4 i~t


(2π) ∆2x + 2m

Its square of modulus reads:

h i
(x−v0 t)2
exp − 2(∆ (t)) 2
2 x
|ψ(x, t)| = √ , (13)
2π∆x (t)
 1/2
~ 2 t2
where ∆x (t) = ∆2x + 4m2 ∆2x is the time-dependent uncertainty of the parti-
2m∆2
cle coordinate. We see that it remains practically unchanged at t  t0 = ~ x .
Asymptotically at t  t0 , the uncertainty of coordinate grows linearly with
time: ∆∆x (t)
x
≈ tt0 . The wave packet smears out with time since it is a superpo-
sition of states with different velocities. It is easy to estimate the uncertainty
of velocity ∆v = ~∆k /m. At sufficiently large time it leads to the smearing of
kt
the wave packet to the width ∆x (t) ∼ ∆v t ∼ ~∆ m ∼ m∆ ~t
x
. This expression
coincides with the asymptotic obtained from exact solution, but this estimate
has more general character and relates to any wave packet. It is valid when the
packet smearing due to the dispersion of velocity becomes larger than the initial
width of the wave packet ∆x , i.e. at t > t0 .