Lecture 13: Free particle, wavepackets, phase and group

velocity

1 Eigenstates of the momentum operator

The momentum eigenstates in the position representation, up (x) defined by
ℏ ∂
p̂up (x) = up (x) = pup (x), (1)
i ∂x
and given by
1
up (x) = √ eipx/ℏ , (2)
2πℏ
cannot be normalized in free space to be interpreted as a probability density since |up (x)|2 = 2πℏ
1
,
R∞ 2
and −∞ dx |up (x)| diverges. However, they do satisfy the continuum orthonormality condition
Z ∞
dxu∗p (x)up′ (x) = δ (p − p′ ) . (3)
−∞

This normalization corresponds to a uniform particle density (particle per meter) given by |up (x)|2 =
1
2πℏ
. Let us calculate the probability current (particles moving past a point x per second), defined
by
  ∗    
ℏ ∗ ∂ψ ∂ψ ℏ ∗ ∂ψ
j(x) = ψ (x) (x) − (x) ψ(x) = Im ψ (x) (x) → see PS (4)
2im ∂x ∂x m ∂x
Another useful expression for the current is:
 
p̂ ∗
j(x) = Re ψ (x) ψ(x) . (5)
m
For ψ(x) = up (x) we find
  
ℏ 1 ip ip
j(x) = − −
2im 2πℏ ℏ ℏ (6)
1 p
= ,
2πℏ m
which is exactly what we expect for a uniform particle density |up (x)|2 = 2πℏ1
moving at velocity
p
v = m.
In general, choosing a wavefunction ψ(x) = Ceipx corresponds to particles moving at velocity
p
m
, a particle density |ψ(x)|2 = |C|2 , and a particle current j(x) = |C|2 mp . Alternatives to deal
with the normalization problem (wavefunction not squareintegrable) for momentum states are:

1
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

1.1 Wavepackets
A superposition of a finite number of momentum eigenstates is not directly normalizable (square-
integrable), but a wavepacket consisting of an infinite number of momentum eigenstates (Fourier
components) is.

Figure 1: A wavepacket ψ(x) in position space or
ϕ(k) in momentum space whose wavefunction
for large x (or k ) falls off faster than x−1/2 k −1/2 can be directly normalized.

1.2 Periodic boundary conditions

Assume box of finite length L, require periodic boundary conditions

ψ(0) = ψ(L) (7)

For plane waves eipx/ℏ this implies that eipL/ℏ = 1 or pLℏ
= kL = n2π, n integer, i.e. momen-
2π
tum is quantized, pn = nℏko with k0 = L . The corresponding momentum states are normalizable
in the interval [0, L],
Z L
2
dx Ceipx/ℏ = L|C|2 = 1
0 (8)
1 ipn x/ℏ
upn (x) = √ e
L
→ normalized momentum eigenstates in box of size L with pn = nℏko
Z L
dxupn ∗ (x)upm (x) = δnm → orthonormality condition in box (9)
0

2
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

We perform all calculations for fixed size box, then take the limit L → ∞ (i.e. k0 → 0,
momentum spectrum becomes continuous). All physically sensible results will be independent of
the initially chosen box size L as long as L is large compared to distances of interest.

Figure 2: Wavefunction in box of length L with periodic boundary conditions.

2 Time evolution of free-particle wavepackets

In free space we often work with normalized Gaussian wavepackets
x2
1 −
4w02
Ψ(x, t = 0) = e (10)
(2π)1/4 w0 1/2

Written in this form we have

x2
1 −
• |Ψ(x, 0)|2 = (2π)1/2 w0
e 2w0 2

R
• dx|Ψ(x, 0)|2 = 1
• ⟨x⟩ = 0
• ⟨x2 ⟩ = w02
• (δx)2 = ⟨x2 ⟩ − ⟨x⟩2 = w02
∆x = w0 is the uncertainty or rms width (root-mean-square width) of the wavepacket. Why
do we prefer this Gaussian form of wavepacket?
1. Particularly simple and symmetric, the Fourier transform is also a Gaussian wavepacket:

k2
1 − 2
4k0
ϕ(k) = e (11)
(2π)1/4 k0 1/2
1
with k0 = 2w0
· (∆k)2 = ⟨k 2 ⟩ − ⟨k⟩ = k02
1 ℏ


2. This is a wavepacket with the minimum uncertainty ∆x∆k = 2
∆x∆p = 2
allowed by
QM
3. Physical system after give rise to Gaussian broadening in momentum or position, e.g., ther-
mal distribution of atomic momenta in a gas is a Gaussian distribution.

3
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

2.1 How do we make a wavepacket move at a velocity?

In order to make a wavepacket moving at velocity v1 , we displace the distribution in momentum
space from ⟨p⟩ = ⟨ℏk⟩ = 0 to ⟨p⟩ = ⟨ℏk⟩ = ℏk1 = mv1 (see Fig. III).
(k−k1 )2
1 − 2
4k0
ϕ(k) = 1/2
e (12)
(2π)1/4 k0
The inverse Fourier transform, i.e., the spatial wavefunction
2
1 − x 2 ik1 x
Ψ(x, t = 0) = e 4wo e (13)
(2π)1/4 w0 1/2
is still a Gaussian, but now with a phase variation eik1 x , rather than a constant phase over the
wavepacket (compare Eq. (14-11). This phase variation eik1 x in position space "encodes" the
motion of the wavepacket at velocity v1 = ℏk m
1
: The dominant de Broglie wavelength in the
wavepacket corresponds to a wavevector k1 , or a momentum ℏk1 .

Figure 3: Moving Gaussian wavepacket with average velocity v1 = ℏk1 /m and spatial wavefunc-
x2 ik1 x
− 2e
1 4w0
tion ψ1 (x) = 1/2 e .
(2π)1/4 w0

2.2 How does a free-space Gaussian wave packet evolve in time?

In general, we expand a wavefunction Ψ(x, 0) into energy eigenfunctions uE (x), and then evolve
the energy eigenfunctions as e−iEt/ℏ .
In free space, there is only KE. Then the momentum eigenstates up (x) are simultaneous eigen-
states of energy:

4
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

p̂2
Ĥup (x) = up (x)
2m
 2
1 ∂ 1
= ℏi √ eipx/ℏ (14)
2m ∂x 2πℏ
p2
= up (x)
2m
or

p2
Ĥup (x) = up (x)
2m (15)
= Ep up (x)
in free space. The energy eigenstates are said to be doubly degenerate: For each eigenvalue
√ of
energy E > 0 there are two different momentum states (namely u±p (x) with p = 2mℏ ) that
have the same energy. It follows that a momentum eigenstate with eigenvalue p evolves in time as
e−iEp t/ℏ , so that the wavefunction in momentum space evolves in time as
p2
Φ(p, t) = Φ(p, 0)e−i 2m t/ℏ (16)
→ Time evolution of momentum eigenfunctions in free space. The wavefunction in real space
is given by the inverse Fourier transform Ψ(x, t), or equivalently, as the superposition of energy
eigenfunctions with their corresponding phase evolution factors e−iEp t/ℏ :
Z
1
Ψ(x, t) = √ dpΦ(p, t)eipx/ℏ
2πℏ 
Z
( = dpΦ(p, t)up (x)
Z
1 p2
=√ dpΦ(p, 0)eipx/ℏ e− 2m t/ℏ (17)
Z 2πℏ
= dpΦ(p, 0)Up (x, t)
Z
= dpΦ(p, t)up (x)
p2 p2
where Up (x, t) = up (x)e− 2m t/ℏ = √2πℏ
1
eipx/ℏ e−i 2m t/ℏ are the time-dependent momentum eigen-
functions in free space. The above equation shows that the phases of different Fourier components
1
up (x) = √2πℏ eipx/ℏ evolve in time at different speeds, the "running out of phase" of different
Fourier components leads to a spreading of the wavepacket in position space. In the problem sets
you will show that the rms width ∆x(t) = w(t) of the wavepacket grows in time as
s
ℏ2 t2
w(t) = w0 1 + 2 4 (18)
m w0

Since a wavepacket contains different momentum components, it changes in time in free spae even
2
though there are non external forces acting. For long times t ≫ t0 = mw
ℏ
o
the wavepacket spreads

5
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

as w(t) ≈ mw ℏ
0
ℏ
t, i.e. at a speed v0 = mw 0
that is inversely proportional to its initial size. That speed
is negligible for macroscopic wavepacket size, but can be appreciable for initially well-localized
microscopic objects. The spreading of a wavepacket in free space was early evidence that the
wavepacket size cannot be identified with the particle size. The spreading is due to the quadratic
(i.e. not linear) dependence of the energy, and hence the phase evolution rate, on momentum. Note
that the wavepacket of a massless particle, e.g. a photon, with E = pc would not spread. (The SE
is non-relativistic and does not apply to photons.)

2.3 Motion of wave packets, group velocity, and stationary phase

Why is it that a wavefunction
1 2
x
− 4w
Ψ(x, 0) = e 02 eik1 x (19)
(2π)1/4 w 0
1/2

ℏk1
represents a particle moving at velocity v1 = m
? Since a crest of a single momentum component
2
√1 e−k1 x e −i ℏk t 2π 2π
uk1 (x, t) = 2π
2m moves forward a distance λ = k1
in a time T = ω1
(remember that
2
ω1 = ℏk1
2m
and e −iE1 t/ℏ
= e−iω1 t ), the velocity of the crest is vph = λ
T
= 2π ω
k1 2π
= ωk11 = ℏk
2m
1

ω1 ℏk1 p1
= vph =
= (20)
k1 2m 2m
This is the phase velocity of a momentum component.
The particle does not move at the phase velocity vph = ωk11 at which the plane wave associated
with a single momentum moves forward. At what velocity then?

• Look at exponent and write:

ℏk12
E1 = E1 (k1 ) = ℏω1 = ℏω (k1 ) = 2m
 
x2
− 2 +ik1 x−iω(k1 )t
4w0
• e

Remember Fermat’s principle of stationary phase: path is defined by region of space where
phasors point mostly in one direction, i.e. where the phase ϕ(k) = kx + ω (k1 ) t does not vary
between different momentum components k to lowest order
 
∂ϕ ∂ω
0= =x− (21)
∂k ∂k
or
 
∂ω
x(t) = t (22)
∂k
Fermat’s principle leads us to the concept of group velocity
∂ω ℏk1 p1
vgr = (k1 ) = = (23)
∂k m m

6
8.04 Lecture 13: Free particle, wavepackets, phase and group velocity Vuletic - Spring 2024

Group velocity of the wavepacket at which the wavepacket, i.e. the region of constructive
interference, propagates. The difference between group and phase velocity is due to the fact the
∂ω
∂k
̸= ωk , or ∂E
∂p
= ∂(ℏω)
∂(ℏk)
̸= Ep , i.e. the quadratic dependence of KE on momentum in free space.
This is in contrast to photons with a linear dispersion relation ∂ω ∂k
= ωk = c in vacuum, where
group and phase velocity are the same. This is the same principle that you already learned in 8.03
about waves: Once the frequency-wavenumber (or energy-momentum) relation has dispersion,
∂E
∂p
= ∂(ω)
∂(k)
̸= ωk , then the group velocity for a wavepacket is different from the phase velocity.