Quantum Mechanics 1 Lecture 3:
Formalism and Postulates of QM
Massimiliano Grazzini
University of Zurich
Outline
Hilbert spaces
Observables and operators
Correspondence principle
Ehrenfest theorem Momentum operator
Wrap Up: The postulates of Quantum Mechanics General form of the uncertainty principle
Time evolution
Wave packets
Observables and Operators
Let us consider a physical observable that characterises a quantum system 𝒜 The values that this quantity can assume are elements of the spectrum of this physical quantity
In classical mechanics physical quantities take in general continuous values In quantum mechanics there can also be quantities with a spectrum of
discrete values: for simplicity we will assume this is the case for the moment We call ( ) the discrete values of and we call the wave
function of the system in the state in which the quantity takes the value a
kk = 0,1,2.... 𝒜 ϕ
k𝒜 a
kWe assume
∫ | ϕ
k|
2dx = 1
If our system is an arbitrary state characterised by a wave function the measurement of the quantity will give one of the values 𝒜 a
kψ
3
According to the superposition principle, we can then say that a generic wave function of the system will be represented by
ψ = ∑
k
c
kϕ
kSuch expansion allows us to determine the probability to obtain (through a measurement) a value of the quantity a
k𝒜
Such probability must be positive definite, must be unity if , and must
vanish if c
k= 0 ψ = ϕ
kc
k∈ ℂ
the probability should be | c
k|
2The sum of these probabilities should be unity, so we have the constraint
∑
k| c
k|
2= ∫ | ψ |
2dx
If we multiply ψ * = ∑ by and we integrate we obtain
k
c*
kϕ*
kψ
∫ | ψ |
2dx = ∑
k
c*
k∫ ϕ*
kψ dx
4
Comparing with the previous expression we find
5
c
k= ∫ ψ ϕ*
kdx
and if we plug in this equation the expansion ψ = ∑ we see that
l
c
lϕ
l∫ ϕ
lϕ*
kdx = δ
klSo, we arrive at a structure of vector space with scalar product (recall MMP1 !) of which the functions form an orthonormal basis ϕ
kWe now want to introduce the concept of mean (expectation) value of the quantity 𝒜
⟨𝒜⟩ ≡ ∑
k
a
k| c
k|
2We can define it as a weighted average of the values where each of them is multiplied by the probability | c
k|
2a
k(1)
6
We now would like to express ⟨𝒜⟩ directly in terms of ψ
Since ⟨𝒜⟩ is bilinear in and such expression must be bilinear in and c
kc*
kψ ψ * We introduce an operator on the functions such that A ψ
⟨𝒜⟩ = ∫ ψ *(Aψ) dx
By using ⟨𝒜⟩ ≡ ∑ and we obtain
k
a
kc
kc*
kc*
k= ∫ ψ* ϕ
kdx
⟨𝒜⟩ = ∫ ψ * ∑
k
c
ka
kϕ
kdx
Comparing with Eq. we see that the action of the operator on is (2) A ψ
(2)
Aψ = ∑
k
a
kc
kϕ
kUsing Eq. for we obtain (1) c
k(Aψ)(x) = ∫ ∑
k
a
kϕ*
k(x′ )ϕ
k(x)ψ (x′ )dx′ ≡ ∫ K
A(x, x′ )ψ (x′ )dx′
7
(Aψ)(x) = ∫ ∑
k
a
kϕ*
k(x′ )ϕ
k(x)ψ (x′ )dx′ ≡ ∫ K
A(x, x′ )ψ (x′ )dx′
We thus conclude that the operator acts as an integral linear operator on A ψ To each physical quantity in quantum mechanics
we associate an appropriate linear operator From Aψ = ∑ we see that if we have and
k
a
kc
kϕ
kψ = ϕ
lc
k= δ
klAϕ
l= a
lϕ
lSo, the action of on is simply to multiply it by the corresponding and the eigenfunctions of the operator are just the solutions of the equation A ϕ
lA a
lAψ = a ψ
The fact that the values that a physical quantity can take are real poses some limitations on the properties of the corresponding operator 𝒜 A
⟨𝒜⟩ = ∫ ψ *(Aψ) dx = ⟨𝒜⟩* = ∫ (Aψ)*ψdx
On the other hand the adjoint operator is defined as ∫ ψ *(A χ) dx = ∫ (A
†ψ)* χdx
8
We thus conclude that the operator corresponding to a physical quantity must
be self-adjoint A A = A
†We now consider the case in which we have two physical quantities and that
can be simultaneously measured 𝒜 ℬ
In this case the two operators must have common eigenfunctions
The product of the corresponding operators and can be defined as A B
(AB)ψ ≡ A(B ψ)
If we apply the product on one of the common eigenfunctions we have AB ϕ
k(AB)ϕ
k= A(Bϕ
k) = A(b
kϕ
k) = b
k(Aϕ
k) = b
ka
kϕ
kOn the other hand
(BA)ϕ
k= B(Aϕ
k) = B(a
kϕ
k) = a
k(Bϕ
k) = a
kb
kϕ
kWe conclude that on each eigenfunction we have but since each wave-function can be expanded in this basis this must hold in general ϕ
k[A, B] ≡ AB − BA = 0
If two physical quantities can be simultaneously
measured their operators commute
Hilbert spaces
9
The superposition principle and the idea that the wave-function describes the state of a system require the mathematical foundations of QM to be laid
down by using linear spaces of functions with scalar products, which allow us to define normalisation and transition rates: these are the Hilbert spaces
Recall MMP1: Hilbert space: “vector space with scalar product which is complete with respect to the metric induced by the norm”
We consider a vector space on a field ( ℋ 𝕂 𝕂 = ℝ or ) ℂ
A scalar product ⟨ ⋅ | ⋅ ⟩ is a mapping from ℋ × ℋ → 𝕂 such that ( | x⟩, | y⟩, | z⟩ ∈ ℋ ) (i) ⟨x | α y + βz⟩ = α⟨x | y⟩ + β⟨x | z⟩
(ii) ⟨x | y⟩ = ⟨y | x⟩*
(iii) ⟨x | x⟩ ≥ 0 and ⟨x | x⟩ = 0 ⇔ x = 0
α, β ∈ 𝕂
Dirac “ket” notation
Hilbert spaces
10
Given a vector space with scalar product, a norm is defined through ℋ
∥x∥ = ⟨x | x⟩
A vector space with scalar product is a Hilbert space if it is complete with respect to the metric induced by the norm ℋ
This means that every Cauchy sequence is such that ϕ
nlim
n→∞
ϕ
n∈ ℋ
In general dim ℋ = ∞
In Quantum Mechanics we will limit ourselves to consider the case in which the Hilbert space is separable: this means that a complete
numerable orthonormal basis exists {ϕ
n, n = 1,2...}
Examples:
- ℋ = ℂ
n: prototype of complex vector space with finite dimension
- ℋ = L
2[a, b] : space of the square-integrable functions in the interval [a, b]
11
Momentum Operator
In position space the probability to find a particle at the position in the
element is given by dx x
| ψ (x, t) |
2dx
Correspondingly, let us define the probability of finding the particle with momentum between and p w( p p, + t)dp dp
We require
∫ dpw( p, t) = 1
By writing ψ (x, t) =
∫ dp
2π ℏ ψ ˜ ( p, t)e
ipx/ℏwe find
∫ dx | ψ (x, t) |
2= ∫ dx ∫ dp
2π ℏ dp′
2πℏ e
ℏi (p−p′)xψ ˜ ( p, t) ˜ ψ ( p′ , t)*
= ∫ dp
2π ℏ | ψ ˜ ( p, t) |
2Use ∫ dx exp{i / ℏ( p − p′ )x} = 2π ℏδ( p − p′ )
12
This result suggests the following definition for the probability density in momentum space
w( p, t) = 1 2π ℏ | ψ ˜ ( p, t) |
2This is consistent with the idea that for a plane wave with momentum the Fourier transform ψ ˜ ( p, t) differs from zero only for p = p
0p
0We can now compute the expectation value ⟨p⟩ of the momentum as
⟨p⟩ = ∫ dp
2πℏ ψ ˜ *( p, t) p ψ ˜ ( p, t) = ∫ dxdx′ ∫ dp
2πℏ e
ipx′/ℏψ*(x′ , t) p e
−ipx/ℏψ (x, t)
If we write pe
−ipx/ℏψ (x, t) = iℏ ( ∂
∂x e
−ipx/ℏ) ψ (x, t) = − iℏe
−ipx/ℏ∂ ψ (x, t)
∂x
integrating by parts
⟨p⟩ = ∫ dp p w( p, t)
We find
˜
ψ ( p, t) = ∫ dx ψ (x, t)e
−ipx/ℏ˜
ψ *( p, t) = ∫ dx′ ψ *(x′ , t)e
ipx′/ℏuse
13
p = − iℏ ∂
∂x
We thus conclude that the operator corresponding to the momentum is p
This result is easily generalised to the tridimensional case in which the wave function is a function of the position vector x
In this case we have
p = − iℏ∇
⟨ p⟩ = ∫ dxdx′ ∫ dp
2πℏ e
ipx′/ℏψ *(x′ , t)(− iℏ)e
−ipx/ℏ∂ψ (x, t)
∂x = ∫ dxdx′ δ(x − x′ )ψ *(x′ , t)(− iℏ) ∂ψ (x, t)
∂x ψ (x, t)
= ∫ dx ψ *(x, t) (− iℏ) ∂
∂x
Continuous spectrum
We have seen that in the case of an observable with a discrete spectrum of values corresponding to the eigenfunctions a general wave function can be expanded as
a
kϕ
k𝒜
ψ = ∑
k
c
kϕ
kc
k∈ ℂ
Aψ = ∑ and
k
a
kc
kϕ
kAϕ
k= a
kϕ
kAnd the corresponding operator is such that A
The momentum operator is an example of operator with continuous spectrum Limiting ourselves to the one-dimensional case the eigenvalue equation reads
pϕ
p(x) = − iℏ ∂ ∂x ϕ
p(x) = pϕ
p(x)
And the solution is
ϕ
p(x) = (2π ℏ)
−1/2e
ipx/ℏThe functions form a complete orthonormal system as the introduced
for the discrete spectrum ϕ
p(x) ϕ
kThe only difference is that the sum over is replaced by an integral over and that the condition is replaced by
k p
∫ ϕ
lϕ*
kdx = δ
kl∫ ϕ*
p′(x)ϕ
p(x)dx = δ( p − p′ )
and the completeness relation reads
∫ ϕ*
p(x′ )ϕ
p(x)dp = δ(x − x′ )
The expansion of an arbitrary wave function in momentum eigenfunctions is nothing but a Fourier expansion and reads
ψ (x) = 1
2π ℏ ∫ dpc
pϕ
p(x) = 1
2π ℏ ∫ dpc
pe
ipx/ℏor, by using k = p / ℏ
ψ (x) = 1
2π ∫ dkc(k)e
ikx16
The Correspondence principle
We can thus summarise what we have learned so far and connect it with the mathematical structure of a Hilbert space
The wave functions ψ (x, t) become vectors in Hilbert spaces | ψ⟩
The observables become linear hermitian operators on such Hilbert spaces 𝒜 The assignment of the classical measured variables to the quantum
mechanical observables is governed by the correspondence principle:
Position x x
Momentum p − iℏ∇
Energy E iℏ∂ / ∂t
To what extent can we assign quantum mechanical relations to classical
relations on the basis of this correspondence ?
17
For example, if we take the classical energy-momentum relation for a massive particle in a potential V(x)
E = p
22m + V
Does it mean that ? iℏ(∂ / ∂t) = − (ℏ
2/2m) ∇
2+ V
This relation cannot be valid at the operator level, but when applied on the wave function it gives exactly the Schrödinger equation
iℏ(∂ψ / ∂t) = Hψ H = − (ℏ
2/2m) ∇
2+ V(x)
where
is the Hamilton operator
18
Linear operators and commutators
In Quantum Mechanics the relevant operators are linear
| ψ
1⟩, | ψ
2⟩ ∈ H c
1, c
2∈ ℂ A ( c
1| ψ
1⟩ + c
2| ψ
2⟩ ) = c
1A | ψ
1⟩ + c
2A | ψ
2⟩
Examples: x
i, ∂ as multiplier….
∂x
i, ∇
2, ∂
∂t , f (x, t)
We have seen that if two physical quantities can be simultaneously measured the corresponding operators must commute
it is clear that [ x
i, x
j] = 0 Since (x
ix
j− x
jx
i)ψ = 0
Let us consider the commutator
[ x
i, ∂
∂x
j]
The opposite is also true
19
Linear operators and commutators
[ x
i, ∂
∂x
j] = − δ
ijWe thus conclude that We have
[ x
i, ∂
∂x
j] ψ =
( x
i∂
∂x
j− ∂ ∂x
jx
i) ψ = x
i∂ψ
∂x
j− δ
ijψ − x
i∂ψ
∂x
j= − δ
ijψ
Since p
i= − iℏ∇
iwe see that [ x
i, p
j] = iℏδ
ijThis result shows that spatial and
momentum components do not commute
This is consistent with the uncertainty principle! it is not possible to exactly
measure position and momentum at the same time
Measurement process
We consider again a physical observable that characterises the quantum system and is the corresponding linear operator A 𝒜
Let ( ) be the discrete values of and the wave function of the system in the state in which the quantity takes the value a
kk = 0,1,2.... 𝒜 ϕ
k𝒜 a
kψ = ∑
k
c
kϕ
kc
k∈ ℂ
A generic state can be expanded as ψ
And the coefficients give the probability that a measurement of returns the value | c
k|
2
𝒜
a
kAfter the measurement returns one of the values the system will be in the
state ϕ
ka
k| ψ⟩ measurement of with result
𝒜 ak| ϕ
k⟩
This is known as “wave function collapse” (or von Neumann projection)
21
General form of the uncertainty principle
We now want to give the general form of the uncertainty principle
We start from a general property of a vector space with scalar product ℋ Cauchy-Schwarz inequality (recall MMP1 !)
∀ | ϕ⟩, | ψ⟩ ∈ ℋ | ⟨ϕ | ψ⟩ | ≤ ∥ϕ∥∥ψ ∥
We then consider two Hermitian operators and and an arbitrary state A
1A
2ψ We define the operators obtained by subtracting from and their expectation value on the state
̂ A
i= A
i− ⟨A
i⟩ = A
i− ⟨ψ | A
i| ψ⟩
A
1A
2ψ
We use the Cauchy-Schwarz inequality with A
1̂ ψ and A
2̂ ψ
| ⟨ ̂ A
1ψ | A
2̂ ψ⟩ | ≤ ∥ ̂ A
1ψ ∥∥ ̂ A
2ψ ∥
or | ⟨ ̂ A
1ψ | A
2̂ ψ⟩ |
2≤ ⟨ ̂ A
1ψ | A
1̂ ψ⟩⟨ ̂ A
2ψ | A ̂
2ψ⟩
22
Using the fact that and are Hermitian we have A
1A
2| ⟨ψ | A ̂
1A ̂
2ψ⟩ |
2≤ ⟨ψ | A ̂
21| ψ⟩⟨ψ | A
22̂ | ψ⟩
Now, even if the operators and are Hermitian their product is not: indeed A ̂
1A
2̂
( A
1̂ A
2̂ )
†= A
†2̂ A
†1̂ = A ̂
2A ̂
1≠ ̂ A
1A ̂
2We decompose the product A
1̂ A
2̂ into an Hermitian and anti-Hermitian part
̂ A
1̂ A
2= 1 2 ( ̂ A
1̂ A
2− ̂ A
2̂ A
1) + 1 2 ( ̂ A
1̂ A
2+ ̂ A
2̂ A
1) = 1 2 [ ̂ A
1, ̂ A
2] + 1 2 { ̂ A
1, ̂ A
2}
Indeed we have
[ A
1̂ , A
2̂ ]
†= − [ A
1̂ , A
2̂ ] { A
1̂ , A
2̂ }
†= { A ̂
1, A ̂
2}
anti-Hermitian Hermitian
Hermitian and anti-Hermitian operators have real and purely imaginary expectation values
⟨ψ | Aψ⟩* = ⟨Aψ | ψ⟩ = ⟨ψ | Aψ⟩ A Hermitian
⟨ψ | Aψ⟩* = ⟨Aψ | ψ⟩ = − ⟨ψ | Aψ⟩ A anti-Hermitian (3)
use ⟨ψ
1| A
1̂ A
2̂ | ψ
2⟩ = ⟨ ̂ A
†1ψ
1| A
2̂ | ψ
2⟩ = ⟨ ̂ A
†2A
†1̂ ψ
1| ψ
2⟩
23
This means that when computing | ⟨ψ | A
1̂ A
2̂ ψ⟩ |
2we can write
| ⟨ψ | A ̂
1A
2̂ ψ⟩ |
2= 1
4 | ⟨ψ | { A
1̂ , A ̂
2}ψ⟩ |
2+ 1
4 | ⟨ψ | [ A
1̂ , A
2̂ ]ψ⟩ |
2Since expectation values are ordinary numbers we have [ A
1̂ , A
2̂ ] = [A
1, A
2] And from the previous result we conclude that
| ⟨ψ | A
1̂ A ̂
2ψ⟩ |
2≥ 1
4 | ⟨ψ | [A
1, A
2]ψ⟩ |
2(4)
The uncertainty Δ A is defined as the root-mean-square deviation
(Δ A)
2≡ ⟨ψ | (A − ⟨A⟩)
2| ψ⟩
Combining the inequalities and we obtain (3) (4)
Δ A
1Δ A
2≥ 1
2 ⟨[A
1, A
2]⟩
This inequality represents the general form of the Heisenberg uncertainty relation An important special case is
A
1= x
iA
2= p
jΔx
iΔp
j≥ ℏ
2 δ
ij24
Continuity equation
The interpretation of as probability density suggests that a probability current density should exist ρ = | ψ (x, t) |
2
We consider the case in which the Hamiltonian is
H = p
2/2m + V(x)
and we compute the derivative of the probability density with respect to t
∂ρ
∂t = ∂ψ*
∂t ψ + ψ * ∂ψ
∂t
Then we use the Schrödinger equation and its complex conjugate
iℏ(∂ψ / ∂t) = Hψ − iℏ(∂ψ */ ∂t) = (Hψ)*
to find ∂ρ
∂t = i
ℏ ( (Hψ)*ψ − ψ *Hψ )
The potential is real: the term in V(x) cancels out and we obtain
25
Continuity equation
If we define j = − iℏ
2m ( ψ * ∇ψ − ψ ∇ ψ * )
we find the continuity equation ∂ρ
∂t + ∇ ⋅ j = 0
∂ρ
∂t = − iℏ
2m ( ( ∇
2ψ*)ψ − ψ* ∇
2ψ ) = iℏ
2m ∇ ⋅ ( ( ∇ ψ*)ψ − ψ * ∇ ψ )
in integral form it gives
d
dt ∫
Vd
3x ρ(x, t) = − ∫
SdS n ⋅ j(x, t)
The meaning of the continuity equation (as for the charge density in
electromagnetism) is that the dynamics cannot create or destroy particles
26
Stationary states and time evolution
We now go back to the Schrödinger equation
And we assume that the Hamiltonian is independent of time
iℏ ∂ ψ
∂t = Hψ
In this case a solution can be found by separation of variables ψ (x, t) = f (t)ψ (x)
Plugging this ansatz in the Schrödinger equation and dividing by ψ (x, t) we find
1
f (t) iℏ ∂ f (t)
∂t = 1
ψ (x) Hψ
Since the left-hand side depends only on and the right-hand side only on both sides must be equal to a constant that we call t E x Thus for we obtain f (t) iℏ ∂ f (t)
∂t = E f (t) with solution f (t) = e −iEt/ℏ
Recall Lecture 2: this is exactly the same procedure used for the free particle
27
Stationary states and time evolution
Hψ (x) = E ψ (x)
The equation for ψ (x) reads
Which is known as time-independent Schrödinger equation and it is an eigenvalue equation for the operator H
The solutions ψ (x, t) = e
−iEt/ℏψ (x) are called stationary states
Note that the associated probability density ρ(x, t) = | ψ (x) |
2is independent of time The solutions of Eq. provide a useful basis for the Hilbert space (5) {ϕ
k}
(5)
Given an initial condition we can proceed as we did in Lecture 2 for the free particles: we expand it on the orthonormal basis ψ (x, t = 0) = ψ
0(x)
ψ
0(x) = ∑
k
c
kϕ
k(x) with c
k= ⟨ϕ
k| ψ
0⟩ Then the solution of the
Schrödinger equation reads ψ (x, t) = ∑
kc
ke
−iEkt/ℏ
ϕ
k(x)
Hϕ k = E k ϕ k
The Heisenberg picture
In classical mechanics there is only one way to describe the time evolution of a system through the time dependence of the observables
In quantum mechanics we can do it in (at least) two ways:
Suppose, as we did up to now, that the state depends on time, and thus it tells us how the probabilities of of observables measurements evolve
Assume that it is the operators associated to the observables to depend on time, and that the state remains fixed
The first possibility leads to the Schrödinger picture, while the second leads to the Heisenberg picture
What is physically observable are the matrix elements of operators
⟨A⟩
ψ(t) = ⟨ψ
S(t) | A
S| ψ
S(t)⟩
Where we have added to denote that we work in the Schrödinger picture S
28
⟨A⟩
ψ(t) = ⟨ψ
S(t) | A
S| ψ
S(t)⟩ = ⟨ψ
S(t
0) | U
−1(t, t
0)A
SU(t, t
0) | ψ
S(t
0)⟩
We can move to the Heisenberg picture by writing
29
Where the operator U(t, t
0) is the time evolution operator defined as
| ψ
S(t)⟩ = U(t, t
0) | ψ
S(t
0)⟩
And it obeys the Schrödinger equation iℏ ∂ U
∂t = H U
We can thus define the operator in the Heisenberg picture as A
HA
H(t) ≡ U
−1(t, t
0)A
SU(t, t
0)
The matrix element of the operator is clearly independent on which representation is used A
With A
H(t
0) = A
Sand | ψ
H⟩ = | ψ
S(t
0)⟩
30
The Ehrenfest Theorem
We have seen with the path integral approach how classical mechanics is obtained in the limit ℏ → 0
Another perspective is offered by the Ehrenfest theorem
We start from the Schrödinger equation and its complex conjugate
iℏ(∂ψ / ∂t) = Hψ − iℏ(∂ψ */ ∂t) = (Hψ)*
We want to construct the dynamical equation for the expectation value of an operator A
⟨A⟩ = ∫ d
3x ψ *(x, t) A ψ (x, t)
d
dt ⟨A⟩ = ∫ d
3x ( ∂ψ *
∂t A ψ + ψ* ∂A
∂t ψ + ψ * A ∂ψ
∂t )
Taking the derivative with respect to we obtain t
31
dt d ⟨A⟩ = ∫ d
3x ( i
ℏ (Hψ)* A ψ + ψ* ∂A
∂t ψ − i
ℏ ψ * AHψ )
By using the Schrödinger equation and its conjugate
we obtain ∂
tψ = − i / ℏ(Hψ) ∂
tψ* = i / ℏ(Hψ)*
Since is Hermitian we have H
d
dt ⟨A⟩ = i
ℏ ⟨[H, A]⟩ + ⟨ ∂ A
∂t ⟩
We can now compare with what we have in classical mechanics: for a system with conjugated variables and we have q p
d
dt f (q, p, t) = ∂f
∂q q · + ∂f
∂p p · + ∂f
∂t = ∂f
∂q ∂H
∂p − ∂ f
∂p ∂H
∂q + ∂f
∂t = {H, f } + ∂f
∂t
where the Poisson brackets are defined as {g, f } = ∂g
∂p
∂f
∂q − ∂ f
∂p
∂g
∂q
We conclude that we
have the correspondence { f, g} ↔ ℏ i [F, G]
Where the and are
observables and and are the corresponding operators
f g
F G
(6)
use Hamilton equations q · = ∂H / ∂p and p · = − ∂H/ ∂q
We now want to apply the theorem to obtain the time evolution of and ⟨x⟩ ⟨p⟩
We need
[H, x
i] = ∑
j
p
j22m , x
i= 1
2m ∑
j
[ p
j2, x
i] = − iℏ p
im
[H, p
i] = iℏ ∂ V
∂x
id
dt ⟨p
j⟩ = i
ℏ ⟨[H, p
j]⟩ = −
⟨
∂V
∂x
j⟩ d
dt ⟨x
j⟩ = i
ℏ ⟨[H, x
j]⟩ = ⟨p
j⟩ m
These equations have the same form as their classical counterparts
use [A
2, B] = [A, B]A + A[A, B]
The above equations (and more generally Eq. ) are known as the Ehrenfest theorem: “Classical equations hold for expectation values of quantum operators” (6)
Thus we can say that classical mechanics emerges as the limit in which the uncertainty (that is the standard deviation in a given measurement) is small with respect to the typical value of the measured quantity
32
Wave packets
The momentum eigenfunctions
ϕ
p(x) = (2π ℏ)
−1/2e
ipx/ℏAre states of definite momentum, and thus, according to the uncertainty principle, in these states the position is completely undetermined
Indeed by squaring the above expression, we obtain a constant independent on x In realistic situations the momentum is measured with a finite resolution, on a system which is localised over a finite region of space
A realistic free particle is thus in a state which is a
superposition of momentum eigenstates, that is a wave packet The most general wave packet can be constructed as
33
ψ (x, t) = 1 2π ∫ dkc(k)e
i(kx−ωkt)We now focus on a special case of wave packet, the one with minimal uncertainty We require that at t = 0 the product ΔpΔx takes the minimum value
ΔpΔx = ℏ
To construct a wave packet with this property, we go back to the proof of the 2
uncertainty principle (slide 21-23)
| ⟨ψ | A ̂
1A ̂
2ψ⟩ |
2= 1 4 | ⟨ψ | { A
1̂ , A
2̂ }ψ⟩ |
2+ 1 4 | ⟨ψ | [ A
1̂ , A
2̂ ]ψ⟩ |
2≥ 1
4 | ⟨ψ | [ A
1̂ , A
2̂ ]ψ⟩ |
2Defining A
î = A
i− ⟨A
i⟩ = A
i− ⟨ψ | A
i| ψ⟩ we used the two inequalities
| ⟨ψ | A
1̂ A
2̂ ψ⟩ |
2≤ ⟨ψ | A
21̂ | ψ⟩⟨ψ | A ̂
22| ψ⟩ (3)
and
(4)
We have to look for the situation in which the two inequalities become equalities
34
As for Eq. (4) it becomes an equality if
⟨ψ | { A
1̂ , A
2̂ } | ψ⟩ = 2Re⟨ψ | A ̂
1A ̂
2| ψ⟩ = 2Re⟨ ̂ A
1ψ | A ̂
2ψ⟩ = 0
| ⟨ ̂ A
1ψ | A
2̂ ψ⟩ |
2≤ ⟨ ̂ A
1ψ | A ̂
1ψ⟩⟨ ̂ A
2ψ | A ̂
2ψ⟩
As for Eq.(3) we can rewrite it as (since the operators are Hermitian)
Which tells us that the equality holds if the vectors and are
proportional A
1| ψ⟩ A
2| ψ⟩
| A
1̂ ψ⟩ = c | A ̂
2ψ⟩
Combining the two conditions we conclude that the coefficient must be purely
imaginary c
c = iλ
c ∈ ℂ
λ ∈ ℝ
If we now set A
1= p and A
2= x we find
( p − ⟨ p⟩ ) | ψ⟩ = i λ ( x − ⟨x⟩ ) | ψ⟩
or −iℏ ∂ ψ (x)
∂x − p
0ψ (x) = iλ(x − x
0)ψ (x) x
0= ⟨x⟩, p
0= ⟨ p⟩
35
∂ψ (x)
∂x = i
ℏ p
0ψ (x) − λ
ℏ (x − x
0)ψ (x)
This is a first order differential equation which can be rewritten as
whose solution is
ψ (x) = C exp { ip
0x
ℏ − λ
2ℏ (x − x
0)
2}
We thus conclude that the state of minimum uncertainty is a Gaussian centred in and modulated by a plane wave which determines the expectation value of the
momentum
x
0p
0By normalising the wave function to unity we obtain C = ( λ
π ℏ )
1/4
36
We can now check that this is indeed a wave packet with minimum uncertainty Position uncertainty
We first check that ⟨x⟩ = x
0We have
⟨x − x
0⟩ = ∫
+∞
−∞
dx(x − x
0) | C |
2exp { − λ
ℏ (x − x
0)
2} = 0 ⟨x⟩ = x
0because the function is odd
Δ
2x = ⟨(x − x
0)
2⟩ = ∫
+∞
−∞
dx(x − x
0)
2| C |
2exp { − λ
ℏ (x − x
0)
2}
The position uncertainty is
= ( λ ℏπ )
1/2
π
2 ( λ ℏ )
−3/2
Use ∫
+∞−∞
y
2dye
−ay2= π
2 a
−3/2= ℏ 2λ
Momentum uncertainty
We first check that ⟨p⟩ = p
0⟨ p⟩ = ∫
+∞
−∞
dx ψ *(x)(− iℏ) ∂ψ (x)
∂x = ∫
+∞
−∞
dx ψ *(x)(− iℏ) ( ip
0ℏ − λ
2ℏ 2(x − x
0) ) ψ (x) = p
0odd
The momentum uncertainty is
Δ
2p = ⟨( p − p
0)
2⟩ = ∫
+∞
−∞
dx ψ*(x) ( − iℏ ∂
∂x − p
0) ( − iℏ ∂
∂x − p
0) ψ (x)
= ∫
+∞
−∞
dx ψ *(x) ( − iℏ ∂ ∂x − p
0) (− iℏ) ( − λ
ℏ (x − x
0) ) ψ (x)
37
Δ
2p = ℏλ
2 Δ
2x = ℏ
2λ so that indeed ΔpΔx = ℏ
2
ψ (x) = C exp { ip
0x
ℏ − (x − x
0)
24Δ
2x }
Expressing in terms of λ Δ
2x we can write ψ (x) as
and we see that really determines the width of the Gaussian Δx
= ∫
+∞
−∞
dx ψ *(x) ( − iℏ ∂ ∂x − p
0) (− iℏ) ( − λ
ℏ (x − x
0) ) ψ (x)
= ℏλ − λ
2∫
+∞
−∞
dx | C |
2(x − x
0)
2exp { − λ
ℏ (x − x
0)
2}
= ℏλ − λ
2( λ ℏπ )
1/2
π
2 ( λ ℏ )
−3/2
= ℏλ
2 Using as before ∫
+∞−∞
y
2dye
−ay2= π
2 a
−3/2In summary we have
38
odd
Dropping as before an
odd term in (x − x
0)
ψ (x, t) = 1 2π ∫ dkc(k)e
i(kx−ωkt)A general wave packet is
Each of the energy eigenstates moves with momentum and phase velocity given
by ℏk
v
k= ℏk / m
The wave packet, however, can be considered as a single object, which moves with a single velocity, known as group velocity
v
g= d
dt ⟨x⟩
The discussion of the Ehrenfest theorem shows that
v
g= d
dt ⟨x⟩ = ⟨ p⟩
m
Such velocity can be interpreted as the velocity of the “centre” of the packet Phase and group velocity
39
Our gaussian wave packet moves
with group velocity v
g= p
0/ m
Time evolution of a Gaussian packet
We have seen that the wave packet has minimum
uncertainty, where we have set the centre of the packet for simplicity and
ψ (x) = C exp { ip
0x
ℏ − x
24Δ
2x } x
0= 0
We now want to study how the packet evolves with time
We can do it by using the free particle propagator (recall Lecture 1)
K(x, t, y,0) = m
2π iℏt exp { im(x − y)
22ℏt }
ψ (x, t) = ∫ dy K(x, t, y,0) ψ ( y,0)
= C ∫ dy m
2π iℏt exp { im(x − y)
22ℏt + ip
0y
ℏ − y
24Δ
2x } C = ( λ
πℏ )
1/4
= ( 2πΔ
2x )
−1/440
σ (t) = σ + iℏt 2m
The integral can be computed by using the standard trick of rewriting the exponent as
imx
22ℏt − ( 1
4Δ
2x − im
2ℏt ) ( (y − y
0)
2− y
02) y
0= i 2ℏ
p
0−
mxt4Δ12x
−
2ℏtimGaussian integral
After some manipulations the result can be written in the following form
ψ (x, t) = ( σ
2πσ
2(t) )
1/4
exp { i
ℏ ( p
0x − p
022m t
) } exp { − (x − p
0t / m)
24σ (t) }
where σ = Δ
2x This is still a Gaussian packet but:
the momentum uncertainty remains constant: Δ
2p(t) = Δ
2p(0) (this could also be expected from the Ehrenfest theorem) The center of the packet moves with velocity p
0/ m
The position uncertainty increases with time Δ
2x(t) = Δ
2x
( 1 + ℏ
2t
24m
2Δ
4x )
41
Wrap up: the postulates of QM
1) The state of a system is represented by a vector (ket) | ψ⟩ in a Hilbert space ℋ The specific Hilbert space depends on the system we consider (for example, for a one-dimensional particle in the interval the wave function
is a vector in the space of the square-integrable functions [a, b] )
ψ (x) ↔ | ψ⟩ L
2[a, b]
The state vector | ψ⟩ ∈ ℋ contains all the information of the system is the norm while is the scalar product in
∥ψ ∥ = ⟨ψ | ψ⟩ = 1 ⟨ ⋅ | ⋅ ⟩ ℋ
The phase of | ψ⟩ is not observable
42
We can now summarise what we have learnt and put it in the form of
postulates of the quantum theory
2) Every observable (position, momentum, angular momentum, energy…) corresponds to a self-adjoint (for us equivalent to hermitian) operator 𝒜 A
The adjoint operator is defined as A
†A self-adjoint operator is such that A = A
†⟨ψ | A χ⟩ = ⟨A
†ψ | χ⟩ ≡ ⟨ψ | A | χ⟩
and ⟨ψ | A | χ⟩* = ⟨ χ | A
†| ψ⟩
For a self-adjoint operator we can write a spectral decomposition
A = ∑
Nk=1