Quantum Information

(ELEC-C9440)
Lecture 3

Lauri Ylinen1

Aalto University

Spring 2024

1
Slides courtesy of Arttu Pönni, Matti Raasakka, with some modifications by LY
Entanglement, entanglement entropy
Entanglement

Entanglement is the fundamental property that distinguishes quantum

mechanics from classical theories of physics and its presence is thus
responsible for many of the strange quantum properties of nature.
Entanglement causes different parts of a system to behave as one in a
fundamental way, even when the parts are not interacting with one
another. Classical intuition is that one should be able to study the
behavior of the non-interacting parts separately and thus understand
the behavior of the full system. This intuition fails if the parts of the
system are entangled, regardless of the parts not interacting with one
another or even being separated by a vast distance.

Entanglement is used as a resource in quantum information. We have

already seen this: superdense coding was possible because we used this
quantum resource. Another example would be quantum teleportation in
which we use pre-shared entanglement and classical communication to
teleport a qubit.
Product, separable and entangled states
Which states have entanglement and which do not? The following
classification of states is useful.
Definition (Product state)
A product state is a tensor product of two or more density operators.
The state |ψ⟩ = |00⟩, ρ = |00⟩⟨00| = |0⟩⟨0| ⊗ |0⟩⟨0| is a product state
because |0⟩⟨0| is a density operator.
Definition (Separable state)
A state ρAB is separable if it can be written as an ensemble of product
P (i) (i) P
states: ρAB = i pi ρA ⊗ ρB , where pi ≥ 0 and i pi = 1.
The state ρAB = 21 |0⟩⟨0| ⊗ |1⟩⟨1| + 12 |+⟩⟨+| ⊗ |−⟩⟨−| would be an
example of a separable state because each of the factors is a density
operator. Note that a product state is also separable.
Definition (Entangled state)
A state ρAB is entangled if it is not separable.
A prominent example of an entangled state is the Bell state
1 1
|ψ⟩ = √ (|00⟩ + |11⟩) → ρAB = (|00⟩⟨00| + |00⟩⟨11| + |11⟩⟨00| + |11⟩⟨11|) .
2 2
Schmidt decomposition
There are many different ways to detect and quantify quantum
entanglement. One such way is given by the Schmidt number which
uses the widely useful Schmidt decomposition.
Theorem (Schmidt decomposition)
Let |ψ⟩ ∈ HAB be a pure state on a bipartite quantum system AB.
Then there exist orthonormal states {|iA ⟩} and {|iB ⟩} for A and B,
respectively, such that
X
|ψ⟩ = λi |iA ⟩ |iB ⟩ ,
i

2
P
where λi ≥ 0 and i λi = 1. The coefficients λi are the Schmidt
coefficients.
One nice thing about this decomposition is that the reduced density
matrices
X X
ρA = trB |ψ⟩⟨ψ| = λ2i |iA ⟩⟨iA | , ρB = λ2i |iB ⟩⟨iB |
i i

have the same eigenvalues! Therefore any functions on reduced density

matrices which depend only on their eigenvalues give the same result
for ρA
and ρB .
For example, if a bipartite state is pure |ψ⟩, then
tr ρ2A = tr ρ2B , tr(ρA log ρA ) = tr(ρB log ρB ) etc.
Schmidt decomposition

Let’s prove the theorem. Any pure bipartite state can be written
X X
|ψ⟩ = ajk |j⟩ |k⟩ = uji dii vik |j⟩ |k⟩ ,
j,k i,j,k

where {|j⟩} and {|k⟩} are orthonormal bases on A and B, respectively.

The second step uses the singular value decomposition of the matrix a:
a = udv , where u and v are unitary matrices andP d is a rectangular
diagonal matrix. Then we define dii ≡ λi , |iA ⟩ ≡ j uji |j⟩, and
P
|iB ⟩ ≡ k vik |k⟩:
  !
X X X X
|ψ⟩ = dii  uji |j⟩ vik |k⟩ = λi |iA ⟩ |iB ⟩ ,
i j k i

which proves the theorem.

Schmidt decomposition and entanglement

The Schmidt number or Schmidt rank is the number of non-zero

Schmidt coefficients λi in the decomposition of a state |ψ⟩. A pure
state is a product state if its Schmidt number is exactly one, otherwise
it is entangled.
Example: If the Schmidt number is one, then we can write

|ψ⟩ = |0A ⟩ |0B ⟩ ∈ HAB ,

for some states |0A ⟩ ∈ HA and |0B ⟩ ∈ HB . This state is clearly a

product state and therefore not entangled.
Example: If we have the Schmidt decomposition
1 1
|ψ⟩ = √ |0⟩ |0⟩ + √ |1⟩ |1⟩ ,
2 2
the Schmidt number is 2 and the state is entangled.
Entanglement entropy
The Schmidt number gives a very crude characterization of
entanglement. A widely used quantity for measuring the amount of
entanglement between subsystems of a quantum state is the von
Neumann entropy. It is often called simply entanglement entropy.
Definition (Entanglement entropy)
The entanglement entropy S(ρ) of a quantum state ρ is
X
S(ρ) = − tr (ρ log ρ) = − pi log pi ,
i

where {pi } are the eigenvalues of ρ.

We will take the logarithm to be in base 2. A few basic properties of
entanglement entropy:
▶ Non-negative: S(ρ) ≥ 0, equality iff ρ is pure.
▶ Invariant under unitary transformations: S(UρU † ) = S(ρ).
▶ S(ρ) ≤ log d, where d = dim(H). Equality iff ρ = I/d.
▶ If ρAB is pure, then S(ρA ) = S(ρB ).
▶ Subadditivity: S(ρAB ) ≤ S(ρA ) + S(ρB ) with equality iff
ρAB = ρA ⊗ ρB .
▶ Araki-Lieb inequality: S(ρAB ) ≥ |S(ρA ) − S(ρB )|.
Entanglement entropy

P
Let’s show that S(ρ) = − P i pi log pi . With the spectral decomposition
we can write any state ρ = i pi |ψi ⟩⟨ψi |. Then
X
log ρ = log pi |ψi ⟩⟨ψi |
i
X X
→ ρ log ρ = pj ψj ψj log pi |ψi ⟩⟨ψi | = pj log pi ψj ψj ψi ⟨ψi |
i,j i,j

X X
= pj log pi ψj δji ⟨ψi | = pi log pi |ψi ⟩⟨ψi |
i,j i
X X
− tr (ρ log ρ) = − pi log pi tr |ψi ⟩⟨ψi | = − pi log pi = S(ρ) .
i i

The eigenvalue expression shows explicitly some of the listed basic

properties: non-negativity because 0 ≤ pi ≤ 1, 2 invariance under
unitary transformations (these do not change eigenvalues), S(ρ) = 0 for
pure states (one pi = 1, others zero).

2
We always define 0 log 0 ≡ limx→0 x log x = 0
Example: Bell state entanglement
√
Example: Consider again the Bell state |ψ⟩ = (|00⟩ + |11⟩)/ 2 ∈ HAB .
We already know this is an entangled state because the Schmidt
number is 2. Let’s study the entanglement between subsystems A and
B by computing S(ρA ) and S(ρB ). The reduced density matrix ρA is

1
|ψ⟩⟨ψ| = (|00⟩⟨00| + |00⟩⟨11| + |11⟩⟨00| + |11⟩⟨11|)
2
1
→ trB |ψ⟩⟨ψ| = (|0⟩⟨0| ⟨0|0⟩ + |0⟩⟨1| ⟨0|1⟩ + |1⟩⟨0| ⟨1|0⟩ + |1⟩⟨1| ⟨1|1⟩)
2

1
= (|0⟩⟨0| + |1⟩⟨1|) = ρA .
2
We see that the eigenvalues of ρA are p1 = 1/2 and p2 = 1/2.
Therefore the entanglement entropy is
1 1 1 1 1 1
S(ρA ) = − log − log = − (−1) − (−1) = 1 .
2 2 2 2 2 2
Because the joint state |ψ⟩⟨ψ| is pure, we know that S(ρA ) = S(ρB ).
This also follows from the Araki-Lieb inequality and S(|ψ⟩⟨ψ|) = 0. We
notice that the subsystems A and B are maximally entangled.
Distance measures for quantum states
How close are two quantum states?

A natural question in the study of quantum information is: what does it

mean that quantum information is preserved in some physical process?
This question arises for example in the following situations:
▶ We store a quantum state in some “quantum memory” to be
retrieved and used later. How to determine whether the
information has degraded and to what extent?
▶ Alice and Bob communicate through a noisy quantum channel.
How close is the message received by Bob and the message sent by
Alice?
▶ We run a program on a noisy quantum computer. Suppose we
apply a gate on some quantum state. What is the “distance”
between the actual output state and the theoretically correct
output state?
In order to study such questions, we need to develop a distance measure
which tells us how similar/close two quantum states are to each other.
We will introduce the two most important distance measures:
▶ Trace distance
▶ Fidelity
Trace distance for classical distributions

We have seen that density operators can be interpreted as ensembles of

quantum states: each state occurring with some probability. In order to
gain intuition about distance measures, let’s first consider classical
probability distributions

{px } and {qx } ,

over the same index set x. There is no unique way to describe their
similarity but the trace distance D(px , qx ) is widely used

1X
D(px , qx ) ≡ |px − qx | .
2 x

This is a sensible metric on probability distributions because it is

1. Symmetric: D(px , qx ) = D(qx , px )
2. Satisfies the triangle inequality: D(px , qx ) ≤ D(px , rx ) + D(rx , qx )
We also see that it is never negative and zero if and only if px = qx , so
calling it a “distance” is reasonable.
Trace distance for classical distributions

Example: What is the trace distance between the probability

distributions (1/2, 1/3, 1/6) and (3/4, 1/8, 1/8)?
 
1 1 3 1 1 1 1
D(px , qx ) = − + − + −
2 2 4 3 8 6 8
 
1 1 5 1 1
= + + = .
2 4 24 24 4

Example: What is the trace distance between the probability

distributions (p, 1 − p) and (q, 1 − q)?

1
D(px , qx ) =(|p − q| + |(1 − p) − (1 − q)|)
2
1
= (|p − q| + | − p + q|) = |p − q| .
2
Fidelity for classical distributions

The other useful distance measure in quantum information is the

fidelity. Let’s again first introduce it for classical probability
distributions. The fidelity F (px , qx ) is
X√
F (px , qx ) ≡ px qx .
x

Note that this has quite differ-

ent properties compared to trace
distance: if {px } and {qx } are
the psame, then F (px , px ) =
px2 = 1.
P
x Interpretation:
F (px , qx ) is the inner product be-
tween two vectors with compo-
√ √
nents px and qx .
Example: The fidelity of proba-
bility distributions (1/2, 1/3, 1/6)
and (3/4, 1/8, 1/8) is
p p
F (px , qx ) = 3/8 + 1/24
p
+ 1/48 ≈ 0.96 .
Trace distance for quantum states
We’ll now extend the classical trace distance for quantum states. The
trace distance between two quantum states ρ and σ is defined to be
1
D(ρ, σ) ≡ tr |ρ − σ| ,
2
√
where the absolute value of a matrix is defined |A| ≡ A† A. Lets first
see how this reduces to the classical case. Consider a situation where ρ
and σ commute [ρ, σ] = 0, which implies that they have the same
eigenstates. Then
X X
ρ= ri |i⟩⟨i| , σ = si |i⟩⟨i|
i i
X
→ρ−σ = (ri − si ) |i⟩⟨i|
i
X
†
→ (ρ − σ) (ρ − σ) = (ri − si )2 |i⟩⟨i|
i
q X
→ (ρ − σ)† (ρ − σ) = |(ri − si )| |i⟩⟨i|
i
1 1X
q
→ tr (ρ − σ)† (ρ − σ) = |(ri − si )| = D(ri , si ) .
2 2 i
Example: computing the trace distance

Example: Find the trace distance between the density operators

ρ = |0⟩⟨0| and σ = |+⟩⟨+|? Let’s do this explicitly in component form3
   
1 0 1 1 1
ρ= σ=
0 0 2 1 1
   
1 1 −1 1 1 0
→ρ−σ = → (ρ − σ)† (ρ − σ) =
2 −1 −1 2 0 1
 
1 1 0
q
|ρ − σ| = (ρ − σ)† (ρ − σ) = √
2 0 1
1 1
→ D(ρ, σ) = tr |ρ − σ| = √
2 2

3 √
Remember that |±⟩ = (|0⟩ ± |1⟩)/ 2.
Fidelity for quantum states
Let’s now define the quantum version of fidelity. The fidelity F (ρ, σ) is
defined
q
F (ρ, σ) ≡ tr ρ1/2 σρ1/2 .

One can show that this is symmetric F (ρ, σ) = F (σ, ρ), invariant under
unitary transformations F (UρU † , UσU † ) = F (ρ, σ), and
0 ≤ F (ρ, σ) ≤ 1. We’ll start
P by checking that Pthis reduces to the
classical case when ρ = i ri |i⟩⟨i| and σ = i si |i⟩⟨i|:
sX !
X√
F (ρ, σ) = tr ri si |i⟩⟨i| = tr ri si |i⟩⟨i|
i i
X√ X√
= ri si tr |i⟩⟨i| = ri si = F (ri , si ) .
i i

We can see from the definition that in general F (ρ, ρ) = 1. Another

useful relation for F (ρ, σ) we can see from the definition is the case
where one of the states is pure
p
F (|ψ⟩⟨ψ| , ρ) = tr |ψ⟩⟨ψ| ρ |ψ⟩⟨ψ|
p p
= tr ⟨ψ|ρ|ψ⟩ |ψ⟩⟨ψ| = ⟨ψ|ρ|ψ⟩ .
Example: computing the fidelity

Example: Find the fidelity between ρ = 34 |0⟩⟨0| + 1

4
|−⟩⟨−| and
σ = |+⟩⟨+|.
 
1 7 −1
ρ=
8 −1 1
p
We could directly use the definition F (ρ, σ) = tr ρ1/2 σρ1/2 but we
avoid somepwork by noticing that σ is pure. Then the fidelity simplifies
F (ρ, σ) = ⟨+|ρ|+⟩.
  
1
7 −1 1 6
⟨+|ρ|+⟩ = 1 1 =
16 −1 1 1 16
r
1 3
→ F (ρ, σ) = .
2 2
Relations between trace distance and fidelity
Similar states have D(ρ, σ) ≈ 0 and F (ρ, σ) ≈ 1. For pure states, these
are effectively equivalent. Consider two pure states |ψ⟩ and |ϕ⟩. We
can always find basis vectors |a⟩ and |b⟩ so that

θ θ
|ψ⟩ = |a⟩ , |ϕ⟩ = cos |a⟩ + sin |b⟩ .
2 2
Then
  2    
1 1 − cos θ2 − sin cos θ2
θ
D(|ψ⟩ , |ϕ⟩) = tr      2
 2  = sin θ
2 − sin θ2 cos θ2 − sin θ 2
2

p θ
F (|ψ⟩ , |ϕ⟩) = ⟨ϕ|ψ⟩ ⟨ψ|ϕ⟩ = cos
2
p
Therefore we see that for pure states D(|ψ⟩ , |ϕ⟩) = 1 − F (|ψ⟩ , |ϕ⟩)2 .
More generally, one can show the following statement about the
qualitative similarity for general states ρ and σ
q
1 − F (ρ, σ) ≤ D(ρ, σ) ≤ 1 − F (ρ, σ)2 .