Department of Physics Indian Institute of Technology Madras

PH5842: Adv. QCQI 2025 — Assignment 2 Solutions

Q1: Fidelity

(i) F(ρ, σ) = 1 when ρ = σ

If ρ = σ, we can express both states in their eigenbasis. Let’s assume ρ is diago-
nalized as: X
ρ= λi |i⟩⟨i|
i

Then, since ρ = σ:
√ √
q p
F (ρ, σ) = Tr ρρ ρ = Tr ρ2 = Tr ρ = 1

This is because ρ is a density matrix, and its trace is 1.

F(ρ, σ) = 0 when ρ and σ are Orthogonal
If ρ and σ have support on orthogonal vector spaces, we can choose a basis where
both are diagonal, but their supports do not overlap. Let’s denote the eigenvectors
of ρ as |i⟩ and those of σ as |j⟩, with ⟨i|j⟩ = 0 for all i, j.
In this basis: X X
ρ= λi |i⟩⟨i|, σ= µj |j⟩⟨j|
i j

Since ⟨i|j⟩ = 0, we have:

√ √ Xp p
ρσ ρ = λi λi λi µj |i⟩⟨i|j⟩⟨j| = 0
i,j

√
Thus: F (ρ, σ) = Tr 0=0

(ii) To show that the fidelity between two quantum states ρ and σ is symmetric, we
use the definition:
√ √
q
F (ρ, σ) = Tr ρσ ρ

We can express this in terms of the trace norm:

q √
√
F (ρ, σ) = ∥ σ ρ∥1

However, to show symmetry directly, we need to prove that:

√ √ √ √
∥ σ ρ∥1 = ∥ ρ σ∥1

1
 This follows from the property of the trace norm that: ∥AB∥1 = ∥BA∥1
Thus, we have:
q √
√ √ √
q
F (ρ, σ) = ∥ σ ρ∥1 = ∥ ρ σ∥1 = F (σ, ρ)

This shows that the fidelity is symmetric in its arguments.

(iii) For a pair of qubit state ρ, σ withp√corresponding Bloch vectors ⃗r, ⃗s the fidelity
2 √ 2
square is given by F (ρ, σ) = (T r ρσ ρ)
√ √
Here, ρσ ρ = M > 0 and, eigenvalues of M be λ1 and λ2

T r(M ) = T r(ρσ) = λ1 + λ2
det(M ) = det(ρ) det(σ) = λ1 λ2
q
√ √ √ p p p p
∴ F 2 (ρ, σ) = (T r ρσ ρ)2 = (T r M )2 = ( λ1 + λ2 )2 = λ1 + λ2 + 2 λ1 λ2
p p
= T r(M ) + 2 det(M ) = T r(ρσ) + 2 det(ρ)det(σ)

I + ⃗r · ⃗σ I + ⃗s · ⃗σ
Now, ρ = and σ =
2 2
1
ρσ = (I + (⃗r + ⃗s) · ⃗σ + (⃗r · ⃗σ )(⃗s · ⃗σ ))
4
1 1
∴ T r(ρσ) = (2 + 0 + ⃗r · ⃗s) = (1 + ⃗r · ⃗s)
4 2
1 − |⃗r|2 1 − |⃗s|2
det(ρ) = and det(σ) =
4 4
p 1p
∴ 2 det(ρ)det(σ) = (1 − |⃗r|2 )(1 − |⃗s|2 )
2
1 p
F 2 (ρ, σ) = (1 + ⃗r · ⃗s + (1 − |⃗r|2 )(1 − |⃗s|2 )) ✓
2

Q2: Worst-case fidelity

(i) The state of the qubit after being sent through the channel without using the
error-correcting code is given by: ρ = (1 − p)|ψ⟩⟨ψ| + pX|ψ⟩⟨ψ|X.
p p
The fidelity is given by: F = ⟨ψ|ρ|ψ⟩ = (1 − p) + p⟨ψ|X|ψ⟩⟨ψ|X|ψ⟩.
The second term under the square root is non-negative and equals zero when
√
|ψ⟩ = |0⟩, so the minimum fidelity is: F = 1 − p.
Suppose the three-qubit error-correcting code is used to protect the state |ψ⟩ =
a|0L ⟩ + b|1L ⟩. The quantum state after both noise and error correction is:

ρ = [(1 − p)3 + 3p(1 − p)2 ]|ψ⟩⟨ψ| + · · · .

2
 The omitted terms represent contributions from bit flips on two or three qubits.
All omitted terms are positive operators, so the fidelity we calculate will be a lower
bound on the true fidelity. We see that:
p p
F = ⟨ψ|ρ|ψ⟩ ≥ (1 − p)3 + 3p(1 − p)2 .

That is, the fidelity is at least:

p
F = 1 − 3p2 + 2p3 ,

so the fidelity of storage for the quantum state is improved provided p < 1/2,
which is the same conclusion we came to earlier based on a much cruder analysis.

(ii) The state of the qubit after being sent through the channel without using the
error-correcting code is given by: ρ = (1 − p)|ψ⟩⟨ψ| + pI/2.
p p p
The fidelity is given by: F = ⟨ψ|ρ|ψ⟩ = (1 − p) + p/2 = 1 − p/2.
Suppose the three-qubit error-correcting code is used to protect the state |ψ⟩ =
a|0L ⟩ + b|1L ⟩. The quantum state after both noise and error correction is:

ρ = [(1 − p)9 + 9p(1 − p)8 ]|ψ⟩⟨ψ| + · · · .

The omitted terms represent contributions from bit flips on two or three qubits.
All omitted terms are positive operators, so the fidelity we calculate will be a lower
bound on the true fidelity. We see that:
p p
F = ⟨ψ|ρ|ψ⟩ ≥ (1 − p)9 + 9p(1 − p)8 .

That is, the fidelity is at least:

p
F = 1 − 36p2 + 168p3 + ... + 8p9 ,

so the fidelity of storage for the quantum state is improved provided p < 1/2,
which is the same conclusion we came to earlier based on a much cruder analysis.

Q3: Dual-rail code:

Consider a two-qubit state of the form, |ψ⟩ = a|01⟩ + b|10⟩, with |a|2 + |b|2 = 1. We
need to show that the action of the two-qubit amplitude damping channel on this state
can be represented by two Kraus operators, of the form:
p √ √
Ẽ0 = 1 − γI; Ẽ1 = γ(|00⟩⟨01|); Ẽ2 = γ(|00⟩⟨10|).

To derive these operators, we consider how the amplitude damping channel affects
the two-qubit state. The channel causes transitions from |1⟩ to |0⟩ with probability p,
but does not affect |0⟩. Thus, for a two-qubit system, we can simplify the action of the
channel as follows:

3
 - Ẽ0 represents no damping occurring on either qubit, which is equivalent to the
√
identity operator I scaled by 1 − γ, where γ is related to the damping probability p.
- Ẽ1 and Ẽ2 represent damping occurring on the first and second qubits, respectively,
causing transitions from |01⟩ and |10⟩ to |00⟩.
The action of these operators on |ψ⟩ results in:
p √ √
Ẽ0 |ψ⟩ = 1 − γ|ψ⟩, Ẽ1 |ψ⟩ = γa|00⟩, Ẽ2 |ψ⟩ = γb|00⟩.

Both Ẽ1 |ψ⟩ and Ẽ2 |ψ⟩ result in states that are orthogonal to the original states |01⟩
and |10⟩ since |00⟩ is orthogonal to both |01⟩ and |10⟩. Therefore, any error caused by
amplitude damping can be detected by measuring whether the state has collapsed to
|00⟩, which is outside the original codespace.
This demonstrates that the span of {|01⟩, |10⟩} forms an error detection code for
amplitude-damping noise.

Q4: Conditions for error correction and error detection

ε ∼ {Ea }N
a=1 is correctable by code C with PC

(i) for, m-dim QEC code C with codewords {|iL ⟩}m−1 i=0 , then QEC conditions are sat-
isfied
iff ⟨iL |Ea† Eb |jL ⟩ = cab δij ∀ i,j ∈ [o, m − 1] (1)
Proof: X
PC = |kL ⟩⟨kL | ,where |kL ⟩ ∈ C
k

PC Ea† Eb PC = cab PC
X X
|kL ⟩⟨kL |Ea† Eb |lL ⟩⟨lL | = cab |mL ⟩⟨mL |
k,l m
! !
X X
⟨iL | |kL ⟩⟨kL |Ea† Eb |lL ⟩⟨lL | |jL ⟩ = ⟨iL | cab |mL ⟩⟨mL | |jL ⟩
k,l m
X X
⟨iL |kL ⟩⟨kL |Ea† Eb |lL ⟩⟨lL |jL ⟩ = cab ⟨iL |mL ⟩⟨mL |jL ⟩ = cab ⟨iL |jL ⟩
k,l m

⟨iL |Ea† Eb |jL ⟩ = cab δij ✓

Here, action of Ea on |iL ⟩ leads to a subspace say, C1 and action of Eb on |jL ⟩ leads
to a subspace say, C2 , (s.t) C1 and C2 are non-overlapping subspaces.

(ii) Ea is correctable by code C if the action of the error leads to a state in a subspace
orthogonal to the codespace. {Ea }N a=1 are detectable by code C with codewords
{|iL ⟩} if they satisfy

⟨iL |Ea |jL ⟩ = αa δij ∀ i, j ∈ [0, m − 1] (2)

4
 Proof: Using eq (1), let Eb = I, i.e., the action of Eb on |iL ⟩ identity, remains
in codespace. The action of Ea on |jL ⟩ leads to a subspace orthogonal to the
codespace.
∴ eq(1) → ⟨iL |Ea |jL ⟩ = αa δij ∀ i, j ∈ [0, m − 1]

(iii) Using eq. (1) we say that QEC code can correct t errors if the set of ε ∼ {Ea }N
a=1
of weight upto t, provided that d ≥ 2t + 1
∴ QEC code with distance d = 2t + 1 can correct for errors on at most t qubits.
d =min(weight of Ea ) s.t ⟨iL |Ea |jL ⟩ ≠ αa δij
Using eq(2), ⟨iL |Ea |jL ⟩ = αa δij ∀ Ea of weight upto t
The action of Ea on |iL ⟩: Ea |iL ⟩ = αa |iL ⟩ + |ϕ⊥
ai ⟩
the first term is the same codeword on codespace, and the second term is an
unnormalized orthogonal vector to codespace.
Therefore, the action on a |ϕ⟩ in the codespace of an error Ea with support on ε
!
X X
|ψ⟩ ⊗ |0⟩A → Ea |ψ⟩ ⊗ |ea ⟩A = |ψ⟩ ⊗ Ca |ea ⟩A + |orthog⟩
Ea ∈ε Ea ∈ε

where, |0⟩A is ancilla and {|ea ⟩A } basis states of ancilla after error. |orthog⟩ denotes
a vector orthogonal to the codespace.
Now, we can perform an orthogonal measurement on the ancilla, with two out-
comes: the state is projected onto either the code subspace or the |orthog⟩ sub-
spaces. if the first outcome is obtained, the undamaged state |ψ⟩ is recovered
(correctable and detectable). If the second outcome is found, an error has been
detected.
We conclude that our QECC that can correct t errors can detect 2t errors since
correctable is detectable too.

Q5: QEC conditions in the stabilizer framework:

Single-Qubit Errors The single-qubit Pauli errors are: - Ii : Identity (no error), - Xi :
Bit-flip error on qubit i, - Zi : Phase-flip error on qubit i, - Yi = iXZi : Combined bit-flip
and phase-flip error on qubit i.
For each error operator Ea ∈ {Ii , Xi , Yi , Zi }, we need to check whether the QEC
conditions are satisfied.
(i) We are tasked with showing that the [5, 1, 3] stabilizer code, defined by the stabi-
lizer generators:
⟨XZZXI, IXZZX, XIXZZ, ZXIXZ⟩,
satisfies the quantum error correction (QEC) conditions for the set of single-qubit
error operators {I, Xi , Yi , Zi }, where i = 1, 2, . . . , 5 labels the five encoding qubits.
These generators satisfy the following properties: 1. They commute with one
another: [Si , Sj ] = 0 ∀i, j, 2. They define a codespace as the simultaneous +1
eigenspace of all stabilizers.

5
 The logical operators for this code are: X̄ = XXXXX, Z̄ = ZZZZZ.
The distance of this code is d = 3, meaning it can correct any single-qubit error
(t = ⌊(d − 1)/2⌋ = 1).
Conclusion: The [5, 1, 3] stabilizer code satisfies the QEC conditions for single-
qubit errors because: - The stabilizers either commute or anticommute with all
single-qubit Pauli errors. - Each single-qubit error produces a unique syndrome
that allows for its detection and correction.
Thus, the code can correct arbitrary single-qubit errors as required.

4. Verifying QEC Conditions

1. Error Detection: - For any pair of errors Ea , Eb ∈ {Ii , Xi , Yi , Zi }, their product
Ea† Eb : - Either commutes or anticommutes with the stabilizers. - If it commutes with
all stabilizers (Ea† Eb ∈ S), it acts trivially on the codespace. - If it anticommutes with
at least one stabilizer (Ea† Eb ∈/ S), it maps the codespace to an orthogonal subspace. -
This ensures that errors can be detected.
2. Error Correction: - For any single-qubit error Ea , measuring the stabilizers produces
a unique syndrome (pattern of +1/ − 1 eigenvalues). - The syndrome identifies both the
type and location of the error. - Since there are four independent stabilizers and each
acts on five qubits, they provide sufficient information to correct any single-qubit error.
Error Correction Conditions: To show that this code satisfies the error correction
conditions, we need to verify that for any pair of errors Ea and Eb , the product Ea† Eb
either commutes with all generators of the stabilizer or anticommutes with at least one
generator.

(i) The single-qubit error operators commutes with S2 , S3 , and S4 , but anticommutes
with S1 . Thus, the syndrome measurement can detect this error. We consider how
each single-qubit error affects the stabilizers.

(ii)

Q6: Measuring a stabilizer

H⊗I C−G 1 H⊗I 1

|0⟩⊗|ψ⟩ −−−→ |+⟩⊗|ψ⟩ −−−→ √ (|0⟩⊗|ψ⟩+|1⟩⊗G|ψ⟩) −−−→ √ (|+⟩⊗|ψ⟩+|−⟩⊗G|ψ⟩)
2 2
 
1 1 1 1
Simplify, √ (|+⟩⊗|ψ⟩+|−⟩⊗G|ψ⟩) = √ |0⟩ ⊗ √ (|ψ⟩ + G|ψ⟩) + |1⟩ ⊗ √ (|ψ⟩ − G|ψ⟩)
2 2 2 2
Eigen decomposition: G = |g+ ⟩⟨g+ |−|g− ⟩⟨g− |, |ψ⟩ in eigen basis of G: |ψ⟩ = α|g+ ⟩+β|g− ⟩
1
|0⟩ measurement: √ (|ψ⟩ + G|ψ⟩) ∝ |g+ ⟩
2
1
|1⟩ measurement: √ (|ψ⟩ − G|ψ⟩) ∝ |g− ⟩
2