These exercises practice the partial trace as a computational rule and as a physical rule for local statistics. All state vectors below are normalized.
Trace over B. The two off-diagonal terms vanish because ⟨1∣0⟩=0 and ⟨0∣1⟩=0:
ρA=21(∣0⟩⟨0∣+∣1⟩⟨1∣)=21I.
The same calculation gives ρB=I/2.
The classically correlated state
ρAB=21∣00⟩⟨00∣+21∣11⟩⟨11∣
has the same one-qubit reduced states as ∣Φ+⟩. Verify this, and state one global difference between the two states.
Solution
Trace over B term by term:
TrB(∣00⟩⟨00∣)=∣0⟩⟨0∣,
and
TrB(∣11⟩⟨11∣)=∣1⟩⟨1∣.
Therefore
ρA=21I,ρB=21I.
The global states are nevertheless different. The Bell state is a pure entangled state with coherence between ∣00⟩ and ∣11⟩. The classically correlated state is mixed and has no such off-diagonal coherence.
The two C states are orthogonal, so the trace over C removes cross terms between the C=0 and C=1 branches. The C=0 branch has norm squared 2/3 and normalized AB state
∣Ψ+⟩=21(∣01⟩+∣10⟩).
Therefore
ρAB=32∣Ψ+⟩⟨Ψ+∣+31∣00⟩⟨00∣.
This reduction remains partly coherent in the one-excitation subspace.
For ∣Φ+⟩, use the reduced state to compute the probabilities for measuring Z on subsystem A. Then compare with measuring the parity observable Z⊗Z.
Solution
From Exercise 2,
ρA=21I.
The projectors for measuring Z are
P0=∣0⟩⟨0∣,P1=∣1⟩⟨1∣.
Thus
p(0)=Tr(ρAP0)=21,p(1)=Tr(ρAP1)=21.
The local outcome is random. But the joint observable has perfect correlation:
(Z⊗Z)∣Φ+⟩=∣Φ+⟩.
Parity is +1 with certainty, so ⟨Z⊗Z⟩=1. Measuring each qubit’s Z separately gives the finer outcomes 00 and 11, each with probability 1/2. An ideal Lüders parity measurement preserves this Bell state, whereas separately measuring both qubits and forgetting their outcomes gives the mixture in Exercise 3. The observable’s expectation alone does not specify a measurement’s disturbance; see local and global observables.
Construct a purification of
ρA=p∣0⟩⟨0∣+(1−p)∣1⟩⟨1∣,0≤p≤1,
and verify that tracing out the reference system gives ρA.
Tracing over R removes the cross terms because the reference states are orthogonal. The result is
TrR(∣Ψ⟩⟨Ψ∣)=p∣0⟩⟨0∣+(1−p)∣1⟩⟨1∣=ρA.
Let
ρAB=41(I⊗I+rZ⊗I+sI⊗Z+tZ⊗Z).
For real r,s,t, find the conditions for this matrix to be positive. Compute ρA and identify which coefficients are invisible to measurements on A alone.
Solution
The matrix is diagonal. Its four eigenvalues are
λab=41+rza+szb+tzazb,z0=1,z1=−1.
All four must be nonnegative; the trace is already one. These inequalities restrict r,s,t jointly, not just one parameter at a time. For example, r=s=1,t=0 gives a negative eigenvalue and is not a state. Next use
Tr(I)=2,Tr(Z)=0.
Then
ρA=TrBρAB=41(2I+2rZ+sITrZ+tZTrZ)=21(I+rZ).
The coefficients s and t do not appear in ρA. Similarly, ρB=(I+sZ)/2. The joint state is a product exactly when t=rs; other allowed values of t describe classical correlations in this diagonal family.
In the ordered qubit–qutrit basis, reduce the complex state
∣ψ⟩=3∣00⟩+i∣11⟩+∣12⟩.
Compute both local purities and the expectation of
Y12=−i∣1⟩⟨2∣+i∣2⟩⟨1∣ on B.
Solution
The coefficient array is C=(100i01)/3. Direct contraction gives
ρA=31(1002),ρB=3110001−i0i1.
Both have nonzero eigenvalues 1/3,2/3, hence purity 5/9. The local expectation is
Tr(ρBY12)=−32.
In this coefficient convention ρB=CTC∗, not C†C. The latter would reverse the imaginary entries and the sign of this measured expectation, while leaving the eigenvalues and purity unchanged. This is why a complex example tests more than normalization alone.