9.2 Quantum Channel and LOCC
Preparation Channel
Preparation Channel: \(\mathcal{E}\circ\Delta=\mathcal{E}\). The channel prepare density matrix
We know \(\Delta(\rho)=\sum^d_{x=1}\lang x|\rho|x\rang|x\rang\lang x|\)
Then \(\mathcal{E}(\rho)=\mathcal{E}\circ \Delta(\rho)=\sum^{d}_{x=1}\lang x|\rho|x\rang \mathcal{E}(|x\rang\lang x|)=\sum^{d}_{x=1}\lang x|\rho|x\rang\sigma_{x}\) where \(\sigma_x:=\mathcal{E}(|x\rangle\langle x|)\)
We can calculate the Choi matrix: \(J_{\mathcal{E}}^{A\tilde{A}}=\mathcal{E}^{A\to \tilde{A}}=\mathcal{E}\circ\Delta (\Omega^{A\tilde{A}})=\sum_{x,y}|x\rang\lang y|\otimes \mathcal{E}\circ\Delta(|x\rang \lang y|)\)
\(=\sum_{x,y}\delta_{xy}|x\rangle\langle y|\otimes\mathcal{E}(|x\rangle\langle x|) =\sum_{x}|x\rangle\langle x|\otimes\sigma_{x}\). This is the Choi matrix of preparation channel.
There is a corresponding between channel and choi matrix
Therefore, \(\mathcal{E} \in \mathrm{CPTP}(A \rightarrow B)\) is a cq-channel if and only if its Choi matrix \(J_{\mathcal{E}}^{A B}\) is a cq-state: \(\rho^{XA}=\sum_{x\in[m]}p_{x}|x\rang\lang x|^{X}\otimes \rho^{A}_{x}\)
H.M. Ex 3.48
Let \(\mathcal{E} \in \text{CPTP}(A \rightarrow B)\) be a cq-channel as above, and set \(m := |A|\) and \(n := |B|\).
-
Show that \(\{M_{xy}\}\) with \(M_{xy} : A \rightarrow B\) and \(M_{xy}:=\sqrt{\sigma_{x}}|y^{B}\rangle\langle x^{A}|,x\in[m]\,,\,y\in[n]\) forms an operator sum representation of \(\mathcal{E}\).
Proof
Let \(M_{xy}:=\sqrt{\sigma_x}\,|y^B\rangle\langle x^A|\) for \(x\in[m],\,y\in[n]\).
\(\sum_{x,y}M_{xy}\,\rho\,M_{xy}^{*}=\sum_{x,y}\sqrt{\sigma_{x}}\,|y\rangle\langle x|\,\rho\,|x\rangle\langle y|\,\sqrt{\sigma_{x}}=\sum_{x} \langle x|\rho|x\rangle \,\sqrt{\sigma_{x}}\Big(\sum_{y} |y\rangle\langle y|\Big)\sqrt{\sigma_{x}}=\sum_{x} \langle x|\rho|x\rangle\,\sigma_{x}=\mathcal{E}(\rho).\)
And \(\sum_{x,y}M_{xy}^{*}M_{xy}=\sum_{x,y}|x\rangle\langle y|\,\sigma_{x}\,|y\rangle \langle x|=\sum_{x}\big(\operatorname{Tr}\sigma_{x}\big)\,|x\rangle\langle x|=\sum _{x}|x\rangle\langle x|=I^{A}.\)
-
Show that the map \(V: A \rightarrow B \otimes \tilde{A}\tilde{B}\) given by \(V=\sum_{x\in[m]}\left(\sqrt{\sigma_{x}^{B}}\otimes I^{\tilde{B}}\right)|\Omega^{B\tilde{B}} \rangle\otimes|x^{\tilde{A}}\rangle\langle x^{A}|\) is a Stinespring isometry (the environment system is \(\tilde{A}\tilde{B}\)) satisfying \(\text{Tr}_{\tilde{A}\tilde{B}}[V\rho V^{*}]=\sum_{x\in[m]}p_{x}\sigma_{x}=\mathcal{E} (\rho)\text{ and }V^{*}V=I^{A}\)
Stinespring isometry(an isometry \(V:A\to BE\) s.t. \(\mathcal{E}(\rho)=\text{Tr}_{E}[V\rho V^{*}],\forall \rho\in\mathcal{L}(A)\))
Proof
Let \(V \;=\; \sum_{x\in[m]}\big(\sqrt{\sigma_{x}^{B}}\otimes I^{\tilde B}\big)\,|\Omega ^{B\tilde B}\rangle \otimes |x^{\tilde A}\rangle\langle x^{A}|\) with \(|\Omega^{B\tilde B}\rangle=\sum_{y}|y^B y^{\tilde B}\rangle\).
Isometry: Let \(|w_x\rangle:=(\sqrt{\sigma_x}\otimes I)|\Omega\rangle\). Then \(V=\sum_{x}|w_{x}\rangle_{B\tilde{B}}\otimes|x^{\tilde{A}}\rangle\langle x^{A}|,\text{ then }V^{*}V=\sum_{x}\langle w_{x}|w_{x}\rangle\,|x^{A}\rangle\langle x^{A}|=\sum _{x}\operatorname{\mathrm{Tr}}\sigma_{x}\,|x\rangle\langle x|=I^{A}.\)
Output state: \(V\rho V^{*}=\sum_{x,x'}\langle x|\rho|x'\rangle\,|w_{x}\rangle\langle w_{x'} |\otimes |x^{\tilde A}\rangle\langle x'^{\tilde A}|.\)
Then \(\operatorname{Tr}_{\tilde A}[V\rho V^{*}]=\sum_{x} \langle x|\rho|x\rangle \,|w_{x}\rangle\langle w_{x}|.\)
Since \(\operatorname{Tr}_{\tilde B}\big[\,|\Omega\rangle\langle\Omega|\,\big]=I^B\), then \(\operatorname{Tr}_{\tilde A\tilde B}[V\rho V^{*}]=\sum_{x} \langle x|\rho| x\rangle\,\operatorname{Tr}_{\tilde B}\big[(\sqrt{\sigma_{x}}\otimes I)|\Omega\rangle \langle\Omega|(\sqrt{\sigma_{x}}\otimes I)\big]=\sum_{x} \langle x|\rho|x\rangle\, \sigma_{x}=\mathcal{E}(\rho).\)
Measurement-Prepare Channel
\(\mathcal{E}:=\mathcal{P}^{X\rightarrow B}\circ M^{A\rightarrow X}\) where \(M^{A\to X}(\rho)=\sum_{x=1}^{m}\operatorname{\mathrm{Tr}}[\underbrace{\Lambda_{x}} _{POVM}\rho]\underbrace{|x\rangle\langle x|^{X}}_{\text{Classical Outcome}}\) POVM Channel
Then \(\mathcal{E}(\rho) = \mathcal{P}\cdot M(\rho)\) \(=\sum_{x=1}^{m}\operatorname{\mathrm{Tr}}[\Lambda_{x}\rho]\mathcal{P}(|x\rangle\langle x|)\)
Let \(\mathcal{E}(\rho)=\sum_{z=1}^{m}\operatorname{\mathrm{Tr}}\left[\Lambda_{z}\rho\right ]\sigma_{z}^{B}\), then Choi Matrix: \(J_{\mathcal{E}}^{AB}=\mathcal{E}^{\tilde{A}\rightarrow B}\left(\Omega^{A\tilde{A}} \right)=\sum_{x,y\in[d]}|x\rangle\langle y|\otimes\mathcal{E}(|x\rangle\langle y| )\)
\(=\sum_{x,y\in[d]}|x\rangle\langle y|\otimes\sum_{z=1}^{m}\operatorname{\mathrm{Tr}} [\Lambda_{z}|x\rangle\langle y|]\sigma_{z}^{B}=\sum_{x,y\in[d]}|x\rangle\langle y |\otimes\sum_{z=1}^{m}\langle y|\Lambda_{z}|x\rangle\sigma_{z}^{B}\)
\(= \sum_{z=1}^{m}\underbrace{\left( \sum_{x,y \in [d]}\langle y|\Lambda_{z}|x\rangle |x\rangle \langle y| \right)}_{\Lambda_{z}^{T}} \otimes \sigma_{z}\)
Thus \(J_{\mathcal{E}}^{AB}=\sum_{z=1}^{m}\Lambda_{z}^{T}\otimes\sigma_{z}^{B}\). This is separable density matrix
Because this channel destroy entangled state
Also when we do \(\text{Tr}_B\) on \(J_\mathcal{E}^{AB}\), we get identity since \(\sigma_z\) has rank 1
Choi matrix is density matrix but not normalized, but even \(\frac{1}{|A|} J_{\mathcal{E}}^{A B}\) is not entangled
Intuition: We make it into classical state, then return it to quantum
The moment it becomes classical, it is no longer quantum entangled. Then, if it returns to quantum, it loses the entanglement.
Entanglement Breaking Channel
Let \(\mathcal{E} \in CPTP(A \rightarrow B)\). We say that \(\mathcal{E}\) is entanglement-breaking if \(\forall \rho \in \mathfrak{D}(AR)\), \(\mathcal{E}^{A \rightarrow B}(\rho^{RA})\) is separable
Theorem
\(\mathcal{E}\) is entanglement breaking iff \(\mathcal{E}\) is Measurement-Prepare Channel
Proof
\(\Leftarrow\)) Since \(\rho\) is convex combination of pure states, then it's enough to prove it for all pure state
Then NTP: \(\mathcal{E}^{A\to B}(\psi^{RA})\) is separable \(\forall\) pure states \(\psi^{RA}\)
We know \(|\psi^{RA}\rang =(M\otimes I^{A})|\Omega^{\tilde{A}A}\rang\), then \(\mathcal{E}^{A\to B}(\psi^{RA})=(M\otimes I^{A})\mathcal{E}\left(\Omega^{\tilde{A}A} \right)(M\otimes I^{A})^{*}\)
\(=\left(M\otimes I^{A})J_{\mathcal{E}}^{AB}\right.(M\otimes I^{A})^{*}\), then since this, \(=\left(M\otimes I^{A}\right)\left(\sum_{z=1}^{m}\Lambda_{z}^{T}\otimes\sigma_{z} ^{B}\right)\left(M\otimes I^{A}\right)^*\)
\(=\sum_{z=1}^{m}q_{z}\frac{M\Lambda_{z}M^{*}}{q_{z}}\otimes\sigma_{z}\) where \(q_{z}:=\operatorname{Tr}\left[M \Lambda_{z} M^{*}\right]=\operatorname{Tr}\left[M^{*} M \Lambda_{z}\right]\)
Thus this is separable
Check: Calculate the reduce density matrix of \(\psi^{RA}\) and check that \(M^*M\) is a density matrix
\(\Rightarrow\)) Let \(J_{\mathcal{E}}^{AB}=\mathcal{E}^{\tilde{A}\rightarrow B}\left(\Omega^{A\tilde{A}} \right)=\sum_{z}q_{z}\omega_{z}^{A}\otimes\tau_{z}^{B}\), let \(\eta_{z}^{A}:=q_{z}\omega_{z}^{A}\), then \(J_{\mathcal{E}}^{AB}=\sum_{z}\eta_{z}^{A}\otimes\tau_{z}^{B}\).
Then we just check that \(\eta_{z}^{A}\) is a POVM: \(\text{Tr}_{B}\left\lbrack J_{\mathcal{E}}^{AB}\right\rbrack=\sum_{z}\eta_{z}^{A} \cdot\text{Tr}[\tau_{z}^{B}]=\sum_{z}\eta_{z}^{A}\)
Since \(\mathcal{E}\) is quantum channel, then \(J_\mathcal{E}^A=I^A\). Thus \(\sum_z\eta_z^A=I_A\)
LOCC(Local Operations Classical Communication)
Definition
Entanglement is a property of a composite physical system that cannot be generated nor increased by LOCC
Example: Separable can be generated by LOCC since Alice and Bob can do local operation
Alice has \(p_x\) and prepare \(\rho_x\) and send \(x\) to Bob, then Bob prepares \(\sigma_x\). Then they all forget what is \(x\)
Then the whole system is \(\sum_xp_x\rho_x^A\otimes \sigma_x^B\)
Similar to Quantum Teleportation
Any measurement on Bob can be simulated by a measurement on Alice, up to a unitary on Bob.
Let \(|\psi^{AB}\rangle\) be a bipartite pure state and let \(\{N_x\}\) be measurement operators on Bob with \(\sum_{x} N_{x}^{*}N_{x} = I_{B}\).
We want to show that for each \(x\) there exist operators \(M_x\) on Alice and unitaries \(U_x\) on Bob such that \((I\otimes N_{x})\,|\psi^{AB}\rangle\;=\;(M_{x}\otimes V_{x})\,|\psi^{AB}\rangle,\) and that \(\sum_{x} M_{x}^{*}M_{x} = I_{A}\) on the support of Alice’s reduced state.
Proof
Let \(|\psi^{AB}\rang\) be a state on Alice with a measurement \(N_x\) where \(\sum_x N^*_xN_x=I\)
We want to prove \(I\otimes N_{x}|\psi^{AB}\rangle\) is equivalent \(I\otimes N|\psi^{AB}\rangle\)
By Schmidt decomposition: \(|\psi^{AB}\rangle=\sum_{y}\sqrt{p_{y}}\,|\phi_{y}\rangle_{A}\,|\varphi_{y}\rangle _{B}=(\Lambda\otimes I)\,|\Omega\rangle\) where \(\Lambda = \sum_y \sqrt{p_y}\,|\phi_y\rangle\langle \phi_y|\) (diagonal in \(\{|\phi_y\rangle\}\)) and \(|\Omega\rangle=\sum_{y}|\phi_{y}\rangle_{A}\,|\varphi_{y}\rangle_{B}\)
Then \((I\otimes N)\,|\psi\rangle=(\Lambda\otimes N)\,|\Omega\rangle=(\Lambda N^{T}\otimes I)\,|\Omega\rangle\) and since \(\Lambda\) is diagonal, then \(\Lambda N^{T}\) is diagonal, by singular value decomposition, we get
\(=UN\Lambda V\otimes I|\Omega\rangle\) \(=UN\Lambda\otimes V^{T}I|\Omega\rangle\) \(=(UN\otimes V^{T})\underbrace{(\Lambda\otimes I)|\Omega\rangle}_{|\psi\rang}\)
Thus \((I\otimes N_{x})|\psi^{AB}\rangle=(\underbrace{U_{x}N_{x}}_{M_{x}}\otimes V_{x}^{T} )|\psi^{AB}\rangle=(M_{x}\otimes V_{x}^{T})|\psi^{AB}\rangle\) where \(\sum M_{x}^{*} M_{x} = \sum N_{x}^{*} N_{x} = I\)
Also \(p_x=\lang \psi^{AB}|I^A\otimes N_x^*N_x|\psi^{AB}\rang=\lang \psi^{AB}|M_x^*M_x\otimes I^B|\psi^{AB}\rang\)
Every measurement bob performed is equivalent to Alice performing measurement and bob do a unitary
This is the most general LOCC(LOCC on pure state)