Is entanglement time-symmetric? It is common to describe an experiment as "causing" entanglement.  For example, two quantum particles that interact become entangled as a result of their interaction, so we are likely to say that the interaction "caused" the entanglement.
However, quantum mechanics is time-symmetric, so a time reversed movie of the experiment would show two entangled particles approaching each other, interacting, and then going on their way unentangled.
I suspect that this picture is incorrect, but can't put my finger on why.
 A: The picture is correct. The key fact is that interaction can both create and destroy entanglement, so there is no conflict with time symmetry.
Suppose for example that two qubits initially in the product state $|\psi_i\rangle = |+\rangle|0\rangle$ are allowed to interact in a way that performs the CNOT gate from the first to the second qubit. Then the final state is $|\psi_f\rangle=(|00\rangle + |11\rangle)/\sqrt{2}$.
Now consider the above process in reverse. Two qubits begin in the Bell state $|\psi_f\rangle=(|00\rangle + |11\rangle)/\sqrt{2}$, interact in a way equivalent to the CNOT gate and finish in the state $|\psi_i\rangle=|+\rangle|0\rangle$.

In terms of spins, you can think of $|1\rangle$ as the "spin up" state, of $|0\rangle$ as the "spin down" state and of $|+\rangle=(|0\rangle+|1\rangle)/\sqrt{2}$ as the uniform superposition of the two states with zero relative phase. A process implementing the CNOT gate is then any interaction that flips the second spin if the first spin is up.
