On the hardness of conversion from entangled proof into separable one

Abstract: A quantum channel whose image approximates the set of separable states is called a disentangler, which plays a prominent role in the investigation of variants of the computational model called Quantum Merlin Arthur games, and has potential applications in classical and quantum algorithms for the separability testing and NP-complete problems. So far, two types of a disentangler, constructed based on $\epsilon$-nets and the quantum de Finetti theorem, have been known; however, both of them require an exponentially large input system. Moreover, in 2008, John Watrous conjectured that any disentangler requires an exponentially large input system, called the disentangler conjecture. In this paper, we show that both of the two known disentanglers can be regarded as examples of a strong disentangler, which is a disentangler approximately breaking entanglement between one output system and the composite system of another output system and the arbitrarily large environment. Note that the strong disentangler is essentially an approximately entanglement-breaking channel while the original disentangler is an approximately entanglement-annihilating channel, and the set of strong disentanglers is a subset of disentanglers. As a main result, we show that the disentangler conjecture is true for this subset, the set of strong disentanglers, for a wide range of approximation parameters without any computational hardness assumptions.