Local entanglement is not necessary for perfect discrimination between unitary operations acting on two qudits by local operations and classical communication

Recently, the problem of discriminating multipartite unitary operations by local operations and classical communication (LOCC) has attracted significant attention. The latest work in the literature on this problem showed that two multipartite unitary operations can always be perfectly distinguished by LOCC when a finite number of runs are allowable. However, in these schemes, local entanglement (an entangled state held by one party) was required, which seems to imply that local entanglement is necessary for perfect discrimination between unitary operations by LOCC. In this paper, we show that a perfect discrimination between two unitary operations acting on a two-qudit system can always be achieved without exploiting any entanglement. As a result, we conclude that local entanglement is not necessary for perfect discrimination between unitary operations acting on a two-qudit system by LOCC. Our result, in general, is based on the assumption made by X. F. Zhou, Y. S. Zhang, and G. C. Guo [Phys. Rev. Lett. 99, 170401 (2007)] that for any unitary operation $U$, we can obtain its reverse transformation ${U}^{\ifmmode\dagger\else\textdagger\fi{}}$ by exchanging the input and output ports of the whole setup. Thus, we further consider whether we can avoid using this assumption, and we find that for unitary operations acting on two qubits, our result can be obtained without using this assumption. However, in higher dimension how to do that seems to be unclear.