LAUSR

Intelligent reflecting surface (IRS) is considered as a promising application in terahertz (THz) communications since it is able to enhance the THz communication with no additional power consumptions. In this letter, we consider a downlink multi‐input multi‐output system, where a base station (BS) communicates to a user with the aid of the IRS. Our goal is to optimize the spectral efficiency by designing discrete phase‐shifts at the IRS and the precoder at the BS, which is an intractable nonconvex optimization problem. To solve it, we approximate the problem into another form where we maximize the sum of path gains by designing the phase‐shifts of the IRS, and we propose a cross‐entropy based algorithm to search the optimal IRS phase‐shifts matrix. After that, we use the classical singular value decomposition and water‐filling solution to design the optimal precoder.