Capacity-Achieving Precoding with Low-Complexity for Terahertz LOS Massive MIMO using Uniform Planar Arrays

Terahertz (THz) line-of-sight (LOS) massive multiple-input multiple-output (MIMO) communication is promising to achieve a few terabit rates for beyond 5G. The tiny wavelength in THz allows us to pack massive antenna elements onto a small aperture size. As the number of antenna elements scales, it is critical to improving the energy efficiency in the baseband processing while maintaining spectral efficiency performance. In this paper, we present a two-stage precoding method that achieves the MIMO channel capacity with significantly reduced computational complexity. The principal idea is to decompose a uniform planner array (UPA) into a superposition of multi-layered uniform circular subarrays (UCAs) at both ends. Thanks to the structure of the circulant channel matrix between the subarrays, the proposed precoding exploits the Fast Fourier Transform (FFT)-based inner precoding per subarray conjunction with singular value decomposition (SVD) outer precoding per eigen-mode. We show that this two-stage precoding method provides a multiplicative gain in the computational complexity reduction compared to the conventional SVD precoding.