Weighted Sum-Rate Optimal in Near-Field THz-MIMO Systems: A Multigroup IRS- Assisted Hybrid Beamforming Approach

In intelligent reflecting surface (IRS)-assisted multi-user terahertz multiple-input multiple-output (THz-MIMO) near-field systems, conventional hybrid beamforming schemes suffer from severe beam misalignment and inter-user interference due to mismatch between planar wave assumptions and spherical wavefront propagation. To address this challenge, we propose a subarray-subgroup-user mapping architecture, where widely spaced base station (BS) subarrays and IRS subgroups are paired with users through sensing of channel. This scheme decomposes near-field spherical wave channels into localized quasi-planar wave subchannels, establishing a joint beamforming optimization framework to maximize the weighted sumrate (WSR) under power and hardware constraints. Specifically, the problem is solved by first fixing the digital beamforming part of the system. Then, the active analog beamforming at the BS/user and the passive analog beamforming at the IRS are updated iteratively based on the singular value decomposition (SVD) of the users’ equivalent channels. Next, a Block Diagonalization (BD) algorithm is used to solve the digital beamforming part of the system. The experiment shows that by pairing the BS subarray, IRS subgroup, and users, the dimensions of the IRS phase shift matrix and the channel matrix are reduced, which significantly lowers complexity. This approach also effectively suppresses inter-user interference in the IRS cascaded channel, enhancing the system’s WSR.