I. Introduction
The appetite for ever higher frequencies on which to communicate wirelessly is insatiable, with the next frontier being the sub-terahertz band (0.1-1 THz) [1]–[3]. At these frequencies, the propagation is ray-like [4], [5] and the availability of a line-of-sight (LOS) component or some strong reflection largely determines the performance. Hence, these frequencies are mostly envisioned for short-range transmissions, which go hand in hand with denser networks. Such densification need not entail additional full-stack access points, and relays could be deployed in their stead. Or, as an emerging alternative, the densification could be aided by intelligent reflecting surfaces (IRSs), which are passive, devoid of radio-frequency chains and baseband processing [6]–[8]. Accordingly, the main motivating application for IRSs at sub-terahertz frequencies is to bypass obstructions, yet, as expounded in this paper, a potentially more impactful benefit may arise from their deployment: an enrichment in the number of spatial degrees of freedom (DOF). While an inherently asymptotic notion, which describes the scaling of the spectral efficiency with the received power [9, Sec. 4.2], the number of DOF translates directly to the number of concurrent signal streams that can be communicated and is therefore a quantity of prime interest.