RIS as a Deep-Fade Shaper: A New Framework for Fluid-Antenna Multiple Access

Fluid-antenna multiple access (FAMA) enables each user device to opportunistically select the port experiencing the most favorable instantaneous interference condition, thereby extracting micro-scale spatial diversity without relying on multiple radio frequency (RF) chains. However, existing FAMA studies assume purely stochastic interference fields and do not provide mechanisms for actively engineering the spatial fluctuations that FAMA relies upon. In parallel, reconfigurable intelligent surfaces (RISs) have been used almost exclusively for coherent signal enhancement, leaving the interference behavior largely unaltered. This paper introduces a fundamentally different integration of RISs with FAMA by using the RIS as a deep-fade engineering device that shapes the second-order statistics of the interference field rather than the desired signal. In near-field user-centric deployments, the RIS induces strong, controllable port-to-port interference variations that significantly increase the likelihood of deep fades at one or more fluid-antenna ports. We develop a full statistical characterization of the cascaded interferer-RIS-UE channel, show that the aggregated interference follows a generalized Gamma convolution law, and obtain a highly accurate moment-matched Gamma model whose parameters are explicitly RIS-controlled. Building on this model, we formulate a variance-maximization RIS design tailored to FAMA port selection and devise a computationally efficient projected gradient-ascent algorithm. We then derive the exact per-port SIR distribution, proving that it follows a Beta-prime law, and provide closed-form outage expressions for both independent and correlated ports. Extensive simulations validate all analytical results and demonstrate substantial performance gains. Compared with conventional FAMA and random-phase RIS configurations, the proposed RIS-enabled deep-fade engineering achieves multi-dB SIR improvements, and markedly lower outage probability.