Nonlinear Stacked Intelligent Surfaces for Wireless Systems

Stacked intelligent surfaces (SIS) are a promising technology for next-generation wireless systems, offering an opportunity to enhance communication performance with low power consumption. Typically, an SIS is modelled as a surface that imparts phase shifts on impinging electromagnetic signals to achieve desired communication objectives. However, this mode of operation results in a linear SIS, which limits its applicability to linear operations. To unlock further SIS potential, we propose a nonlinear SIS that can mimic the behaviour of nonlinear neural networks. We discuss the feasibility and potential of this idea and introduce a nonlinear unit cell that changes between absorptive and transmissive states in response to signal amplitude. To evaluate the system-level performance of the nonlinear SIS, we present a case study where the SIS structures are optimized to minimize the symbol error rate (SER) in an MIMO system with SIS deployed at both the transmitter and receiver sides using only statistical channel information. We demonstrate that a nonlinear SIS can improve communication reliability compared to a linear SIS by forming complex signal patterns across the SIS surface, which provide higher diversity against noise disturbances, while still allowing the receiver to discern these patterns. Finally, we outline several potential applications of nonlinear SIS in wireless communication scenarios.