When designing next-generation wireless networks (5G-Advanced and 6G), Reconfigurable Intelligent Surfaces (RIS) and Dynamic Metasurface Antennas (DMAs) are often conceptualized as "ideal programmable mirrors." In standard theoretical models, an engineer assumes that if an RIS element is programmed to introduce a phase shift of $\pi/2$, it will apply that exact same phase shift to every incoming electromagnetic wave.
This assumption is known as a frequency-flat response. While mathematically convenient, it hides a fundamental physical reality: reconfigurable surfaces are constructed from resonant metamaterial elements. In wideband communication systems—where signals span a broad range of frequencies divided into multiple subcarriers—treating these surfaces as frequency-flat leads to severe performance degradation.
A rapidly growing body of research focuses on frequency-selective hardware modeling. By replacing abstract mathematical simplifications with physics-compliant models, researchers are uncovering how to control and exploit the frequency-dependent nature of real-world hardware.
The Illusion of the Frequency-Flat Surface
In wideband communication systems, such as those utilizing Orthogonal Frequency Division Multiplexing (OFDM), the wireless channel is inherently frequency-selective; different frequencies experience different fading. However, as noted by Meng Hua et al. in Intelligent Reflecting Surface-Aided Joint Processing Coordinated Multipoint Transmission, conventional optimization models mistakenly assume that the reflecting surface itself treats all frequencies identically.
In reality, an RIS element cannot decouple its phase response from the frequency of the incident wave. If you optimize the surface configuration perfectly for the center subcarrier of a wideband signal, the phase and amplitude responses will distort at the outer edges of the frequency band. As Wenjing Yan et al. demonstrate in Frequency Reflection Modulation for Reconfigurable Intelligent Surface Aided OFDM Systems, ignoring these hardware constraints introduces severe signal mismatch, degrading the efficiency of reflection modulation and beamforming schemes.
The Physical Consequence: Beam Split
To understand why this matters intuitively, consider what happens to a signal's spatial direction. When a wideband signal hits a frequency-selective surface, the phase shift varies per subcarrier. Because the spatial steering of a wave depends directly on the phase progression across the surface elements, different frequency components of the exact same signal end up being redirected toward different physical locations.
This phenomenon is known as beam split. Shuhao Zeng et al. explore this in Beamforming Design for Wideband Near-Field Communications With Reconfigurable Refractive Surfaces, demonstrating that the combined effects of near-field propagation and frequency selectivity exacerbate beam split. Instead of focusing energy tightly on a target user, the energy disperses across space, drastically reducing the achievable data rate.
Ideal (Frequency-Flat) Model:
[Wideband Signal] ---> [ RIS Element ] ---> All frequencies reflect to User A
Realistic (Frequency-Selective) Model:
/ Subcarrier 1 ---> Reflects to User A
[Wideband Signal] ---> [ RIS Element ] ---> Subcarrier 2 ---> Reflects to Space B (Beam Split)
\ Subcarrier 3 ---> Reflects to Space C
Grounding Models in Physical Reality
To resolve this discrepancy, recent literature replaces abstract phase-shift matrices with physics-aware analytical frameworks. Two primary methods are used to capture how hardware components influence signal phase and amplitude across a wideband spectrum.
1. The Lorentzian Resonator Model
Instead of modeling an RIS element as a static complex number, it can be modeled as a physical oscillator. Nir Shlezinger and Yonina C. Eldar utilize this approach in Dynamic Metasurface Antennas for 6G Extreme Massive MIMO Communications, applying a Lorentzian equation to characterize individual metamaterial elements.
As expanded by Konstantinos D. Katsanos et al. in Wideband Multi-User MIMO Communications with Frequency Selective RISs: Element Response Modeling and Sum-Rate Maximization, the Lorentzian function maps the element's response to tangible physical parameters:
- Oscillator strength: Dictates the magnitude of the coupling.
- Damping factor: Represents energy losses within the material.
- Resonance frequency: The specific frequency where the phase transitions most sharply.
When an engineer tunes an RIS element, they are not directly changing a "phase angle"; they are shifting its resonance frequency. Shifting this peak alters the entire Lorentzian curve, meaning a single adjustment changes the phase and amplitude response across all subcarriers simultaneously in a highly structured, predictable way.
2. Equivalent Resonant Circuits
An alternative, hardware-equivalent approach models each reflective element as a microscopic circuit containing resistors, inductors, and tunable varactor capacitors.
In Multi-RIS-Empowered Multiple Access: A Distributed Sum-Rate Maximization Approach, Konstantinos D. Katsanos et al. use these equivalent resonant circuits to model "Beyond Diagonal" RIS devices. By adjusting the voltage across a tunable capacitor, the circuit's impedance changes, which alters how the circuit reflects different frequencies. This provides a mathematically tractable way to optimize hardware configurations based on actual circuit components.
Turning a Limitation into a Design Parameter
Once optimization algorithms incorporate these physics-based models, frequency selectivity stops being an unpredictable bottleneck and becomes a structured design parameter.
For instance, Yajun Wang et al. embed a Lorentzian-like profile directly into their signal processing framework in Efficient Joint Precoding Design for Wideband Intelligent Reflecting Surface-Assisted Cell-Free Network. Because the algorithm understands the non-linear relationship between frequencies, it can jointly design the active precoding at the base stations and the passive response at the surface to ensure robust signal alignment across the entire bandwidth.
Similarly, in Distributed Sum-Rate Maximization of Cellular Communications with Multiple Reconfigurable Intelligent Surfaces, Konstantinos D. Katsanos et al. develop a Distributed Successive Concave Approximation (D-SCA) algorithm. Instead of assuming independent phase control per subcarrier (which is physically impossible), the algorithm optimizes the shared underlying physical parameters—like damping and oscillator strength—to maximize the total network sum-rate. This cooperative approach is extended to multi-cell interference channels by Konstantinos D. Katsanos et al. in The Interference Broadcast Channel with Reconfigurable Intelligent Surfaces: A Cooperative Sum-Rate Maximization Approach, proving that modeling circuit-level constraints prevents optimization mismatch in complex multi-user topologies.
Finally, this hardware-aware paradigm is proving vital for resource allocation beyond pure communication. In Power Minimizing MEC Offloading with QoS Constraints over RIS-Empowered Communications, Mattia Merluzzi et al. demonstrate that accounting for the frequency-dispersive nature of RIS elements is necessary to meet strict Quality of Service (QoS) and energy constraints in Mobile Edge Computing (MEC). When the system accurately models how the surface behaves across subcarriers, it avoids over-provisioning power, leading to highly energy-efficient offloading.
Key Foundational Work
Intelligent Reflecting Surface-Aided Joint Processing Coordinated Multipoint Transmission (2020) — contributed 1 concept elements to this direction.
Dynamic Metasurface Antennas for 6G Extreme Massive MIMO Communications (2021) — contributed 1 concept elements to this direction.