The Abstraction Gap in Wireless Communications
For decades, wireless communication design has relied on Claude Shannon’s Classical Information Theory (CIT). CIT abstracts the physical world by compressing the complex behavior of radio waves into discrete mathematical entities: channel matrices ($H$) filled with statistical random variables representing fading effects (such as Rayleigh or Rician distributions).
This abstraction works remarkably well when antennas are separated by many wavelengths and signals arrive as uniform, parallel plane waves. However, as wireless engineering transitions toward 6G, networks are deploying extremely large-scale antenna arrays (XL-arrays) and continuous, software-defined surfaces known as Holographic MIMO (HMIMO).
As Tierui Gong et al. point out, standard information theory frameworks completely omit the underlying electromagnetic and circuit-level behaviors of these dense elements. By ignoring the physical constraints of wave propagation, a significant gap opens between what mathematical models predict and what physical systems can achieve. To close this gap, a new paradigm is required: Electromagnetic Information Theory (EIT).
Why Far-Field Models Break Down
To understand why EIT is necessary, we must look at how traditional models fail when antennas are packed tightly together or expanded across massive surfaces.
In conventional systems, a signal is assumed to travel as a flat, planar wavefront. This "far-field" assumption holds true only when the user is far away from the base station. But as antenna surfaces grow larger, the surrounding environment falls within the antenna's near-field region.
Far-Field (Planar Waves) Near-Field (Spherical Waves)
| | | ) ) )
| | | ( ( (
[======Array======] [======Array======]
Linear phase shifts only 3D Beam Focusing Possible
In the near-field, waves are inherently spherical. As explained by Dezhi Wang et al., modeling signals as spherical wavefronts unlocks a new physical capability called beam focusing. Instead of merely steering a signal in a certain direction, a continuous surface can focus electromagnetic energy onto a precise three-dimensional point in space, drastically reducing interference and boosting capacity.
Furthermore, classical models ignore boundary-level physics. When antennas are placed closer than half a wavelength apart, they experience mutual coupling—where energy from one element leaks into another. They also interact with evanescent waves, which are electromagnetic fields that do not propagate into the distance but store energy right at the surface interface.
By treating the channel as a simple matrix of independent random variables, conventional statistical models omit these phenomena. As a result, they significantly underestimate the true data throughput potential of dense holographic arrays, a limitation highlighted by Shuai S. A. Yuan et al..
The Mechanism of Electromagnetic Information Theory
EIT bridges this gap by unifying classical information theory directly with Maxwell's equations. Instead of replacing Shannon's capacity formulas, EIT changes how the channel itself is calculated, moving from discrete matrix abstractions to continuous wave-domain modeling.
The core mechanism relies on Green's functions. A Green's function is a mathematical tool that solves differential equations governing wave propagation; it describes exactly how an electromagnetic field behaves at any point in space when excited by a source. By embedding Green's functions into the information-theoretic framework, the channel is no longer modeled as a collection of random numbers, but as a continuous physical field.
Classical Information Theory:
[Data Source] ---> [Discrete Matrix H] ---> [Capacity Limit]
(Statistical Abstraction)
Electromagnetic Information Theory:
[Data Source] ---> [Continuous Green's Functions] ---> [Capacity Limit]
(Maxwell's Equations Physics)
As Jieao Zhu et al. demonstrate, this framework ensures that the physical layer is redefined by integrating fundamental electromagnetic propagation laws directly into system design. This consistency is vital when dealing with advanced wave properties.
For instance, consider Orbital Angular Momentum (OAM), a technique that twists electromagnetic wavefronts into helical shapes to multiplex data. Traditional scalar models fail to capture the spatial structure of these twisted beams. By applying an EIT lens, Runyu Lyu et al. demonstrate that accounting for the precise physical field details reveals significantly higher calculated channel capacities and yields far more realistic performance predictions.
Overcoming Computational and Design Bottlenecks
If modeling every ripple of an electromagnetic wave is so powerful, why hasn't wireless engineering always done it? The barrier has always been computational complexity. Running full-wave electromagnetic simulations for a complex environment can take hours or days.
EIT addresses this by developing fast, hybrid mathematical frameworks. For example, Ruifeng Li et al. introduced an EMIT model that combines a fast "Group T-matrix" scattering algorithm with mode-decomposition. By mapping physical wave propagation directly to informatic parameters, their approach calculates channel capacity up to 10 times faster than traditional full-wave simulations, giving engineers a practical way to optimize antenna power and phase in real time.
[Physical Wave Propagation]
│
▼ (Group T-Matrix + Mode Decomposition)
[Informatic Parameters]
│
▼ (10x Faster Capacity Calculation)
[Real-Time Optimization]
This physical grounding also changes how we view the wireless environment. Historically, the propagation channel was a passive, adversarial constraint that engineers had to overcome. However, by applying EIT to Reconfigurable Intelligent Meta-surfaces (RIS)—thin, AI-driven electronic skins coated on walls and objects—the environment becomes a programmable asset.
As Marco Di Renzo et al. explain, integrating Maxwell's equations allows operators to shape radio waves at the boundary level. By recycling existing signals and dynamically altering the channel's probability distribution, systems can break the traditional link between higher data rates and increased power consumption.
Reshaping the Wavefront
Ultimately, EIT works because it treats the physical world not as a source of random noise to be mitigated, but as a structured wave-domain computing medium.
When antenna patterns are fixed, they cannot adapt to how physical radiation interacts with environmental scattering paths. By using EIT, designers can decouple complex capacity maximization problems into sequential, physically meaningful steps. As shown by Haonan Wang et al., decoupling the problem into a sequential correlation reduction step followed by a closed-form power allocation step allows pattern-reconfigurable systems to systematically maximize capacity.
Whether it means accounting for arbitrary surface orientations in line-of-sight holographic systems as done by Tierui Gong et al., or embedding physics-informed components into digital twin radio maps as explored by Yashuai Cao et al., grounding information theory in the laws of electromagnetics ensures that our mathematical models finally match the true physical limits of the universe.
Key Foundational Work
Smart radio environments empowered by reconfigurable AI meta-surfaces: an idea whose time has come (2019) — contributed 2 concept elements to this direction.
Active Reconfigurable MIMO Communications: Capacity Maximization Pattern Design (2022) — contributed 1 concept elements to this direction.