Abstract
As wireless communication moves toward 6G, antenna arrays are evolving into "extremely large" (XL-MIMO) systems, pushing signal propagation into the complex near-field region. This paper analytically identifies the capacity saturation point—the specific array size beyond which adding more antennas yields diminishing returns—and introduces a "Modified ZF" beamforming algorithm. This algorithm utilizes closed-form mathematical expressions to bypass the massive computational overhead typically required for processing thousands of antenna elements.
The Near-Field Challenge in XL-MIMO
In traditional far-field communications, radio waves are assumed to arrive as flat planes. Under this model, increasing the number of antennas ($M$) suggests a linear increase in received power and capacity. However, in the near-field (the Fresnel region), waves are spherical. As antenna arrays grow massive and frequencies climb into the millimeter-wave and Terahertz bands, the far-field model becomes physically inaccurate.
Researchers face two primary hurdles in this regime:
- The Saturation Dilemma: In the near-field, received power does not grow infinitely with $M$; it hits a physical limit. Identifying this "saturation point" is critical to avoid wasting hardware resources.
- The Computational Wall: Standard beamforming techniques like Zero-Forcing (ZF) require calculating inner products of channel vectors. For an XL-MIMO system with tens of thousands of antennas, this matrix math becomes a significant bottleneck.
Analytical Saturation Modeling
The study utilizes the Projected Non-Uniform Spherical Wave (PNUSW) model, which provides higher accuracy for near-field scenarios. By defining a power saturation threshold ($t$) and applying the Stationary Phase Method (SPM), the researchers derived closed-form expressions for the array coordinates that capture the bulk of available signal power.
A key finding is the "Short Board" effect: for a group of users, the saturation point of the entire system is dictated by the outermost users in the distribution. Once the array is large enough to cover the boundaries of the user area, adding more antennas provides almost zero capacity gain. Interestingly, as shown in the simulation results, this physical saturation point remains constant regardless of the Signal-to-Noise Ratio (SNR).
Accelerating Beamforming: The Modified ZF Algorithm
To address the computational complexity of XL-MIMO, the authors proposed a Modified ZF Algorithm. Instead of performing $O(M)$ inner product operations for every beamforming step, they substitute these heavy calculations with derived closed-form approximations for power and correlation. This "accelerated" version relies on simple system parameters like user location and wavelength rather than raw antenna data.
Numerical simulations validated these proofs using a carrier frequency of 100 GHz and up to $2 \times 10^5$ antenna elements. The results demonstrate that the Modified ZF achieves nearly identical capacity performance to the original ZF once the system reaches the saturation point, while significantly reducing the mathematical workload.
Key Insights and Future Directions
- ZF Superiority: While Maximal Ratio Combining (MRC) is often preferred in far-field for its simplicity, user correlations in the near-field do not drop to zero as $M$ increases. This makes ZF significantly more effective at managing interference in XL-MIMO.
- Formula-Based Prediction: The success of the Modified ZF suggests that if a user's general location is known (e.g., via Integrated Sensing and Communication), physics-based formulas can accurately predict matrix values, allowing systems to skip exhaustive channel measurements.
Future work will look to expand these findings beyond Line-of-Sight (LoS) channels into complex multipath environments and transition from Uniform Linear Arrays (ULA) to Uniform Rectangular Arrays (URA) to better reflect practical 6G base station designs.
Summary updated