Abstract: Terahertz (THz) communication is emerging as a pivotal enabler for 6G and beyond wireless systems owing to its multi-GHz bandwidth. One of its novel applications is in wireless data centers, where it enables ultra-high data rates while enhancing network reconfigurability and scalability. However, due to numerous racks, supporting walls, and densely deployed antennas, the line-of-sight (LoS) path in data centers is often partially obstructed, resulting in "quasi-LoS" propagation and degradation of spectral efficiency. To address this issue, Airy beam-based hybrid beamforming is investigated as a promising technique to mitigate quasi-LoS propagation. This paper proposes a cascaded geometrical and wave-based channel model (CGWCM) for quasi-LoS scenarios and introduces efficient beam search methods. Simulation results validate that Airy beams significantly outperform traditional Gaussian beams in mitigating blockages, verifying their potential for practical THz wireless communication in data centers.
The Challenge of Quasi-LoS in THz Networks
As we move toward 6G, Terahertz (THz) frequencies (0.1–10 THz) are prime candidates for wireless data centers because of their massive bandwidth. However, THz waves behave much like light: they travel in straight lines and are easily blocked by the "clutter" of a data center—server racks, cooling pillars, and moving technicians.
Standard wireless systems use Gaussian beams, which focus energy in a straight line. If an obstacle sits in that path, the link fails. While researchers usually categorize paths as either Line-of-Sight (LoS) or Non-Line-of-Sight (NLoS), data centers often exist in a "Quasi-LoS" state where an object partially clips the beam. In these scenarios, Gaussian beams suffer massive efficiency losses. The challenge is finding a way to "bend" the wireless signal around these obstructions without needing a complex mirror system.
Modeling and Beam Design
The authors tackle this through three main innovations:
- Cascaded Geometrical and Wave-based Channel Model (CGWCM): Existing models are either too simple (Geometrical models ignore diffraction) or too complex (Wave-based models require heavy calculus). CGWCM bridges this gap by using "virtual arrays" to represent blocked regions, turning complex integrals into simple matrix math that is both accurate and fast enough for system-level simulations.
- Airy Beamforming Design: Unlike Gaussian beams, Airy beams possess "self-acceleration" (they curve) and "self-healing" (they can reconstruct their shape after hitting a small obstacle). The authors show that these beams can be generated in a standard hybrid beamforming architecture simply by applying a cubic phase profile to the antenna array.
- Hierarchical Beam Search: Since the exact position of a blockage (like a person walking by) isn't always known, the authors developed a two-stage search. First, the system finds the best focal point (distance and angle), then it "refines the curve" by testing different cubic coefficients to find the path that best dodges the obstacle.
Key Insights: The "Curving" Advantage
The most striking insight is that Airy beams don't just "spread" around an object; they follow a curved trajectory. This allows them to concentrate energy at the receiver even when the direct path is physically blocked. Furthermore, because the energy of an Airy beam is distributed across several "side lobes," if the main lobe is blocked, the side lobes can interfere constructively to "re-grow" the main lobe further down the path—a property known as self-healing.
Key Results
- Model Accuracy: The proposed CGWCM model showed an approximation error 7 dB lower than standard geometric models when blockages were present.
- Spectral Efficiency: In scenarios where more than 50% of the LoS path was blocked, Airy beams outperformed traditional Gaussian beam steering by approximately 3.23 bits/s/Hz.
- Search Efficiency: The Hierarchical Focusing-Airy beam search achieved near-optimal performance with only 156 training overhead, compared to over 1,000 required for an exhaustive search.
- Robustness: Even as the blockage moved closer to the receiver, the Airy beam maintained a stable link, whereas Gaussian beams saw their performance plummet to near-zero.
Conclusion
While Airy beams are primarily a near-field phenomenon suited for the short distances inside a data center (3–10 meters), their ability to mitigate blockages without complex mirror systems makes them a formidable candidate for 6G wireless infrastructure. Future work will need to address hardware realization constraints, such as phase shifter quantization, and multi-user interference in dense, multi-rack environments.
Summary updated