6G Sub-Terahertz Wireless Communication

• 6G sub-terahertz wireless communication is defined for the 90–300 GHz band, delivering Tbps data rates and sub-millisecond latencies for high-capacity networks.
• It employs advanced physical-layer techniques including high-order modulations, hybrid beamforming, and adaptive sub-band selection to overcome propagation challenges like path loss and molecular absorption.
• Emerging hardware innovations and wavefront engineering methods enable practical implementation in diverse applications such as wireless backhaul, immersive AR/VR, and chip-to-chip links.

Sub-terahertz (sub-THz) wireless communication, typically defined for the 90–300 GHz band, is a central pillar of projected sixth-generation (6G) wireless systems. Sub-THz communication promises Tbps-class data rates, sub-millisecond latencies, and supports new application domains such as wireless backhaul, data center interconnects, immersive extended reality, and ultra-high-resolution sensing. This regime leverages the vast, underutilized spectral resources between conventional millimeter wave and the lower end of the terahertz band, exploiting unique propagation windows and deploying advanced physical-layer solutions to overcome formidable challenges from path loss, molecular absorption, and hardware constraints (Bicaïs et al., 2022).

1. Physical Layer Paradigms and Modulation

Two complementary physical-layer (PHY) design paradigms dominate sub-THz 6G research: (1) spectral-efficiency maximization and (2) complexity/power minimization (Bicaïs et al., 2022).

Spectral-Efficiency-Oriented PHY aims at maximizing SE [bit/s/Hz] through high-order modulations (e.g., QAM, OFDM), multi-carrier signaling, and full-digital MIMO. This paradigm demands RF chains with high linearity, strict phase-noise control (requirement: PN PSD ≤ –100 dBc/Hz at >100 GHz), and high-resolution, multi-GS/s DAC/ADCs. Under ideal channel and hardware, the peak SE is given by $SE = \log_{2}(1 + SNR)$, with practical rates capped by hardware impairments. Data rates are set by $R = BW \cdot SE$. For example, optimized single-carrier constellations (e.g. Polar-QAM) achieve SE ≈ 5.5 bit/s/Hz at 30 dB SNR even with strong phase noise, surpassing standard QAM (2.5 bit/s/Hz) (Bicaïs et al., 2022).

Low-Complexity/Low-Power PHY minimizes device cost and energy, critical for small cell, device-to-device (D2D), and hotspot scenarios. These exploit single-carrier waveforms with low PAPR, impulse-radio, various index-modulation (IM) schemes (e.g., filter-shape IM, or FSIM), and non-coherent detection. IM architectures embed additional bits into dimensions such as pulse shape, subcarrier selection, or antenna index, yielding $SE_{FSIM} = \log_{2} M + \log_{2} L$ where $L$ is the number of pulse shapes (Bicaïs et al., 2022).

Trade-off: By tuning modulation order, number of RF chains, and IM dimensionality, SE vs. hardware complexity and power can be navigated to fit use-case constraints.

Molecular Absorption Adaptation: Sub-THz channels are affected by deep molecular absorption lines (e.g., from water vapor), prompting adaptive sub-band selection—avoiding frequencies with elevated attenuation (Bicaïs et al., 2022, Shafie et al., 2022).

2. Propagation, Channel Modeling, and Link Budget

Sub-THz propagation is dominated by free-space path loss and pronounced molecular absorption. The total path loss in dB combines spreading and absorption:

$$PL(f,d)[\mathrm{dB}] = 20\log_{10}(4\pi f d/c) + 4.34 \kappa(f) d,$$

where $\kappa(f)$ is the absorption coefficient (m⁻¹), highly frequency-selective due to atmospheric H₂O and O₂ resonances (Bicaïs et al., 2022, Han et al., 2019, Yalavarthi et al., 13 Mar 2025). For example, at 300 GHz over 10 m, path loss is ≈ 100–120 dB. Empirical measurements confirm this model in both indoor (Ju et al., 2022) and outdoor (Bicaïs et al., 2022) scenarios.

Sparsity and Blockage: Sub-THz channels are often LoS-dominated, with few resolvable paths and minimal diffuse scattering. Blockage by objects (walls, foliage) causes severe attenuation, while diffraction is negligible (Shafie et al., 2022).

Absorption Windows: Usable sub-THz bands (“transmission windows”) interleaved with absorption lines are identified through spectroscopy. For instance, in the 1–3 THz range (with 35% RH), seven 100 GHz windows were measured, supporting propagation distances from 5 m to 210 m at SNR >30 dB with 100 GHz beamformed antennas, each enabling ≈1 Tb/s channel capacity (Yalavarthi et al., 13 Mar 2025).

3. Antennas, Beamforming, and Wavefront Engineering

High-gain, pencil-beam antennas and hybrid beamforming architectures are essential for sub-THz links (Bicaïs et al., 2022, Shafie et al., 2022). The gain of a planar array is $G \approx 4\pi A/\lambda^2$, enabling commercial arrays to reach 40–70 dBi. The beamwidth (half-power) scales as $\theta_{3dB} \approx 0.886\lambda/D$.

Beamforming Strategies:

• Fully digital: One RF chain per antenna, offering maximal spatial multiplexing but infeasible in power/complexity for large N at sub-THz.
• Hybrid: Digital baseband with analog phase-shifters or true-time-delay lines, allows multi-user MIMO with reduced RF chain count. Widely-spaced subarray architectures mitigate channel correlation and restore multiplexing (Shafie et al., 2022, Han et al., 2023).
• Analog: Single RF chain with phase network, low power, single-beam only.

Wavefront Engineering: The massive near-field effect (large array, small λ, link $d < d_F = 2 D^2/\lambda$) necessitates aperture phase profiles beyond planar steering, including near-field beamfocusing (spherical profiling), generation of Bessel and Airy beams (nondiffracting, self-healing, obstacle-avoiding), and OAM multiplexing for spatial channels (Singh et al., 2023). For example, Bessel beams maintain nearly constant gain over their depth-of-field (~20 m at 1 THz), and 32 OAM modes at 1 m provide 1 Tbps with only 3–10 GHz bandwidth (Singh et al., 2023).

Polarization Multiplexing: Empirical factory measurements demonstrate gross cross-polarization discrimination XPD ≈ 27.7 dB (mean), enabling dual-polarized arrays to implement nearly orthogonal channels (Ju et al., 2022).

4. Hardware Technologies and Implementation Challenges

Semiconductor Frontends: InP/InGaAs HEMTs, GaN HEMTs, SiGe BiCMOS, 28 nm CMOS, and FD-SOI platforms are all employed (Tataria et al., 2021, Lee et al., 20 May 2025). Critical figures: f_max of 300–700 GHz, PA output power up to 32 dBm at 100 GHz (decreasing at higher frequencies), power-added efficiencies (PAE) 3–24%, and noise figures (NF) ≈ 3–15 dB.

Key PA Architectures in sub-THz CMOS:

• Common-source (multi-stage, cascode, pseudo-differential)
• Stacked-FET for voltage swing
• Transformer-based power combining (on-chip DAT, coupled lines)
• Passive gain-boosting, load-pull, and neutralization techniques for bandwidth and linearity (Lee et al., 20 May 2025).

ADC/DAC bottlenecks: With 1+ GHz bandwidths, converters require >2 GS/s at medium resolution—expensive and power-hungry (Bicaïs et al., 2022). RF parallelization and channel bonding are solutions.

Phase Noise: Severe at f >100 GHz, limiting high-order modulations. Optimized SC waveforms (e.g. Polar-QAM), pilot-aided estimation, and envelope detection in non-coherent schemes mitigate impairment (Bicaïs et al., 2022).

Emerging Graphene Receivers: Zero-bias, CMOS-compatible, <1 mm² area, supporting up to 3 Gbit/s over ~3 m at 0.2–0.3 THz with responsivity R = 0.16 A/W and NEP ≈ 60 pW/√Hz. These are promising for ultra-low SWaP frontends in dense device-to-device and chip-to-chip links (Soundarapandian et al., 2024).

5. Experimental Demonstrations and Standards

IEEE 802.15.3d Standard: Defines P2P links in 252–322 GHz, channel bandwidths up to 69 GHz, and up to 315 Gbit/s using 64-QAM at 69 GHz. PHY supports single-carrier and OOK. Maximum spectrally efficient range for 100 Gbit/s is ~100 m with 25 dBm TX power and >30 dBi antennas (Petrov et al., 2020).

Photonic Sub-THz Links Above 300 GHz: Demonstrations at 560 GHz use Kerr micro-resonator soliton microcombs for <100 kHz carrier linewidth with injection-locked DFB lasers, photomixed in UTC-PDs. Recent records include:

• 112 Gbit/s at 560 GHz (16QAM, 28 GBaud, BER ≈ 3×10⁻³) (Tokizane et al., 20 Oct 2025).
• Near-error-free OOK at 1 Gbit/s, EVM 8.1% (compliant with IEEE 802.15.3d), and robust BPSK/QPSK at 1 GBaud (Tokizane et al., 2023).
• 2 Gbit/s OOK over 0.6 m at 560 GHz with Kerr-microcomb source and UTC-PD (Tokizane et al., 2023).

Time-Reversal Multiple Access at 273 GHz: Multi-user focusing, 3 mm Rx separation, 2 GHz bandwidth, 343 Mbps per user, enabled by TR-precoding and non-coherent PPM, highlights path to high-rate, low-power IoT links (Mokh et al., 2022).

6. System and Architectural Considerations

Backhaul: Multi-hop sub-THz–FSO hybrid chains offer tens of Gbps with five-nines (99.999%) availability when hop lengths ≈ 200–300 m, combining the weather-diversity of sub-THz (opaque to fog/cloud) and FSO (opaque to rain) (Singya et al., 2023).

Dynamic Spectrum Management: Integrated photonic real-time spectrum sensing on TFLN enables ultrawide (~120 GHz) sub-100 ns latency channel state measurement for agile spectrum allocation (ISAC) (Tao et al., 4 Sep 2025).

Cross-Field Channel Modeling: Hybrid spherical-planar models and widely-spaced subarray hybrid beamforming allow accurate, efficient MIMO processing across near/far-field, enabling 3×–5× SE gains and accurate channel estimation at realistic pilot overheads (Han et al., 2023).

Network Layer & MAC: Severe “deafness,” blockage susceptibility, and highly directional cell architectures call for beam-aware MACs, out-of-band discovery and control, and fast handover protocols. Multi-connectivity and distributed, AI-driven resource allocation will be required for robustness (Shafie et al., 2022, Han et al., 2019).

Deployment Use Cases: Include wireless backhaul (100+ Gbit/s, 100 m), ultra-dense indoor access (hotspots, AR/VR, >10 Gbit/s), server/rack interconnects (multi-100 Gbit/s at 10–20 m), wireless chip-to-chip links (multi-Gbit/s, cm range), and time-critical sensing (sub-ms latency, super-resolution positioning) (Petrov et al., 2020, Soundarapandian et al., 2024, Ju et al., 2022).

7. Open Challenges and Future Directions

• Hardware scaling: Pushing CMOS, SiGe, and III–V PAs and LNAs above 300 GHz with acceptable PAE, output power, and device reliability remains a priority (Lee et al., 20 May 2025).
• Phase-noise-tolerant waveforms: SC and advanced IM designs, as well as phase-noise-aware coding, are critical at extreme frequencies (Bicaïs et al., 2022).
• Energy efficiency: Large-N arrays and high-resolution ADC/DAC pose severe power challenges. Mixed-architecture beamforming and energy-proportional transceiver design are active areas (Shafie et al., 2022, Tataria et al., 2021).
• Wavefront and beam management: Scalable, low-loss true-time-delay networks for array phase control are required for broadband, near-field, and multi-user beams (Singh et al., 2023).
• Cross-layer integration: Close coordination among waveform, MIMO architecture, RF/hardware, MAC, and network layers, with dynamic adaptation to channel and application context, is necessary for sustained Tbps 6G performance.
• Standardization and coexistence: Ongoing IEEE and ITU-R activities are defining regulatory, coexistence, and spectrum-sharing frameworks for the 140–450+ GHz band (Petrov et al., 2020).
• Advanced channel models: Further empirical measurement and modeling—especially in industrial, urban, and dynamic environments—and extending to OAM and non-Gaussian beams are required for robust system design (Ju et al., 2022, Singh et al., 2023).
• Device integration: Monolithic CMOS/graphene, chip-scale photonic, and packaged sub-THz transceivers, with on-chip antenna and lens arrays, are critical enablers for mass-market adoption (Soundarapandian et al., 2024).

In summary, 6G sub-terahertz wireless embodies richly cross-disciplinary advances: from new frontiers in semiconductor and photonic device physics to innovation in spatial-spectral waveform design, channel modeling, large-array architectures, and intelligent multi-layer networking. Realizing this vision requires sustained co-design across propagation, hardware, algorithms, and distributed protocols, exploiting the unique properties of sub-THz channels while overcoming their distinctive challenges (Bicaïs et al., 2022, Singh et al., 2023, Shafie et al., 2022, Soundarapandian et al., 2024, Lee et al., 20 May 2025, Han et al., 2023).