Terahertz Wireless Channels

Updated 11 March 2026

Terahertz wireless channels are defined by unique propagation physics—including free-space spreading, molecular absorption, and weather-related losses—that distinguish them from lower-frequency regimes.
They are modeled using deterministic, stochastic, and hybrid approaches, with measurement techniques like THz-TDS capturing their sparse multipath and frequency-selective characteristics.
Insights from these channels guide 6G system design, emphasizing adaptive beamforming, advanced modulation, and error correction to tackle path loss and environmental effects.

Terahertz wireless channels denote the propagation environments and their associated electromagnetic characteristics for signals in the terahertz frequency range, typically 0.1–10 THz. They are distinguished from microwave and millimeter-wave regimes by unique propagation physics, statistical channel properties, and pronounced implications for 6G and beyond system architectures. This article surveys their essential features, mathematical models, empirical findings, and their central role in emerging wireless systems.

1. Fundamental Propagation Phenomena

Terahertz propagation is governed by three dominant mechanisms: free-space spreading, molecular (atmospheric) absorption, and weather/surface interactions (Han et al., 2019). The total path loss for a link of length $d$ at frequency $f$ is

$L(f,d) = 20\log_{10}\!\Bigl(\frac{4\pi f d}{c}\Bigr) + 4.343\,k_a(f)\,d,$

where $k_a(f)$ [m⁻¹] denotes the molecular absorption coefficient, composed primarily of H₂O and O₂ resonance lines derived from spectroscopic databases (e.g., HITRAN) (Han et al., 2019). Spreading loss grows quadratically with frequency and distance; at 300 GHz over 10 m, $L_{\rm spreading}\approx 100$ dB. $k_a(f)$ exhibits deep frequency selectivity, with "windows" of low absorption and strong attenuation at line centers (e.g., $k_a$ as low as $10^{-4}$ – $10^{-3}$ m⁻¹ near 300 GHz, exceeding $10^{-1}$ m⁻¹ at resonance lines).

Weather effects further impose frequency-dependent losses. Diffuse scattering increases when surface roughness approaches the wavelength, while specular reflections remain dominant but are diminished due to material electrical thickness. Diffraction is negligible beyond several GHz, and THz links experience higher shadowing variance, somewhat milder scintillation (atmospheric turbulence) than optical links, and susceptibility to Mie/Rayleigh scattering in fog or rain (Han et al., 2019).

2. Channel Modeling and Statistical Characterization

2.1 Large-Scale Path Loss and Absorption

The large-scale loss is often modeled as

$L(f,d) = 20\log_{10}\!\Bigl(\frac{4\pi f\,d}{c}\Bigr) + 4.343\int_0^d k_a(f, \ell)\,d\ell,$

where $k_a$ is allowed to vary along the path with environmental parameters (Han et al., 2019). At intermediate and high altitudes, particularly for aerospace and satellite scenarios, molecular absorption drops sharply as H₂O density decays exponentially, rapidly widening spectral windows (Gao et al., 25 Feb 2025, Kokkoniemi et al., 2020).

2.2 Channel Impulse Response and Transfer Function

THz impulse responses are notably sparse. For indoor and short-range LoS, a tapped delay line with $N_{\rm paths}$ often $1$–$5$ suffices (Han et al., 2019): $h(t;f) = \sum_{n=0}^{N_{\rm paths}-1} \alpha_n(f) e^{-j2\pi f \tau_n} \delta(t-\tau_n),$ with $\alpha_n(f)$ absorbing both path-loss and frequency-dependent absorption. The frequency response $H(f)$ thus becomes highly selective over wide bands and generally smooth over narrow bands (Han et al., 2019, Zhao et al., 5 Oct 2025).

2.3 Small-Scale Fading and Parameter Statistics

The dominant LoS component at THz yields Rician fading with high $K$ -factor ( $K > 10$ dB), which can be reduced to Nakagami- $m$ (with $m\approx 2$ –$4$) or Weibull distributions under severe misalignment or blockage (Han et al., 2019, Li et al., 2024, Ge et al., 23 Jun 2025). For example, indoor campaigns have measured $K$ -factors in the $10$–$20$ dB range, with delay spreads typically $0.1$–$1$ ns in LoS, and coherence bandwidths $B_c \approx 160$ –$1600$ MHz (Han et al., 2019, Zhao et al., 5 Oct 2025).

Antenna misalignment emerges as a dominant impairment, causing path loss to deviate significantly even over a few degrees of error (Karakoca et al., 2023, Ekti et al., 2017). At 140 GHz, a 10° Tx tilt incurs 2–3 dB additional loss, escalating to 13 dB at 20° (Ekti et al., 2017).

3. Environmental and Scenario-Dependent Effects

3.1 Atmospheric and Weather Effects

Molecular absorption loss and scattering by rain, fog, and aerosols generally restrict practical THz links to short- and mid-range under standard terrestrial atmospheric conditions (Ma et al., 2024, Han et al., 2019). Scattering loss may be modeled as

$L_{\rm scat}(f,d) = 10\log_{10} e^{k_{\rm scat}(f) d}$

with $k_{\rm scat}$ informed by Mie theory and empirically correlated to particle size distributions (e.g., Marshall–Palmer for rain). Empirical data indicate $L_{\rm abs} \sim 145$ dB/km at 300 GHz under typical air, $L_{\rm scat}$ for rain at 625 GHz may reach $170$ dB/km at 100 mm/h (Ma et al., 2024).

Adverse weather, such as heavy snow at 140 GHz, can introduce additional path loss of $\sim$ 13 dB over a 70 m link but generally does not interrupt well-designed 6G links employing adaptive margin and powerful error correction (Sen et al., 2022).

3.2 Water/Vapor Dynamics and Dynamic Surfaces

Links traversing or reflecting from water surfaces are heavily modulated by dynamic reflectivity effects, best captured by dual-scale scattering models such as I2EM that incorporate both macroscopic (wave) and microscopic (roughness) surface features (Ge et al., 23 Jun 2025). Reflection coefficients are decomposed into coherent and diffuse terms, with statistical fading over water being best fit by Weibull distributions. For instance, increased surface wave height and frequency cause up to 15 dB mean loss increase and broader SNR variance (Ge et al., 23 Jun 2025).

Dynamic environments necessitate BER modeling under these conditions (e.g., using integrated BER expressions over the Weibull SNR distribution), calling for modulation/coding adaptive to instantaneous channel statistics.

4. Channel Modeling Methodologies and Tools

4.1 Measurement Techniques

Frequency-domain sounders using VNAs, sliding-correlation time-domain setups, and THz time-domain spectroscopy (THz-TDS) are the main modalities, each with tradeoffs in bandwidth, dynamic range, and temporal/spatial resolution (Han et al., 2021). Most campaigns focus on the 100–300 GHz region due to equipment limitations, with dynamic range tapering above 0.5 THz (Han et al., 2021).

4.2 Modeling Approaches

Deterministic: Ray-tracing (incorporating frequency-dependent absorption and material properties), finite-difference time-domain (FDTD) for local regions with roughness.
Stochastic/Statistical: Extensions of Saleh–Valenzuela cluster models for sparse MPCs, geometry-based stochastic models (GBSMs) for mobile/A2G/UAV links, and heavy-tailed fading mixture models (e.g., Dirichlet process Gamma mixture models) for fine-grained SNR histograms in micro-scale scenarios (Karakoca et al., 2022, Karakoca et al., 2023, Li et al., 2024).
Hybrid: Quasi-deterministic schemes and map-based stochastic models that combine deterministic modeling for strong paths with statistical filling of diffuse components (Han et al., 2021, Tarboush et al., 2021).

Full-system simulators such as TeraMIMO implement these physical and statistical models, including ultra-massive MIMO, beam split, near-/far-field propagation, and misalignment fading (Tarboush et al., 2021).

4.3 Atmospheric Dispersion

Ultra-wideband THz communication is limited by group-velocity dispersion, modeled through refractivity and group delay derived from spectroscopic lines. Compensation schemes, such as stratified-media Gires–Tournois Interferometer cohorts, restore waveform fidelity with near-unity in-band efficiency (Strecker et al., 2019).

5. Channel Parameters: Empirical Ranges and System Design Implications

Channel Parameter	Typical Range	Reference Frequency or Context	Reference Papers
Spreading Loss	~100 dB (10 m, 300 GHz)	Indoor LoS	(Han et al., 2019)
$k_a(f)$ (Absorption)	$10^{-4}$ – $10^{-1}$ m⁻¹	0.1–1 THz, function of humidity/Temp.	(Han et al., 2019, Gao et al., 25 Feb 2025)
RMS Delay Spread	0.1–1 ns (indoor), <0.5 ns (outdoor)	Quasi-static THz channels	(Han et al., 2019, Zhao et al., 5 Oct 2025)
Coherence Bandwidth	160–1600 MHz (typical)	Inverse of $\tau_{\rm rms}$	(Han et al., 2019)
Angular Spread	$<10^\circ$ (LoS), higher in NLoS	Sparse reflections	(Han et al., 2019)
Fading Distribution	Rician (high $K$ ), Weibull, Nakagami	LoS, hovering UAV, water reflection	(Li et al., 2024, Ge et al., 23 Jun 2025)
Beam Misalignment Loss	2–13 dB per 10°–20° tilt	Short links, horn antennas	(Ekti et al., 2017)

Coherence times are $\sim$ 1 ms (pedestrian, 300 GHz), implying quasi-static channels for packet-scale protocols in static environments but necessitating fast tracking for mobile/UAV/aerospace (Gao et al., 25 Feb 2025, Li et al., 2024).

Design implications:

Antenna arrays: Ultra-massive MIMO (>10⁴ elements), razor-sharp hybrid beamforming to overcome path loss and maintain directional integrity (Han et al., 2019, Tarboush et al., 2021).
Modulation/coding: Pulse-based schemes (OOK, PPM) for short-range; adaptive multi-wideband OFDM aligned with spectral windows at longer ranges; FEC with LDPC, polar, or turbo codes tailored for frequency-selective errors (Han et al., 2019).
MAC/PHY: Protocols address deafness, beam alignment, and leverage frequency–distance selectivity, supporting non-orthogonal access (Han et al., 2019, Tarboush et al., 2021).
Physical-layer security: Differential path loss and absorption sharply localize eavesdropper threat zones, suggesting phy-layer schemes based on artificial-noise beamforming, frequency-hopping, and secure-zone radii (Ma et al., 2024).

6. Special Scenarios: Aerial, UAV, and Water Surface Channels

Aerospace (A2S, A2G, A2A): At cruise altitudes, molecular absorption windows open, supporting 20–120 Gbit/s over 5–1000 km; at ground level, water-vapor absorption sharply limits range and mandates window selection (e.g., 300–380 GHz, 600–700 GHz under clear air) (Gao et al., 25 Feb 2025, Kokkoniemi et al., 2020).
UAV Links: Ground-to-UAV measurements at 140 GHz reveal path-loss exponents $n\sim1.7$ , Rician to Weibull small-scale fading, and hovering jitter requiring beam-tracking rates of 100 Hz or higher for link reliability (Li et al., 2024, Gao et al., 2023).
Water-surfaces: Reflection channels over dynamic aquatic surfaces must model fast randomization of specular/diffuse components, with Weibull fading providing best statistical fit (laboratory and field) (Ge et al., 23 Jun 2025).

7. Open Research Problems and Future Directions

Extending measurement and modeling beyond 300 GHz with accurate, dynamic, and spatially consistent models is a critical challenge. Non-stationarity induced by ultra-massive MIMO, adaptive intelligent reflecting surfaces (IRS), and ultra-wideband operation calls for new physical/statistical modeling frameworks (Han et al., 2021, Song et al., 6 May 2025). Hybrid data-driven approaches, leveraging AI/ML for real-time parametric estimation and efficient channel state inference from sparse observation (e.g., radio radiance field fusion), show promise for scaling THz system design to complex, dynamic environments (Song et al., 6 May 2025).

Efforts to standardize channel models, establish open-data repositories, and develop AI-accelerated simulators are pivotal for the convergence of research and practical THz wireless deployment.

References: