PaperlayerView exact revision on arXiv

Source-linked AI summary

Terahertz Wireless Channels: A Holistic Survey on Measurement, Modeling, and Analysis

Chong Han, Yiqin Wang, Yuanbo Li, Yi Chen, Naveed A. Abbasi, Thomas Kürner, Andreas F. Molisch

arXiv:2111.04522v2cs.ITeess.SP

TL;DR

Reliable THz communications require understanding propagation channels that underpin system and application design. This article comprehensively reviews channel measurement, modeling, simulation, and characterization, then identifies open research problems. It concludes that no reviewed measurement technique satisfies every high-performance requirement, so reliable THz channel measurement remains an open need.

  • Problem

    Reliable THz communication systems require propagation-channel knowledge, while THz-specific channel measurement, modeling, and characterization remain necessary because THz behavior differs from lower-frequency channels.

  • Method

    The article reviews and compares VNA, correlation-based, and THz-TDS measurement methods; surveys deterministic, stochastic, and hybrid models and simulators; and analyzes THz channel characteristics.

  • Results

    The survey concludes that none of the three measurement techniques meets all high-performance THz channel-measurement requirements, while reviewing channel models, simulators, characteristics, and non-stationary properties.

  • Takeaways & Limitations

    Reliable THz channel measurement and validated modeling remain necessary foundations for future 6G communications, localization, and imaging research.

Abstract

from arXiv · show

Terahertz (0.1-10 THz) communications are envisioned as a key technology for sixth generation (6G) wireless systems. The study of underlying THz wireless propagation channels provides the foundations for the development of reliable THz communication systems and their applications. This article provides a comprehensive overview of the study of THz wireless channels. First, the three most popular THz channel measurement methodologies, namely, frequency-domain channel measurement based on a vector network analyzer (VNA), time-domain channel measurement based on sliding correlation, and time-domain channel measurement based on THz pulses from time-domain spectroscopy (THz-TDS), are introduced and compared. Current channel measurement systems and measurement campaigns are reviewed. Then, existing channel modeling methodologies are categorized into deterministic, stochastic, and hybrid approaches. State-of-the-art THz channel models are analyzed, and the channel simulators that are based on them are introduced. Next, an in-depth review of channel characteristics in the THz band is presented. Finally, open problems and future research directions for research studies on THz wireless channels for 6G are elaborated.

I. INTRODUCTION

THz channels are central to realizing 6G’s ambitious performance targets, but their propagation behavior differs from lower-frequency bands and requires dedicated measurement, modeling, and characterization. This survey reviews those foundations, analyzes THz channel properties, and identifies open research problems.

  • Motivation: 6G systems are expected to target Tbps peak rates, 10–100 Gbps experienced rates, sub-0.1 ms latency, and 100 bps/Hz spectral efficiency.Additional targets include µs-level jitter, 99.99999% success probability for data rate and latency, and millimeter-level positioning and sensing accuracy.
  • Motivation: THz communications span 0.1–10 THz and offer broad bandwidth for applications including Tera-WiFi, Tera-IoT, wireless data centers, and Tera-IAB networks.Current research interest mostly concentrates on the 0.1–0.5 THz range.
  • THz Channel Challenges: THz waves interact with smaller environmental structures and experience molecular and particulate absorption and scattering, producing propagation behaviors distinct from mmWave and cmWave channels.These effects motivate THz-specific channel modeling and parameterization for diverse 6G scenarios.
  • THz Channel Challenges: Ultra-large-scale antenna arrays can increase capacity, but their channel models require higher accuracy because channel statistics may be non-stationary across the array.The paper defines non-stationarity as variation in channel statistics across temporal, spatial, and frequency domains.
  • Survey Scope: The article surveys three measurement methodologies: VNA-based frequency-domain measurement, correlation-based time-domain measurement, and THz-TDS pulse measurement.It also reviews measurement campaigns, deterministic, stochastic, and hybrid modeling approaches, corresponding simulators, channel characteristics, and open problems.

II. CHANNEL SOUNDING METHODOLOGIES AND MEASUREMENT CAMPAIGNS

THz channel sounding uses wideband measurements because channel characteristics vary across frequency, with VNA, sliding-correlation, and emerging multitone approaches trading measurement quality against practical constraints. Directional antennas and positioning systems address link-budget and multipath requirements, while VNA implementations face power, noise, duration, stability, and range limitations.

  • Measurement requirements: Wideband THz sounders are needed because THz signal bandwidth typically exceeds channel coherence bandwidth, producing frequency-dependent channel characteristics.Measurement systems adapt to THz frequencies through frequency conversion, while horn antennas compensate for path loss and conversion limitations.
  • Measurement methodologies: The three reviewed wideband techniques are VNA frequency-domain sounding, sliding-correlation time-domain sounding, and THz-TDS time-domain sounding.The choice trades off bandwidth, speed, distance, power consumption, cost, and system complexity.
  • VNA-based sounding: VNA measurements obtain the channel frequency response from S21 and derive the time-domain impulse response through inverse Fourier transformation.The radio channel, including propagation and antennas, is treated as a two-port device under test.
  • VNA-based sounding: VNAs provide calibrated precision measurements but generally suffer from low output power, high receiver noise, and long measurement duration.Because THz channel coherence time is shorter than practical sweep duration, VNA measurements are restricted to quasi-static scenarios.
  • Alternative sounding: Multitone excitation can reduce measurement time and potentially capture time-varying channels, but requires phase design for low PAPR and has not yet been applied to THz measurements in the cited discussion.The approach remains constrained by device linearity, including power-amplifier limitations.
  • Directional scanning: Directional scanning uses rotating horns or phased arrays to capture multipath components, with phased arrays reducing scan time through fast electronic switching.Suitable phased arrays were not commercially available for THz frequencies at the time described, while increasing distance and antenna gain tighten alignment requirements.

2) Measurement Campaigns:

THz measurement campaigns span short-range line-of-sight, specialized indoor environments, multipath-rich rooms, and longer links enabled by optical-fiber extension. The reviewed methods include VNA systems and correlation-based sounders, with sliding correlation trading sampling rate and measurement duration for SNR.

  • VNA-based campaigns: VNA-based campaigns primarily covered 140–750 GHz using 15–26 dBi directional horns, with Tx/Rx distances typically from 0.1–14 m.RFoF extension increased the maximum reported distance to 100 m.
  • Short-range environments: Short-range campaigns measured line-of-sight channels below 1 m and specialized desktop, motherboard, and data-center environments around 300 GHz.Reported systems included 270–300 GHz measurements over 0.1–1 m and 110–170 GHz measurements across LoS, OLoS, and reflected NLoS settings.
  • Multipath characterization: Fixed-orientation campaigns characterize large-scale path loss and shadowing, whereas increasing distance makes multipath components from multiple directions non-negligible.Directional scanning is therefore applied when spatial multipath characterization is required.
  • Indoor campaigns: A 220 GHz campaign covered 1.8–7.3 m in a meeting room and 2–30 m in an office, including LoS office, LoS hallway, and NLoS cases.The receiver was mounted on a motor-driven rotation unit, and path loss was studied across the indoor scenarios.
  • Extended-range campaigns: Optical-fiber extension expanded a 140 GHz VNA system from an original 8–10 m range to 100 m, enabling medium- and long-distance campaigns.The resulting system supported measurements from 1–35 m and links up to 100 m.
  • Correlation-based campaigns: Correlation-based sounding uses low-PAPR sequences and matched filtering to obtain channel impulse responses, enabling real-time measurements relative to VNA-based methods.Sliding correlation is used to avoid the very high sampling rate required by direct wideband digitization.
  • Correlation-based campaigns: Sliding correlation reduces the required sampling rate by γ while increasing measurement duration and SNR by the same factor.This explicitly trades acquisition speed against sampling burden and signal-to-noise performance.

2) Measurement Campaigns:

THz measurement campaigns span sliding-correlation and THz-TDS systems across several frequency bands and propagation scenarios. THz-TDS offers scalable bandwidth but is mainly used for short-range, narrow-beam measurements.

  • Measurement Campaigns: Sliding-correlation campaigns have covered 140–300 GHz, including indoor, urban microcell, and transportation-related scenarios.Reported examples include directionally resolved measurements and train-to-train, infrastructure-to-infrastructure, train-to-infrastructure, and intra-wagon channels.
  • Measurement Campaigns: THz-TDS directly samples channel impulse responses by transmitting a train of narrow pulses whose period exceeds the channel’s maximum excess delay.Each sampling instance corresponds to the channel response at a delay relative to pulse transmission.
  • Measurement Campaigns: A typical THz-TDS uses a femtosecond laser, beam splitter, THz emitter, mechanical delay line, and detector.The probe beam travels through the delay line, whose path length is swept to align detection with the THz signal.
  • Measurement Campaigns: THz-TDS provides huge, scalable bandwidth but has low output power, a large setup, and a narrow beam, limiting measurements mainly to distances below a few meters.Lenses are typically used to increase pulse intensity, and the narrow beam makes the system more suitable for material characterization than directional scanning.

2) Measurement Campaigns:

The survey compares THz channel measurement systems, campaigns, and methodological trade-offs. Existing measurements concentrate below 300 GHz, while hybrid sounding systems are being explored to balance speed, dynamic range, and scenario requirements.

  • Measurement Campaigns: Most existing THz channel measurements focus below 300 GHz, leaving extensive channel measurements above 300 GHz missing from the literature.Systems have been built at 140, 220, and 300 GHz using VNA, sliding correlation, and THz-TDS technologies.
  • Comparison and Discussion: VNA systems directly measure calibrated channel S-parameters through frequency stepping, providing high time-domain resolution and synchronized transceivers at low measurement-system complexity.Frequency sweeping is time-consuming, and the basic setup can constrain transceiver separation through cable connections.
  • Comparison and Discussion: Sliding-correlation sounders transmit PN sequences and cross-correlate received signals to obtain channel impulse responses, enabling fast instantaneous broadband measurements.Separate synchronization is required, but rubidium-clock synchronization can avoid cable connections; speed, sampling rate, and SNR are tradeable.
  • Comparison and Discussion: THz-TDS offers huge scalable bandwidth but low power, large size, narrow beams, and short measurement distances, while also requiring standardized calibration and analysis.These properties make TDS unsuitable for directional scanning and generally restrict measurements to a few meters.
  • Comparison and Discussion: Multi-mode sounders trade dynamic range against sounding speed, motivating systems that combine temporal and frequency-domain methods for varied THz scenarios.The NYU Wireless sounder combines sliding correlation and real-time spread spectrum; the latter is faster but reduces dynamic range.

A. Overview

THz channel modeling must represent diverse propagation paths, environments, and antenna effects while accounting for frequency-dependent THz behavior. The survey organizes approaches into deterministic, stochastic, and hybrid methods, including ray-tracing refinements for efficiency and scalability.

  • A. Overview: THz channel analysis must model line-of-sight, reflected, scattered, and diffracted paths across static and time-varying environments, potentially including antenna-array effects.These requirements reflect the variety of propagation mechanisms and communication scenarios in the THz band.
  • A. Overview: Physical channel models are categorized as deterministic, stochastic, or hybrid approaches.Deterministic methods solve approximate Maxwell equations and can be accurate but require detailed environmental data and high computational complexity.
  • Ray-Tracing: Ray tracing is site-specific and uses geometric information and electromagnetic material properties to model propagation paths, with approximations such as GO, GTD, UTD, and Kirchhoff theory.Its accuracy depends on the quality of the environmental description and propagation models.
  • Ray-Tracing: A visibility tree represents scenario objects as nodes and line-of-sight connections as branches, with leaves corresponding to paths identified by ray tracing.Backtracking from leaves to the transmitter applies geometric-optics rules to determine each ray path.
  • Ray-Tracing: Ray launching follows rays across directional grids until they leave the area or fall below a strength threshold, efficiently supporting many receivers from one transmitter.It requires longer simulation times for a single transmitter–receiver pair but scales well across a cell.
  • A. Overview: THz propagation models must include spreading loss, molecular absorption, and frequency-, angle-, and polarization-dependent multipath effects.Reflection and scattering models commonly modify Kirchhoff theory, while UTD and Fresnel knife-edge diffraction provide approximations for diffraction.
  • Ray-Tracing: Virtual point approximation reduces multi-antenna ray-tracing cost by assuming subchannels share similar rays and mapping results back to actual antenna positions.A channel matrix can then be obtained from a double-directional channel between ray-traced virtual points.
  • Ray-Tracing: Map-based deterministic models simplify the three-dimensional environment, and their complexity can be adjusted through the number of rays and included propagation mechanisms.The METIS model accounts for line-of-sight, diffraction, specular reflection, diffuse scattering, and blocking.

2) Finite-Domain Time-Domain:

Finite-difference time-domain (FDTD) modeling directly solves Maxwell’s equations and can accurately represent small, complex THz scatterers and rough surfaces. Its fine spatial discretization, however, creates severe memory, time, and computational demands, especially as frequency increases.

  • FDTD, also known as Yee’s method, numerically solves Maxwell’s equations by discretizing electromagnetic fields across space and time.The method partitions space into Yee cells and alternately samples electric and magnetic fields.
  • FDTD can accurately model small, complex scatterers and rough surfaces that are difficult for ray-based methods to represent.Its accuracy is retained in principle under arbitrary conditions, without requiring a calibrated scattering model.
  • FDTD requires very fine grids to resolve the smallest wavelength and geometric feature, increasing memory use and computational time.The required resolution rises with frequency, making THz computations potentially exceed advanced-computer capabilities.
  • Statistical modeling offers a lower-complexity alternative by describing channel behaviors for environment types rather than specific locations.These models support fast channel construction and system simulation but do not reproduce a particular propagation environment.
  • Stochastic models include geometry-based approaches, which randomly place scatterers before applying simplified reflection, diffraction, and scattering laws.Nongeometrical stochastic models instead generate parameters such as DoD, DoA, and delay from probability distributions without modeling the environment.
  • Wideband THz statistical models must represent resolvable multipath components and, for MIMO, include directional information such as DoD, DoA, ToA, and complex amplitudes.Physical models describe double-directional channel statistics, whereas analytical models directly represent channel and antenna impulse responses.

1) Physical Model:

Physical and analytical stochastic models describe THz channels through statistical representations of multipath, angular, delay, and antenna-related behavior. Their reduced complexity supports simulation, but common correlation-based formulations can lose spatial realism and omit THz-specific effects.

  • The Saleh–Valenzuela model represents multipath components as clusters whose arrival times follow Poisson processes and whose powers decay with delay.Spatial extensions additionally model DoA and DoD distributions, including Gaussian-mixture approximations in THz channels.
  • The Zwick model describes individual multipath components by loss, delay, and DoA/DoD, while modeling their temporal appearance and disappearance as a birth–death process.It is designed for indoor scenarios and omits amplitude fading.
  • Analytical stochastic models represent impulse responses between transmit and receive antenna connectors and organize them into channel matrices with statistical correlations.This treatment incorporates both channel and antenna characteristics rather than modeling propagation separately.
  • The Kronecker-based stochastic model assumes transmit- and receive-array correlations are separable, an assumption that becomes less valid with more antennas and dominant single reflections.This limits its fidelity for increasingly large THz MIMO arrays.
  • Correlation-based stochastic models offer low complexity for massive-MIMO evaluation but sacrifice spatial determinism capability.Conventional forms also omit near-field effects and non-stationarity, limiting their suitability for THz ultra-massive MIMO modeling.
  • Hybrid modeling is motivated by the complementary trade-off between deterministic accuracy and statistical efficiency.The surveyed approaches combine individual modeling strategies to balance accuracy and computational cost.

2) Deterministic-Stochastic Hybrid Approach:

Deterministic–stochastic hybrid approaches combine environmental or ray-based structure with statistical generation to improve realism without the full cost of deterministic modeling. Examples include RT–FDTD, quasi-deterministic, geometry-based stochastic, and modified standardized models.

  • Statistical channel models are efficient but cannot easily reproduce spatial consistency or temporal evolution.This motivates hybrid approaches that add geometric or deterministic structure to stochastic generation.
  • RT–FDTD uses FDTD near complex discontinuities and ray tracing elsewhere, addressing the limits of geometrical optics at rough THz surfaces.FDTD resolves small scatterers and rough structures, while RT handles propagation outside those regions.
  • Quasi-deterministic models obtain dominant multipath components from a simplified environmental map and add statistically generated clusters around them.Additional clusters can represent small or mobile scatterers, with angular and delay spreads centered on dominant paths.
  • Hybrid indoor THz models combine geometric or deterministic structure with statistical angular and arrival distributions, including GMM, von Mises, and Poisson formulations.Examples include a 300 GHz indoor model and a semi-deterministic 140 GHz indoor model.
  • Geometry-based stochastic models randomly place scatterers but apply simplified deterministic propagation laws to the resulting paths.Their design depends on scenario-specific scatterer-placement distributions, scatterer counts, and cluster assignments.
  • 3GPP and WINNER models generate multipath parameters stochastically after limited geometric placement, but their usual independent drops omit cross-location correlations.The models also generally keep large-scale parameters unchanged as the mobile station moves, except for partial newer spatial-consistency support.
  • 3GPP validity claims up to 100 GHz rely on few measurements above 6 GHz and omit effects important to mmWave and THz channels.Measurements indicate that one temporal cluster may contain several spatial lobes arriving from diverse angles nearly simultaneously.
  • RT is broadly applicable in THz scenarios, whereas FDTD is more accurate for complex small-scale structures but is limited to small areas such as intra-device channels.Hybrid deterministic methods target acceptable accuracy and low complexity, particularly for indoor and intra-device modeling.

IV. THZ CHANNEL SIMULATOR

THz channel simulators reproduce validated channel characteristics for new scenarios without repeating expensive measurements. The survey covers deterministic, stochastic, hybrid, academic, and commercial tools, alongside measurement campaigns and system-relevant channel characteristics.

  • THz channel simulators use channel models and measurements to reproduce channel characteristics for system evaluation without repeated measurements.Validated models can generate numerous scenarios while avoiding additional expensive and time-consuming campaigns.
  • Deterministic simulators use environmental databases, antenna descriptions, and transmitter–receiver or scatterer trajectories to generate channels.CloudRT combines validated V2V and ultrawideband THz ray-tracing simulators.
  • Planning and ray-launching tools extend deterministic simulation to applications such as 300 GHz backhaul and realistic scenario descriptions.The ThoR project incorporates atmospheric effects in backhaul planning, while another simulator generates deterministic channels through ray launching.
  • Commercial propagation software supports urban, indoor, rural, and mixed environments across frequency ranges extending to 100 GHz.Examples include EDX Advanced Propagation and Wireless InSite.
  • Quadriga supports purely stochastic, 3GPP-compliant, map-based, and geometry-based stochastic models, with parameterization based on measurements up to 80 GHz.NYUSIM supports indoor office and several outdoor environments, with extensions based on measurements up to 150 GHz underway.
  • The survey summarizes state-of-the-art academic and commercial THz channel simulators in a non-exhaustive table.The simulator taxonomy is presented in Table IV.
  • Channel-characteristic analysis supports link-budget, inter-symbol-interference, and non-stationarity evaluations through metrics such as path loss, shadowing, RMS delay spread, and coherence bandwidth.Indoor and outdoor measurement setups, distance ranges, antenna configurations, and campaign data are summarized in Tables V and VI.

1) Path Loss and Shadow Fading:

THz path loss and shadow fading reflect frequency-dependent propagation, atmospheric, weather, geometry, and blockage effects. Measurements show distinct LoS/NLoS and indoor/outdoor patterns, while sampling choices can bias observed path loss.

  • Path Loss: Atmospheric attenuation is more noticeable in THz than microwave or mmWave bands and exhibits several frequency-dependent peaks.ITU-R P.676 evaluates attenuation from oxygen, water vapor, and other factors under specified atmospheric conditions.
  • Path Loss: 20 dB/km rain attenuation remains nearly constant at 50 mm/hr, whereas dense fog reaches 100 dB/km at 330 GHz.The comparison reflects different frequency and particle-size dependencies for rain and fog.
  • Interpretation: Measured path loss is affected by pre-selecting locations where received power is measurable, potentially excluding high-loss outage locations.Deterministic simulations on regular grids can avoid this selection bias.
  • Path Loss: LoS path loss exponents are close to 2, while reported NLoS cases have larger exponents; indoor and outdoor PLE values show no significant difference.Indoor measurements contain substantially more LoS than NLoS cases, whereas outdoor LoS and NLoS counts are similar.
  • Shadow Fading: Shadow-fading standard deviations are generally below 1 dB indoors, larger outdoors and in NLoS cases, and range from nearly zero to over ten dB.The substantial variation is attributed to dependence on detailed propagation geometry.

2) K-factor:

The K-factor measures whether a THz channel is dominated by its strongest path, while related delay, angular, and polarization metrics describe multipath dispersion and diversity. Available measurements indicate scenario- and frequency-dependent behavior, but several trends remain insufficiently established.

  • K-factor: A larger K-factor indicates stronger dominance by a single path and therefore a weaker multipath effect.The strongest path is often, but not always, the LoS path.
  • K-factor: Indoor K-factor values at 300 GHz are smaller than at 140 GHz, but limited results prevent a firm frequency trend.At 300 GHz, outdoor values are slightly larger than indoor values, possibly because enclosed indoor reflections are stronger.
  • Delay Spread: Outdoor THz delay spreads are typically larger than indoor values, while delay spreads decrease as frequency increases.The paper attributes these patterns to longer outdoor reflected-path delays and fewer received multipath components at higher frequencies.
  • Angular Spread: THz angular spreads range from several degrees to tens of degrees, exceeding the typical several-degree beamwidth and allowing directional antennas to reduce multipath effects.Azimuth spreads are generally larger than elevation spreads in the reported scenarios.
  • Cross-polarization Ratio: Large XPR supports orthogonal-polarization multiplexing, whereas small XPR makes polarization diversity useful for link enhancement.Existing THz XPR results find larger values in LoS than NLoS cases, with no clear scenario or frequency trend.

5) Cross-correlation among Channel Characteristics:

THz channel characteristics are cross-correlated, so models should account for joint spatial-temporal behavior rather than treating parameters independently. Reported relationships include strong azimuth coupling and positive delay-angular correlations, while the evidence base remains incomplete.

  • Cross-correlation: Shadow fading and XPR are weakly correlated with other channel parameters, unlike established lower-frequency shadowing relationships with RMS delay spread.The paper identifies this contrast as an observation requiring further investigation.
  • Cross-correlation: DS and AS are positively correlated, with azimuth spreads more correlated with DS than elevation spreads.The paper relates this pattern to their shared dependence on multipath-component powers and limited terrestrial elevation dispersion.
  • Cross-correlation: ASA and ASD are strongly positively correlated, whereas ESA and ESD show weaker correlation in terrestrial scenarios.The reported explanation is that departure and arrival azimuths track the same multipath directions, while elevation dispersion is limited.
  • Comparison Scope: THz comparisons across bands are sensitive to Tx/Rx positions, sounder setup, and post-processing methods, limiting direct interpretation of measured differences.The paper reports alignment with several multiband measurement campaigns despite these sources of variation.

1) Channel Characteristics in Chip-scale Channels:

Chip-scale THz channels involve environment-specific path loss and multipath over very short distances, with structures such as ground planes and DIMM backsides shaping propagation. Characterization remains incomplete beyond a few studied enclosures and motherboards.

  • Path Loss: Chip-scale THz path loss differs from macro-scale behavior because it depends strongly on the specific short-range environment.A chip-to-chip model inside a desktop-size metal enclosure includes multiple physical terms.
  • Multi-path Effect: Ground planes and parallel-plate structures introduce multipaths in motherboard chip-to-chip channels, while DIMM backsides can provide strong reflections.These nearby structures materially shape the short-distance multipath environment.
  • Open Characterization: Chip-scale THz delay spread and angular spread remain unexplored, although both are important for system design.Existing studies mainly address path loss and multipath in computer motherboards or metal enclosures.
  • Nano-scale Channels: Nano-scale THz path loss combines spreading, scattering, and molecular absorption losses, with absorption depending on the medium and dielectric properties.Spreading follows free-space propagation, scattering arises from particles and biological structures, and absorption is characterized using Beer-Lambert law.
  • Nano-scale Channels: An empirical human-skin path-loss model depends on frequency, distance, and sweat-duct count and is validated only within the considered scenario.Further studies are needed to verify its generality.
  • Open Characterization: Existing nano-scale THz channel studies mainly rely on analytical methods, so experimental measurements are needed to validate theoretical results.This limitation applies to the current characterization evidence base.

VI. OPEN PROBLEMS AND FUTURE DIRECTIONS

THz channel research still requires better measurement systems, broader campaigns, validated simulators, and models that capture large-array and near-field effects across diverse scenarios.

  • Measurement systems: High-performance THz measurements must extend toward 10 THz, provide tens of GHz bandwidth, and retain sufficient sensitivity and dynamic range for large path losses.The requirements also include accurate phase measurement and support for narrow-beam scenarios.
  • Measurement campaigns: Measurement campaigns should cover UAV, ship, vehicle, indoor, outdoor hotspot, and inter-chip scenarios across 0.1-1 THz despite their cost and complexity.The paper calls for collaboration because these setups are difficult, time-consuming, and expensive.
  • Simulators: Ray-tracing simulators can supplement campaigns with statistically larger results, but require extensive measurements to validate material properties and support intelligent reflecting surfaces.The paper identifies accuracy, stability, and efficiency as fundamental simulator requirements.
  • Channel models: Hybrid channel models seek balanced accuracy and low complexity, while accurate electromagnetic modeling of diffuse non-line-of-sight scattering remains missing.Deterministic propagation environments should be selected according to the communication scenario.
  • Large arrays and IRS: Very large arrays and intelligent reflecting surfaces create modeling challenges involving mutual coupling, antenna spacing, frequency, and antenna technology.The paper specifically calls for efficient coupling models among proximal antennas rather than modeling every antenna pair.
  • Near-field effects: Near-field effects arise below the increased Rayleigh distance of THz very large arrays, requiring spherical-wave or nonconstant-angle models and experimental characterization.Most existing investigations rely on simplifying modeling assumptions.

1) Temporal non-stationarity:

THz channels require further study of temporal, frequency, and spatial non-stationarity, alongside AI-assisted analysis and future standardization of channel models.

  • Temporal non-stationarity: Temporal birth and death of multipath components requires validation of microwave- and mmWave-derived models through THz measurements or simulations.Vehicular, railway, and UAV time-varying channels are identified as relevant settings.
  • Frequency non-stationarity: Ultra-wideband THz channels need measurements of stationarity and coherence bandwidth to determine frequency selectivity and physical-layer design requirements.The paper calls for additional ultra-wideband measurement campaigns in the THz band.
  • Spatial non-stationarity: UM-MIMO introduces spatial non-stationarity, but effective multipath birth-and-death models and representative array stationarity distances remain missing.These quantities are needed for typical indoor and outdoor scenarios.
  • AI-powered analysis: AI techniques are expected to support THz channel modeling, including multipath clustering and path-loss prediction, because unique propagation features create modeling challenges.The cited discussion positions robust intelligent algorithms and models as an area for further effort.
  • Standardization: Future THz standardization is expected to specify additional application scenarios and expand channel-model support beyond current hotspot-focused 5G standardization.IEEE 802.15.3d-2017 already standardized 100 Gbps wireless communication at 252-321 GHz.
Loading 2111.04522v2…