PaperlayerView exact revision on arXiv

Source-linked AI summary

Seven Defining Features of Terahertz (THz) Wireless Systems: A Fellowship of Communication and Sensing

Christina Chaccour, Mehdi Naderi Soorki, Walid Saad, Mehdi Bennis, Petar Popovski, Merouane Debbah

arXiv:2102.07668v2cs.IT

TL;DR

THz systems offer abundant bandwidth and high-resolution sensing but are constrained by uncertain, short-range, blockage-sensitive channels and molecular absorption. This paper analyzes these constraints through seven defining features and proposes a forward-looking deployment roadmap. Its supported conclusion is that THz communication and sensing should be addressed together, with architectures and techniques tailored to the band’s distinctive conditions.

  • Problem

    THz links must provide high-rate communication and sensing despite short range, blockage, molecular absorption, sparse channels, and rapidly varying channel state information.

  • Method

    The paper presents a holistic roadmap covering seven THz system features, including tailored architectures, joint sensing and communication, PHY procedures, spectrum access, and network optimization.

  • Results

    The paper concludes that THz systems can combine high-rate communication with high-resolution sensing, while RISs, NOMA, sensing, and learning-based approaches address distinctive connectivity and optimization challenges.

  • Takeaways & Limitations

    THz deployment requires opportunistic, distributed, RIS-enabled architectures that exploit sensing and cooperation with lower-frequency links to improve coverage, reliability, and scalability.

Abstract

from arXiv · show

Wireless communication at the terahertz (THz) frequency bands (0.1-10THz) is viewed as one of the cornerstones of tomorrow's 6G wireless systems. Owing to the large amount of available bandwidth, THz frequencies can potentially provide wireless capacity performance gains and enable high-resolution sensing. However, operating a wireless system at the THz-band is limited by a highly uncertain channel. Effectively, these channel limitations lead to unreliable intermittent links as a result of a short communication range, and a high susceptibility to blockage and molecular absorption. Consequently, such impediments could disrupt the THz band's promise of high-rate communications and high-resolution sensing capabilities. In this context, this paper panoramically examines the steps needed to efficiently deploy and operate next-generation THz wireless systems that will synergistically support a fellowship of communication and sensing services. For this purpose, we first set the stage by describing the fundamentals of the THz frequency band. Based on these fundamentals, we characterize seven unique defining features of THz wireless systems: 1) Quasi-opticality of the band, 2) THz-tailored wireless architectures, 3) Synergy with lower frequency bands, 4) Joint sensing and communication systems, 5) PHY-layer procedures, 6) Spectrum access techniques, and 7) Real-time network optimization. These seven defining features allow us to shed light on how to re-engineer wireless systems as we know them today so as to make them ready to support THz bands. Furthermore, these features highlight how THz systems turn every communication challenge into a sensing opportunity. Ultimately, the goal of this article is to chart a forward-looking roadmap that exposes the necessary solutions and milestones for enabling THz frequencies to realize their potential as a game changer for next-generation wireless systems.

I. INTRODUCTION

THz wireless systems promise abundant bandwidth and high-resolution sensing for demanding 6G services, but uncertain channels, hardware barriers, short range, blockage, and molecular absorption complicate deployment. The paper develops a holistic treatment of seven defining features and a roadmap for communication–sensing systems.

  • 6G applications require a 1000× capacity increase over expected 5G systems and multi-purpose communication, sensing, localization, and control functions.
  • The THz gap arose from inefficient transceivers and antennas because semiconductor devices cannot effectively convert electrical energy into electromagnetic energy at THz frequencies.
  • The paper uses “THz” for the combined 0.1–0.3 THz sub-THz and 0.1–10 THz ranges, while noting that frequencies above 275 GHz exhibit unique THz properties.
  • THz systems face highly varying channels, short-range links, narrowbeam LoS dependence, and modeling, analysis, design, and optimization challenges for real-world IoE deployment.
  • The sub-6 GHz–THz spectrum is becoming jointly populated by communication and sensing, with lower bands offering longer range and reliability and higher bands offering greater spatial resolution.
  • The article examines seven defining features, including quasi-opticality, THz-tailored architectures, lower-band synergy, joint sensing and communication, PHY procedures, spectrum access, and real-time optimization.

2) THz-tailored architectures:

THz-tailored architectures must accommodate dense small-base-station deployments, short range, multifunctionality, and unique channel conditions. The paper connects these architectures with lower-frequency cooperation, joint sensing and communication, PHY-layer adaptation, and application-specific use cases.

  • THz-tailored architectures: THz deployments require higher small-base-station density, shorter communication range, multifunctionality, and architectures adapted to unique channel conditions.The paper emphasizes opportunistic THz-tailored architectures, including cell-less designs and their accompanying challenges.
  • Synergy with lower frequency bands: THz systems are expected to cooperate and coexist seamlessly with sub-6 GHz and mmWave technologies in already populated spectrum.Immersive remote presence is presented as a use case that could opportunistically use all available wireless frequencies for its end-to-end experience.
  • Joint sensing and communication: The quasi-optical THz band supports a fellowship of high-rate communication and high-resolution sensing through joint communication-and-sensing systems.The paper emphasizes mutual feedback between sensing and communication to improve overall system performance.
  • PHY-layer procedures: Spatially sparse, low-rank THz channels create PHY-layer challenges for channel estimation and initial access.The paper highlights generative learning for predicting full THz CSI and sensing for enhanced initial network access.
  • Spectrum access techniques: Conventional access schemes cannot be directly applied to THz bands because of hardware constraints and the distinctive propagation environment.The paper examines THz-suitable spectrum access techniques, including orbital angular momentum and non-orthogonal approaches.
  • THz-enabled use cases: The paper concludes by surveying prospective THz-enabled use cases, their challenges, open problems, and ways to exploit the seven defining features.Four major 6G use cases are presented in relation to the defining features.

II. FUNDAMENTALS OF THZ FREQUENCY BANDS

THz communications offer abundant bandwidth but face short, intermittent links caused by severe losses, narrow LoS availability, blockage, and molecular absorption. These effects create a trade-off: higher frequencies provide more bandwidth while increasing absorption-related losses and worst-case performance risks.

  • THz channel characteristics: Abundant bandwidth is THz communication’s key advantage, but high path and reflection losses, sporadic LoS availability, and molecular absorption limit range and reliability.These characteristics produce intermittent on/off link behavior, with molecular absorption additionally degrading received power and intensifying noise.
  • THz channel characteristics: 5–10 dB and more than 15 dB: first- and second-order reflected paths are attenuated relative to LoS at 300 GHz–1 THz.The resulting LoS dominance requires concentrating power into very narrow beams.
  • Blockage and propagation: Dynamic and self-blockage models must generalize across environments because human behavior varies between settings such as roads and indoor VR spaces.Static blockages are more deterministic, whereas human-dependent blockage is harder to predict.
  • Blockage and propagation: 10×: mmWave communication range exceeds THz range, while THz beams are narrower and less penetrative, limiting direct reuse of mmWave blockage mitigation.These differences motivate more accurate THz-specific mitigation methods.
  • Deployment trade-offs: Dense THz deployment alleviates short range but can increase LoS interference, handovers, intermittent cell association, and cell-edge performance degradation.Small cells improve best-case proximity to serving base stations while potentially worsening worst-case outcomes.
  • Molecular absorption: Molecular absorption varies non-monotonically with frequency and atmospheric composition, making low-absorption windows potentially unreliable outdoors but more suitable indoors or in controlled environments.Water vapor and meteorological conditions affect absorption levels, while higher frequencies offer more bandwidth but incur a higher absorption baseline.
  • Molecular absorption: Higher frequencies can improve average content freshness and communication reliability through increased bandwidth but worsen worst-case content freshness because of molecular absorption.The result illustrates the opposing effects of THz bandwidth and absorption.
  • Communication–sensing interplay: Molecular absorption is communication-detrimental yet creates sensing opportunities, making THz channel impairments relevant to both communication design and sensing functionality.The paper identifies quasi-opticality as the first defining feature of THz systems.

III. QUASI-OPTICALITY OF THE THZ BAND

THz quasi-opticality couples communication and high-resolution sensing: molecular interactions support environmental identification, while large arrays and bandwidth enable radar, localization, mapping, and orientation sensing. The same channel properties also require architectures tailored to sparse LoS propagation, dense deployment, and multifunctional operation.

  • Environmental sensing: Molecular absorption enables electronic smelling through rotational spectroscopy, which identifies gases by measuring energy transitions between quantized molecular rotational states.The technique is described as specific, sensitive, and capable of determining concentration.
  • Environmental sensing: Neural networks can process THz sensing measurements to learn complex patterns for substance or fluid-dynamics identification, although this remains an open research area.The processing architecture depends on the desired identification or classification task.
  • High-resolution sensing: Large arrays combined with abundant THz bandwidth enable high-resolution radar, centimeter-scale localization, precise 3D mapping, and fine-grained orientation sensing.These capabilities support sensing across multiple contexts, including indoor personal-radar applications.
  • Communication–sensing fellowship: THz systems form a communication–sensing fellowship in which communication blockages can represent sensing opportunities, using measurements such as angles, delay, Doppler, and object patterns.The resulting sensing outcomes include object detection, UE tracking, and 3D mapping.
  • THz-tailored architectures: THz architectures must sustain highly varying channels, short-range links, narrowbeam LoS dependence, high traffic capacity, connection density, and multifunctional services.These requirements motivate architectures designed specifically for THz propagation and system functions.
  • THz-tailored architectures: Ultra-massive MIMO can fit within a few square millimeters and provide scalable pencil beamforming, while array-of-subarrays designs can improve beamforming gain and energy efficiency.Unlike sub-6 GHz settings with favorable propagation, THz channels are sparse, LoS-dominated, and low rank.
  • THz-tailored architectures: Massive-MIMO small-cell densification supports seamless coverage but introduces intermittent cell-edge connectivity, inter-cell interference, and substantial global-CSI overhead.These challenges arise alongside THz’s short range and beam-alignment sensitivity.

2) Opportunities:

THz-tailored architectures can address short-range, intermittent links through cell-free cooperation, clustering, and RIS-assisted propagation control. These approaches improve link continuity and sensing potential, but scalability and CSI acquisition remain challenges.

  • Cell-free massive MIMO unties THz architectures from cellular boundaries while potentially providing more uniform user experience and improved edge-user QoE.
  • Local CSI requirements in cell-free massive MIMO can reduce overhead and improve THz reliability and latency, although other channel-estimation challenges remain.
  • Dynamic, possibly overlapping user-centric BS clusters can suppress interference and activate multiple cooperative links, improving THz reliability and continuity.
  • Cell-free THz architectures still face unresolved scalability issues, including power control, synchronization, and pilot assignment.
  • RISs can re-engineer uncertain THz propagation environments to support continuous LoS links and improve beam and mobility management.
  • Holographic RISs provide higher spatial resolution, richer propagation paths, and potentially precise 3D environmental mapping for THz sensing.

3) Opportunities in Near-field RIS communication at THz:

Near-field and holographic RIS techniques exploit THz propagation characteristics to focus beams, enrich channels, and support sensing. Their deployment introduces substantial cooperation, CSI, latency, and integration challenges.

  • Near-field THz communication can focus reflected beams on new focal points and exploit wavefront curvature rather than relying only on far-field anomalous reflection.
  • Multiple cooperative RISs increase channel-estimation overhead and processing delays, creating stringent real-time cooperation and scheduling requirements.
  • Accurate CSI acquisition for multiple THz RISs is difficult because many metasurfaces have nonlinear hardware characteristics and rapidly changing channels limit CSI validity.
  • Sparse-channel and geometry-based data-driven methods are proposed to infer accurate CSI within THz coherence times and 6G latency requirements.
  • Integrating THz with lower-frequency bands can support realistic coverage, scalable networks, and reliable real-time reconfiguration of blockage-prone THz links.

B. Opportunities

Coexisting THz, mmWave, and sub-6 GHz bands can combine high-rate short-range services with broader and more reliable connectivity. Effective hybridization requires new interfaces, control methods, and resource-management strategies.

  • Lower-frequency links can improve THz reliability and continuity through blockage prediction and control-information exchange.
  • THz and mmWave bands can provide high-resolution short-range sensing alongside longer-range radar-like sensing.
  • Spatial correlation across three frequency bands can support traffic scheduling and reduce training overhead.
  • Multi-band support enables network slices with different rate, latency, and synchronization requirements through band selection and handoff.
  • THz information showers can prefetch high-rate content while existing lower-frequency base stations continue serving conventional lower-rate services.
  • Hybrid networks need robust lower-frequency interfaces for beamforming, initial access, and channel estimation, while THz interfaces handle bandwidth-intensive transmission.
  • Open problems include cross-band control and payload mapping, joint scheduling, spectrum management, and user association.
  • Successful THz deployment depends on THz-tailored architectures and coexistence with lower-frequency bands, while exploiting their distinct complementarities.

VI. JOINT SENSING AND COMMUNICATION SYSTEMS

Joint sensing and communication systems use THz measurements and feedback to support beam management, predictive resource allocation, spectrum sharing, and new applications. Reliability, continuity, coverage, and sensing range remain central constraints.

  • Joint sensing and communication creates mutual feedback in which communication blockages can represent sensing opportunities and sensing can support communication functions.
  • Continuous sensing feedback can improve initial access, beam tracking, and user association for mobile THz pencil beams.
  • Sensing information can enable predictive communication-resource allocation by characterizing environmental changes and user gestures.
  • Dynamic spectrum sharing between THz sensing and communication can increase spectrum efficiency, while OAM can multiplex both functions.
  • Joint THz sensing and communication supports applications such as XR that require high-rate service and instantaneous high-resolution localization.
  • Integrating sensing and communication on one platform can reduce system cost and size, but short-range intermittent links constrain communication and sensing coverage.
  • Multiple independent paths can extend communication reliability and enrich sensing measurements for longer-range objects and more detailed situational awareness.
  • Integrating THz with mmWave can extend joint sensing and communication to outdoor applications, including radar and environmental sensing for autonomous vehicles.

C. Challenges

Joint sensing and communication at THz frequencies creates resource, waveform, coexistence, and design-complexity challenges that require THz-tailored solutions.

  • Resource sharing and allocation: Dynamic resource allocation must balance stable, sharply pointed communication beams against time-varying directional sensing beams.The central tradeoff is between high data rates and high-resolution sensing.
  • Coexistence schemes: Sensing and communication can coexist through spectral overlap, cognition, or functional integration, each with application-dependent advantages and drawbacks.These schemes respectively target interference mitigation, channel sensing, or hardware-level integration.
  • Waveform design: Radar and communication waveforms differ in modulation, power radiation, and receiver complexity, requiring joint systems to characterize their tradeoffs.Radar commonly uses unmodulated signals, pulses, or chirps, whereas communications combine pilots and modulated signals.
  • Design complexity: Extending joint sensing and communication to long-range, ubiquitous coverage requires THz integration with mmWave, complicating feedback and scheduling.The two bands differ in interference, accuracy, and range, creating additional coordination requirements.
  • PHY-layer procedures: THz PHY procedures face high pilot overhead, beam-deaf initial access, and coherence times too short for training and payload exchange.Sparse, highly dimensional channels and narrow beams make channel estimation and initial access especially difficult.
  • Channel estimation: Dense RIS-enabled THz architectures produce time-varying compound channels, making instantaneous CSI essential for reliable beam alignment, association, and optimization.Partial CSI can yield poor decisions when user-centric clusters and RIS operating modes change rapidly.
  • Channel estimation: User-centric clusters require supplemented channel data and cooperative learning because local data are scarce and cluster conditions are correlated.Existing approaches remain limited by single-agent learning and inadequate compound-channel modeling for RIS-enhanced networks.

B. Initial Access

THz initial access and spectrum access require alternatives to conventional lower-frequency procedures, using sensing, lower-band assistance, and THz-tailored multiplexing.

  • Initial access: Initial access must make narrow SBS or RIS beams and UE beams meet in space before information exchange, creating a THz deafness problem.Two proposed directions are lower-frequency assistance and sensing-based situational awareness.
  • Initial access: Lower-frequency integration can configure links, associate beams, and gather control information before THz data exchange.This approach uses existing lower-band connectivity to support THz access procedures.
  • Initial access: THz sensing and localization can sweep beams to estimate UE location and orientation before communication, while RISs can provide multiple paths for richer environmental information.A single narrow-beam LoS sensing link is time-consuming and vulnerable to blockage.
  • Spectrum access: Conventional OFDMA and CDMA cannot be directly transferred to THz systems because quasi-optical propagation and hardware constraints alter spectrum-access requirements.THz access schemes must also address stringent emerging 6G requirements.
  • OAM: THz quasi-opticality supports robust OAM modes that add dimensions for LoS multiplexing, multiple access, and spectral-efficiency improvement.Orthogonal topological-charge modes can reduce reliance on conventional spatial or frequency resources.
  • OAM: Metasurface-based OAM generation can exploit RIS architectures for spectrum efficiency, resource allocation, multiplexing, and multiple access.RIS phase regulation provides a route to integrate OAM into THz network designs.
  • NOMA: NOMA can improve fairness and spectral efficiency in THz networks, but narrow beams complicate user pairing and power and bandwidth allocation.Its benefits are especially relevant to users facing blockage, deep fades, mobility, and large channel disparities.
  • NOMA: RIS-NOMA architectures can combine passive beamforming, energy-efficiency gains, massive connectivity, and spectral-efficiency improvements.The proposed synergy can apply whether RISs are active or passive.

IX. REAL-TIME THZ NETWORK OPTIMIZATION

Real-time THz optimization must handle uncertain, rapidly changing, heterogeneous, and compound network conditions while jointly supporting communication and sensing objectives.

  • Optimization requirements: THz optimization must account for short coherence times, blockage, molecular absorption, beam misalignment, heterogeneous bands, and conflicting rate-resolution objectives.These conditions undermine robust real-time communication, control, and computing for 6G services.
  • Compound channels: RIS architectures mitigate channel challenges but create compound channels requiring continuous synchronization across multiple links.The resulting tradeoffs lack explicit models and motivate data-driven performance assessment.
  • Non-stationary data: Non-stationary, jointly correlated channel and network-performance distributions make prediction and generalization inherently complex.Breaking correlations among events can simplify prediction.
  • Distributed optimization: Centralized methods conflict with low-latency requirements and cannot generalize location-specific decisions across changing THz network conditions.The paper therefore motivates distributed learning approaches for network control.
  • Learning challenges: Location-specific data are insufficient for reliable learning, while current ML training periods are too long for real-time operation under non-stationary conditions.The paper proposes supplementing channel data and developing real-time multi-agent learning.

A. Towards Generalizeable and Specialized Learning

The paper develops a learning perspective that combines generalized knowledge with specialized agent capabilities, then extends it toward multi-task and meta-learning for complex THz environments.

  • Generalized and specialized learning: Multi-agent learning can separate generalizable traits from service-, mobility-, UE-, and resource-specific characteristics.Dynamic clustering assigns agents specialties while preserving shared knowledge.
  • Generalized and specialized learning: Shared common skillsets improve prediction across agents, while isolated skillsets depend on their complexity and available data.The framework combines common denominators with shared specialties.
  • Generalized and specialized learning: Democratized learning offers hierarchical sharing from common knowledge to specialized skills, but prior evaluations used only MNIST and Fashion MNIST.Those studies did not test time-sensitive, scarce, heterogeneous THz data or real-time reinforcement learning.
  • Learning robustness: Deep Q-learning may fail on slightly out-of-distribution THz data because blockages and deep fades create heavy-tailed conditions.This exposes a robustness challenge even after clustering related users or services.
  • Multi-task learning: Multi-task learning partitions beam alignment into distinct environmental tasks, such as high-blockage and nominal-operation conditions.The number and type of tasks can differ across learning problems and may not be known beforehand.
  • Multi-task learning: Riskier beam-alignment tasks can consume substantial frequency, energy, and spatial resources for continuous sensing, whereas nominal tasks can allocate more resources to communication.This creates an operational resource distinction between extreme and ordinary conditions.
  • Meta-learning: Multi-task and meta-learning are proposed to address data scarcity and improve generalization, but task definition is difficult and wireless problems are stochastic reinforcement-learning settings.Existing meta-learning work has focused mainly on supervised learning with labeled data.

X. USE CASES FOR THZ WIRELESS SYSTEMS

The paper surveys 6G use cases for THz systems, emphasizing demanding XR, holographic, industrial, and digital-twin requirements alongside architectural and reliability challenges.

  • Use-case overview: The use cases can be deployed on different THz architectures according to their distinct requirements and operating modes.The paper frames these use cases as applications of the seven defining THz-system features.
  • XR and holographic services: XR and holographic services require extremely high rates together with continuous low latency, low jitter, fresh information, synchronization, and precise delivery.Momentary disruption of a THz line-of-sight link can disrupt the user experience, motivating indoor architectures that increase LoS likelihood.
  • XR and holographic services: Outdoor XR and holographic deployments face molecular absorption, longer ranges, mobility, beam-tracking overhead, and tradeoffs between operational versatility and user QoE.AR for assisted driving additionally makes uplink information freshness safety-critical.
  • Industry 4.0 and digital twins: Industry 4.0 automation demands Tbps data rates, hundreds-of-microseconds latency, and connection density of 10^7/km^2 for instantaneous control.Digital twins further require real-time synchronization between physical and cyberspace models.
  • Industry 4.0 and digital twins: Digital-twin systems are error-sensitive because faulty cyberspace initialization can propagate biases through subsequent predictions and decisions.Dual-band mmWave/THz cooperation is proposed for HRLLC model updates, while multiplexing and multiple access can improve scalability and spectral efficiency.

C. CRAS

CRAS requires high-rate exchange and accurate environmental sensing, but mobility makes continuous THz line-of-sight connectivity unreliable. The paper therefore combines THz with lower-frequency links and distributed learning.

  • CRAS requirements: CRAS systems exchange large data volumes, including high-resolution real-time maps, while sensing and tracking their environment for routing, traffic, and safety.Autonomous driving, drone swarms, and vehicle platoons are representative CRAS services.
  • CRAS connectivity: Although THz can provide CRAS rate requirements, high mobility disrupts reliability through intermittent links and unavailable continuous line-of-sight paths.THz indoor stations can provide high-rate map delivery while mmWave and sub-6 GHz links carry less data-intensive content.
  • Integrated sensing and communication: Sensing data can augment communication measurements to support predictive control driven by high-reliability, low-latency communications.Integrated THz and mmWave sensing and communication also require new network models, broader coverage, and predictive resource management.
  • Learning for CRAS: Centralized black-box ML can learn spurious relationships and lacks strategic decision learning under out-of-distribution events, with scarce datasets worsening risks in vehicular settings.The paper calls for trustworthy, real-time ML mechanisms in high-risk autonomous systems.
  • Learning for CRAS: Multi-agent reinforcement learning enables local decision making, while THz and mmWave can share local models or data to improve agent generalizability.Distributed schemes can reduce bidirectional overhead associated with centralized ML.

D. NTNs

THz-enabled non-terrestrial networks combine high-capacity links with lower-frequency coverage to support continuity, ubiquitous access, and scalable backhaul. The paper places this integration within a broader seven-feature deployment roadmap.

  • NTN objectives: 6G non-terrestrial networks target service continuity for mobile platforms, ubiquitous access to underserved areas, greater scalability, and more efficient backhaul.Integrated space-air-ground communications support these objectives through satellites, UAVs, and related platforms.
  • THz–mmWave integration: Synergistic THz and mmWave air-to-air and air-to-ground links can provide continuous links, ubiquitous communication, and high-capacity backhauls.Longer communication ranges in these settings necessitate mmWave links alongside THz.
  • Spectrum coexistence: Integrated satellite-terrestrial THz/mmWave backhaul must coexist with current systems, where terrestrial transmitters can interfere with satellite backhaul terminals.Flexible spectrum-sharing techniques are therefore identified as necessary.
  • Deployment roadmap: The paper’s roadmap covers THz fundamentals, quasi-optical opportunities, architectures for directional LoS links, lower-band synergy, and joint sensing and communication.It also addresses channel estimation, initial access, and ML approaches for network design and optimization.
  • Recommendations: The recommendations favor indoor or overlaid initial deployment, versatile joint sensing-communication systems, metasurfaces, integrated frequency bands, and specialized ML.These measures address short range, intermittent links, coverage, sensing, channel estimation, and real-time optimization.
Loading 2102.07668v2…