Terahertz Spectrum for 6G: Opportunities, Propagation, and Device Challenges

Why the THz Frontier Is on the 6G Roadmap

By the late 2020s, the densification of mid-band and lower millimeter-wave networks will struggle to deliver the data rates, latency, and spatial fidelity that immersive video, machine-to-machine swarms, and joint sensing-communication systems demand. The 100 GHz–1 THz range offers contiguous bandwidths of tens of gigahertz and wavelengths measured in fractions of a millimeter, opening a path to raw capacities in the tens or hundreds of gigabits per second. The same physical properties make the band attractive for high-resolution radar-like sensing and positioning, giving 6G a potential sensing layer that shares spectrum with payload data.

Propagation Physics: More Than Free-Space Loss

Free-space path loss scales roughly with the square of frequency, so a 300 GHz link suffers about 20 dB more loss than a 30 GHz link at the same distance. The real penalty is larger because of molecular absorption: water vapor and oxygen have resonant absorption lines clustered between 100 GHz and 1 THz, adding frequency- and humidity-dependent attenuation that can reach several to tens of decibels per kilometer. Indoor and outdoor channel measurements show that walls, foliage, and human bodies block THz signals, while non-line-of-sight paths are weak, diffuse, or absent. Rain and atmospheric scintillation compound the problem. These characteristics do not eliminate the use case, but they restrict it to short-range, high-density deployments where line-of-sight is the norm and repeaters or intelligent surfaces extend coverage.

Absorption Peaks and Frequency-Selective Fading

The spectrum is not uniformly usable. At 183 GHz, 325 GHz, 380 GHz, 448 GHz, and 557 GHz, water-vapor absorption peaks create deep nulls. Engineers model these as frequency-selective attenuation superimposed on distance-dependent path loss, which means a system may need to hop between lower-absorption windows or adapt bandwidth and modulation based on the link budget.

Antenna Arrays and the Beam-Narrowing Paradox

Short wavelengths allow thousands of antenna elements to fit within a credit-card-sized aperture, enabling massive spatial multiplexing and extremely narrow beams. A 256-element array at 300 GHz can produce a beamwidth of only a few degrees, which boosts antenna gain and suppresses interference, but also makes alignment and tracking critical. Sub-millimeter movement by a user or device can place the signal outside the main lobe, forcing beam-management cycles at sub-millisecond timescales. Hybrid analog-digital beamforming architectures are the likely compromise: analog beamforming provides array gain, while digital processing handles a small number of spatial streams to keep power and complexity manageable.

Beam Management at Scale

In mobile or cluttered environments, base stations and devices must continuously predict the best beam pair. Machine-learning algorithms can compress the search space using pose, history, and channel-state information, but the trade-off between training data and real-time latency remains an active research question.

Semiconductor and Packaging Requirements for THz Radios

Silicon CMOS can operate above 100 GHz in advanced FinFET or gate-all-around nodes, but transistor cutoff frequencies and noise figures degrade as frequencies approach 1 THz. III-V materials such as indium phosphide and gallium arsenide offer higher electron mobility and remain favored for high-output-power sources and low-noise amplifiers, although they are harder to integrate with digital baseband. Silicon-germanium BiCMOS and RF silicon-on-insulator provide intermediate options for switches and front-ends. The bottleneck, however, extends across the entire signal chain.

From Transistors to System-in-Package

• Active devices: CMOS offers density and digital integration but lower raw power and efficiency; III-V compounds deliver output power and low noise but cost more and complicate heterogeneous integration.
• Local oscillators and converters: Low-phase-noise sources, broadband digital-to-analog converters, and high-sampling-rate analog-to-digital converters must be co-designed with the antenna array.
• Packaging and thermal: Parasitic capacitance, interconnect loss, and thermal density become first-order design variables, pushing systems toward chiplet-based or wafer-scale integration and advanced packages with embedded antennas and low-loss dielectrics.

Roadmap and Open Problems

Before THz bands can be standardized for 6G, channel models must capture both deterministic ray-tracing and statistical behavior across indoor, outdoor, and body-centric scenarios. Regulators must allocate spectrum and power limits, and coexistence with existing sensing and scientific services must be resolved. Energy efficiency is another concern: massive arrays, high-rate converters, and complex beam management can consume several watts even at short range, making the technology viable for fixed and pedestrian applications before wide-area mobile use. The next few years will determine whether the THz frontier becomes a core 6G pillar or remains a niche capability for specialized hotspots and sensing.