As the demand for more spectrum increased with the extensive usage of mobile data, XR/VR, sensing, and autonomous systems, the sub-THz region (100–300 GHz and beyond) emerges as a compelling frontier. In effect, we are approaching the limits of what mm Wave alone can deliver at scale. The THz band promises immense contiguous spectrum, enabling links well above 100 Gbps, and the possibility of co-designing communication and high-resolution sensing (imaging/radar) in a unified platform.
Yet this promise confronts severe physical obstacles: high path loss, molecular absorption, component limitations, packaging losses, and system complexity. This article traces how the industry is navigating those obstacles, what is working now, what remains open, and where the first real systems might land.
The Early Milestones: Lab Prototypes That Matter
A landmark announcement came in October 2024 from NTT: a compact InP-HEMT front-end (FE) that achieved 160 Gbps in the 300 GHz band by integrating mixers, PAs, LNAs, and LO PAs in a single IC.
Key technical innovations in that work include:
- A fully differential configuration to cancel local-oscillator (LO) leakage, critical at THz frequencies.
- Reduction of module interconnections (thus insertion loss) by integrating discrete functions into a monolithic chip.
- Shrinking module size from ~15 cm to ~2.8 cm, improving form factor while widening operational bandwidth.
More recently, in mid-2025, NTT (with Keysight and its subsidiary NTT Innovative Devices) demonstrated a power amplifier module capable of 280 Gbps (35 GBaud, 256-QAM) in the J-band (≈220–325 GHz), albeit at 0 dBm output power. This points toward simultaneous scaling of both bandwidth and linear output power, a crucial step forward.
On the standardization/architectural front, partnership experiments like Keysight + Ericsson’s “pre-6G” prototype show how new waveforms and stacks might evolve. In 2024, they demonstrated a base station + UE link (modified 5G stack) over new frequency bands, signaling industry interest in evolving existing layers to support extreme throughput. Ericsson itself emphasizes that 6G will mix evolved and new concepts spectrum aggregation, ISAC, spatial awareness, and energy-efficient designs.
These milestones are not “toy results” they validate that the critical component blocks can already support high-throughput, multi-GHz signals, albeit in controlled lab settings.
Technical Foundations: Devices, Architectures, and Packaging
To move from prototypes to systems, several technical foundations must be matured in parallel:
Device and Front-End Technologies
- InP / III–V HEMTs and HBTs remain leading candidates for mixers, LNAs, and PAs at high frequencies, thanks to superior electron mobility and gain.
- SiGe BiCMOS bridges the gap, often handling LO generation, control logic, and lower-frequency blocks, while III–V handles the toughest RF segments.
- Schottky diodes, resonant tunneling diodes (RTDs), and nonlinear mixers play roles for frequency translation and LO generation.
- Photonic sources such as UTC photodiodes or photomixing supplement generation in narrowband, coherent applications. For example, a modified uni-traveling-carrier photodiode (MUTC-PD) has been proposed for 160 Gbps over D-band in a fiber-THz hybrid link.
The challenge is achieving sufficient output power, flat gain over multi-GHz bandwidth, linearity, and noise performance, all within thermal and size constraints.
Architectures and Signal Processing
- Multiplication chains (cascaded frequency multipliers) remain the standard path for elevating microwave frequencies into THz.
- Harmonic or sub-harmonic mixing eases LO generation but while managing phase noise is critical.
- Beamforming / phased arrays are essential. Directive beams offer path-loss mitigation and interference control. True-time delay or phase shifting (with very fine resolution) is a design hurdle at THz.
- Waveforms must tolerate impairments (phase noise, CFO). Hybrid schemes combining single-carrier plus OFDM and FMCW / chirp waveforms are under study.
- Joint sensing-communication (ISAC): Using the same waveform for data and radar-like imaging is central to future designs.
- Channel modeling, beam training, blockage prediction, and adaptive modulation are crucial companion software domains.
Packaging, Antennas, and Interconnects
At THz, packaging and interconnect losses can kill performance faster than device limitations.
- Antenna-in-package (AiP) and antenna-on-substrate (e.g. silicon lens, meta surfaces, dielectric lens) help reduce the distance from active devices to radiating aperture.
- Substrate-integrated waveguides (SIW), micromachined waveguides, quasi-optical coupling replace lossy microstrip lines and CPWs.
- Thermal spreaders, heat conduction, and material selection (low-loss dielectrics) are critical for sustaining device stability.
- Calibration and measurement: On-wafer TRL/LRM up to sub-THz, over-the-air (OTA) test setups, and real-time calibration loops are required for production test.
Propagation, Channel, and Deployment Constraints
Propagation in THz is unforgiving:
- Free-space path loss (FSPL) scales with frequency. Every additional decade in frequency adds ~20 dB loss.
- Molecular absorption, especially from water vapor, introduces frequency-specific attenuation notches; engineers must choose spectral windows (D-band, G-band, J-band, etc.).
- Blockage: Humans, objects, and materials often act as near-total blockers at THz.
- Multipath is limited — channels tend toward sparse tap-delay profiles.
Thus, THz is suited for controlled, short-range, high-throughput links or co-located sensing+ communication. Outdoor macro coverage is generally impractical unless beams are extremely narrow and paths well managed. Backhaul and hotspot links are more feasible use cases than full wide-area coverage.
Imaging and Sensing Use Cases
Unlike pure communication, imaging demands high dynamic range, spatial resolution, and sometimes passive operation. THz enables:
- Active coherent imaging (FMCW, pulsed radar) for 3D reconstruction, industrial NDT, and package inspection.
- Passive imaging / thermography for detecting emissivity contrasts.
- Computational imaging via coded apertures, compressed sensing, and meta surface masks to reduce sensor complexity.
In system designs, the same front-end and beam infrastructure may handle both data and imaging tasks, subject to power and SNR trade-offs.
Roadmap & Open Problems
While lab successes validate feasibility, many gaps remain before field-ready systems:
- Watt-class, efficient THz sources at room temperature (particularly beyond 200 GHz).
- Low-loss, scalable passives and interconnects (waveguide, delay lines) at THz frequencies.
- Robust channel models across environments (indoor, outdoor, humidity, mobility) with validation data.
- Low-cost calibration / test methodologies for mass production.
- Integrated ISAC signal processing and software stacks that abstract complexity from system integrators.
- Security and coexistence in pencil-beam, high-frequency environments.
Conclusion: What’s Realistic, What’s Ambitious
The next decade will see THz systems not replacing, but supplementing existing networks. They will begin in enterprise, industrial, and hotspot contexts (e.g. 100+ Gbps indoor links, wireless backhaul, imaging tools in factories). Over time, integrated sensing + communication systems (robotics, AR, digital twins) will leverage THz’s ability to see and talk in the same hardware.
The core enablers: heterogeneous integration (III-V + CMOS/BiCMOS), advanced packaging and optics, robust beamforming, and tightly coupled signal processing. Lab records such as 160 Gbps in the 300 GHz front-end by NTT, and 280 Gbps in a J-band PA module show that neither bandwidth nor throughput is purely theoretical — the next steps are scaling power, cost, and reliability.
