In December 2024, engineers at South Korea’s Electronics and Telecommunications Research Institute (ETRI) achieved something wireless researchers had chased for years. Inside a controlled laboratory in Seoul, South Korea, they transmitted data across open air at terahertz frequencies, sustaining a link that carried more information than any wireless system before it.
It was one of the first demonstrations to show that communication at these frequencies could exist outside simulation and theory, under conditions that engineers could observe and measure.
This achievement drew attention for its speed. What mattered more was where it operated. Terahertz frequencies sit in a region of the spectrum long considered inhospitable to practical communication. Signals here behave less like the radio waves that support today’s networks and more like light. And at such frequencies, the rules change because energy interacts with matter in unfamiliar ways. Atmospheric humidity absorbs specific bands. Molecular resonances carve the spectrum into narrow windows, interrupting continuity. Surface texture alters reflection and scattering at scales that previously averaged out.
Sustaining a stable link under those conditions requires abandoning many assumptions that have guided wireless design for generations. In recent years, research has shifted toward understanding how propagation, antennas and devices behave specifically in this range, rather than extending models developed for lower frequencies.
The Seoul experiment served as a quiet turning point. Terahertz communication crossed from speculation into measurement. That transition matters because it defines where sixth-generation wireless will begin. The next era of connectivity will take shape in frequency bands where physics dictates design and where progress advances through evidence rather than projection.
Why 6G is happening now
Though commercial deployment of 6G is expected closer to 2030, the research underway today is born out of necessity. Wireless systems mature slowly and the limits shaping the next generation already press against current networks.
Those limits became difficult to ignore as fifth generation systems expanded into millimeter-wave spectrum. In controlled environments, mmWave links delivered impressive throughput. In operational settings, performance proved fragile. Field measurements showed rapid degradation outside narrow line-of-sight paths, with throughput collapsing under modest blockage and user movement. Ensuring reliability required both dense infrastructure and continuous beam management.
As networks compensated with more base stations and more coordination, returns diminished. The mechanisms that once enabled scaling began to define its ceiling. Engineers recognized that further extensions of millimeter wave would require increasing complexity without delivering proportional benefit. Attention shifted upward toward spectrum where bandwidth itself could justify that complexity. Frequencies above 100 GHz provided that space.
Operating above 100 GHz
Operating in this domain alters the problem in fundamental ways. Wavelengths shrink to fractions of a millimeter, approaching the scale of surface roughness found on building materials, circuit boards and even antenna substrates. Reflection coefficients become highly sensitive to finish and composition. Scattering increases where lower-frequency models assume smooth behavior.
Atmospheric absorption further fragments the spectrum. Molecular resonances, particularly those from water vapor, introduce frequency-dependent attenuation that divides the band into usable corridors rather than a continuous block. Measurements reveal sharp notches and distance-dependent loss profiles that cannot be treated as secondary effects and must be accounted for during link planning.
Channel modeling
At lower frequencies, channel models often serve as abstractions. They describe average behavior and validate design after the fact. Above 100 GHz, that role changes. Channel modeling becomes a gating function for feasibility.
Once wavelength approaches surface texture and absorption dominates link behavior, assumptions inherited from sub-6 GHz and even millimeter-wave models fail to predict performance with useful accuracy. Irregularities don’t average out, they persist.
Researchers have responded by developing hybrid modeling approaches that combine deterministic ray tracing with parameters defined by measurement. Ray tracing captures geometry-driven paths and reflection behavior, while empirical data accounts for variability introduced by real materials, alignment tolerance and environmental conditions such as humidity.
What separates current terahertz models from earlier attempts is how tightly they couple theory to observation. These models preserve the irregularities. Coherence bandwidth, delay spread and spatial consistency emerge directly from measured behavior, particularly across short indoor and urban links where surface interaction dominates propagation and small changes in material or geometry reshape the channel itself.
When engineers treat absorption windows, material composition and antenna directivity as primary inputs, predicted channel behavior begins to align closely with what instruments record in practice. At that point, channel modeling stops approximating feasibility and starts setting its own boundaries. Decisions about frequency selection, antenna placement and beam strategy move upstream, before hardware is committed.
Antennas as architecture
At terahertz frequencies, the antenna stops behaving like a component that can be optimized in isolation. Its influence spreads outwards, shaping how the entire system behaves. The reason lies partly in scale. When wavelengths shrink to fractions of a millimeter, the physical size of the array, the spacing between elements and the materials used to support them all begin to matter in ways that are difficult to decouple.
Dense phased arrays become possible at these frequencies, packing hundreds of elements into apertures that would have supported only a handful at lower bands. That density enables extremely narrow beams and aggressive special reuse, fundamental to keeping link quality as path loss rises. It also compresses tolerance margins. Small differences in alignment or phase coherence immediately surface.
Beamforming reflects this shift and becomes the mechanism that holds the link together. Phase noise introduced by oscillators affects beam coherence rather than appearing as a secondary impairment. Thermal drift alters carrier stability and, with it, the coherence of the array. Moreover, calibration, once an occasional adjustment, becomes a continuous concern.
Recent hardware demonstrations illustrate how tightly these factors are intertwined. Fully integrated phased-array transceivers operating in the 220 GHz to 260 GHz range have shown that electronically steerable beams are achievable with usable performance, but only when antenna geometry, RF circuitry, packaging and thermal design are addressed together rather than sequentially.
Materials at the limit of radio
While engineers are pushing wireless systems beyond 100 GHz, materials have moved to the center of the problem. Copper, long the foundation of radio-frequency design, becomes increasingly lossy as skin effects confine current to ever thinner layers. Silicon, prized for its scalability, struggles to deliver efficiency when frequencies rise. Even packaging, often treated as a supporting detail at lower bands, begins to assert itself. Parasitics introduced by substrates and interconnects approach the scale of the signal paths they carry.
In fact, these limitations explain why terahertz research has drifted away from the conventional RF material palette. Compound semiconductors such as indium phosphide and gallium-based technologies offer higher electron mobility and improved gain at extreme frequencies. Their performance advantages are well documented, as are the manufacturing and integration obstacles that accompany them.
Attention has also turned toward materials that behave differently at these scales. Two-dimensional systems, particularly graphene, offer electrical properties that can be tuned in new ways. High carrier mobility persists even as dimensions shrink, and conductivity can be adjusted dynamically, characteristics that align well with modulation and detection in the terahertz range.
At these frequencies, material behavior feeds directly into system performance, which makes direct measurement the only reliable way to understand how theory survives contact with real hardware.
Measurement as the new grammar of wireless design
As terahertz research has progressed, measurement has taken on a different role. It no longer sits at the end of the process, verifying designs that are already complete. It has moved upstream, shaping decisions before architectures harden and long before systems leave the laboratory.
This shift reflects the nature of the problem. At frequencies where wavelength approaches surface texture and material behavior varies across narrow bands, small assumptions propagate quickly into system failure. Simulation still guides exploration, but it cannot stand alone. What matters is how theory behaves when exposed to real materials, real alignment tolerances and real environmental variation.
That reality has driven the development of experimental testbeds designed to highlight interactivity. These platforms bring wideband transceivers, highly directional antennas, adaptive beam control and calibrated measurement chains together so that propagation, phase noise, alignment sensitivity and link stability can be observed as a coupled system. Recent surveys show how this approach allows researchers to examine behavior across multiple terahertz frequency windows without changing experimental context, placing modeled absorption regions directly alongside measured performance.
What emerges from this work is a change in practice. Channel models stop predicting in the abstract and begin bounding what is feasible. Antennas stop compensating for uncertainty and begin enforcing it. Materials stop offering incremental improvement and start defining limits.
Sixth-generation wireless will not arrive fully formed. It is being assembled now, piece by piece, through measurement that clarifies where radio still works and where it does not. In that sense, terahertz research is less about reaching higher frequencies than about learning how to design again, in a domain where assumptions no longer hold and evidence leads the way.
