Terahertz Communications
Introduction to Terahertz Communications
Terahertz (THz) communications exploits the sparsely used electromagnetic spectrum between roughly 0.1 and 10 THz. This range, long called the "terahertz gap," sits between microwave and infrared frequencies and offers contiguous bandwidth far beyond anything available at lower frequencies. Atmospheric absorption and device limitations make the band difficult to use, yet laboratory links already reach hundreds of gigabits per second, and the technology is a leading candidate for short-range, very high capacity applications such as wireless data centers, kiosk downloading, and chip-to-chip interconnects.
Terminology varies across the literature. Radio engineers often reserve "terahertz" for frequencies above 300 GHz and describe the 100–300 GHz range as "sub-terahertz," while the communications community frequently treats the whole span above about 100 GHz as one design regime because the same engineering problems—low source power, high path loss, quasi-optical propagation, and blockage—dominate throughout. Nearly all deployed and standardized systems to date sit in the sub-terahertz portion of that span, and this article follows the broader usage while noting where the distinction matters.
The band remained underused for decades because both generating and detecting these waves fell awkwardly between electronic and photonic techniques. Advances in indium phosphide and silicon-germanium transistors, uni-traveling-carrier photodiodes, quantum cascade lasers, and nanofabrication have since made practical terahertz systems viable. Regulators have responded: the 2019 World Radiocommunication Conference identified 137 GHz of spectrum between 275 and 450 GHz for fixed and land mobile use, and the IEEE published the first terahertz communication standard, IEEE 802.15.3d, in 2017. Beyond communications, terahertz radiation interacts distinctively with many materials, which supports imaging, security screening, medical diagnostics, and quality control—applications that share most of the same hardware.
The Terahertz Spectrum Gap
The terahertz frequency range occupies a distinctive position in the electromagnetic spectrum, corresponding to wavelengths from 3 millimeters down to 30 micrometers. A photon at 1 THz carries about 4.1 millielectronvolts of energy, more than three orders of magnitude below the roughly 10 electronvolts needed to ionize a typical molecule, which is why terahertz radiation is non-ionizing. The region is a transition zone where electronic approaches scaled up from microwave frequencies and photonic approaches scaled down from the infrared converge. The gap exists because transistor gain falls away as frequencies approach the device cutoff, while the intersubband and interband transitions that drive optical sources become difficult to sustain as photon energies drop toward the thermal energy scale, roughly 25 millielectronvolts at room temperature.
The bandwidth available in the terahertz region is the principal attraction. A single IEEE 802.15.3d channel may occupy up to 69.12 GHz, whereas millimeter-wave 5G New Radio carriers in the FR2-1 range are limited to 400 MHz. That raw bandwidth translates directly into capacity: even a modest spectral efficiency of a few bits per second per hertz yields hundreds of gigabits per second. Regulators have begun to reflect this potential in spectrum policy while acknowledging the inherently short range of terahertz links, which limits the interference footprint and permits lighter licensing regimes than lower bands require.
Terahertz Generation Methods
Generating terahertz radiation requires specialized techniques that bridge the gap between electronics and photonics. Electronic approaches extend microwave methods upward using advanced semiconductor devices. Frequency multiplication chains drive a microwave or millimeter-wave synthesizer into cascaded Schottky-diode doublers and triplers, and this remains the workhorse method for radio astronomy and instrumentation. Efficiency falls steeply with each stage, so a chain that delivers tens of milliwatts near 100 GHz may yield only microwatts near 1 THz. Fundamental solid-state oscillators avoid that penalty but face their own ceiling: high-electron-mobility transistors (HEMTs) and heterojunction bipolar transistors (HBTs) oscillate reliably into the low hundreds of gigahertz, while resonant tunneling diodes have demonstrated fundamental oscillation approaching 2 THz, albeit with output falling to the microwatt level at the highest frequencies.
Photonic generation converts optical power into terahertz output. Photomixing beats two lasers with slightly different wavelengths in a photoconductive antenna or a uni-traveling-carrier photodiode, producing radiation at the difference frequency. The approach offers continuous tunability across hundreds of gigahertz and inherits the phase stability and modulation bandwidth of optical components, which makes it the preferred transmitter for many record-setting communication experiments. Quantum cascade lasers (QCLs) take a different route, emitting directly from intersubband transitions in engineered semiconductor heterostructures. Terahertz QCLs deliver milliwatt to hundreds-of-milliwatt output but do not yet operate at room temperature: the reported maximum operating temperature has advanced to roughly 250 to 261 kelvin in pulsed mode, which brings them within reach of compact thermoelectric coolers rather than liquid helium, but still short of uncooled operation. Room-temperature terahertz emission is available instead from mid-infrared QCLs that generate the terahertz wave by intracavity difference-frequency mixing, at correspondingly lower power.
Ultrafast optical techniques provide another pathway, using femtosecond laser pulses to switch photoconductive antennas or to drive optical rectification in nonlinear crystals such as zinc telluride and lithium niobate. These sources emit broadband single-cycle pulses spanning several terahertz and underpin most time-domain spectroscopy and imaging instruments, though their low average power and bulky pump lasers limit direct use in communication links. Vacuum electronic devices, including backward-wave oscillators, extended interaction oscillators, and gyrotrons, reach far higher power than solid-state alternatives—watts to kilowatts depending on frequency and device class—at the cost of size, high operating voltages, and expense, which confines them to scientific instruments and specialized industrial uses.
Terahertz Detection Techniques
Detection poses challenges comparable to generation. Direct detectors rectify the incident field and report only its envelope. Gallium arsenide Schottky barrier diodes work into the low terahertz range at room temperature with response times fast enough for video-rate imaging and for demodulating amplitude-modulated data. Thermal detectors—Golay cells, pyroelectric elements, and bolometers—respond to absorbed power rather than field, which makes them broadband but slow. Superconducting hot-electron bolometers achieve the highest sensitivity of any terahertz detector by sensing the resistance change of a nanoscale superconducting bridge, and they operate as mixers as well as direct detectors, though both roles demand cryogenic cooling near 4 kelvin.
Field-effect transistors built on high-mobility semiconductors, and increasingly on graphene, detect terahertz radiation through plasma-wave rectification in the transistor channel. These detectors work uncooled, need no exotic materials, and can be fabricated in standard CMOS, which makes dense focal-plane arrays practical for imaging. Electro-optic sampling offers a different capability: a terahertz field modulates the birefringence of a nonlinear crystal, and a synchronized femtosecond probe pulse reads the instantaneous field amplitude, recovering the full waveform with sub-picosecond resolution. That phase-resolved measurement is what gives time-domain spectroscopy its ability to extract complex refractive index directly, without resorting to Kramers-Kronig inversion.
Communication receivers almost always use coherent detection, which preserves both amplitude and phase. Heterodyne and homodyne front ends mix the incoming signal with a local oscillator and down-convert to an intermediate frequency where conventional analog-to-digital conversion and digital signal processing apply. Room-temperature Schottky mixers reach receiver noise temperatures of a few thousand kelvin, adequate for short links; superconductor-insulator-superconductor mixers approach the quantum noise limit and serve radio observatories to roughly 1.2 THz, with hot-electron bolometer mixers taking over above that. Coherent reception is the precondition for higher-order modulation, polarization multiplexing, and multiple-input multiple-output (MIMO) techniques, all of which are needed to convert raw terahertz bandwidth into usable capacity.
Terahertz Integrated Circuits
Integrated circuit technology is the critical enabler, combining active devices, antennas, and signal processing on single chips. Silicon-based processes, including CMOS and silicon-germanium (SiGe) BiCMOS, now reach into the low terahertz range. Research SiGe heterojunction bipolar transistors developed in the European DOTSEVEN project achieved a transit frequency of 505 GHz and a maximum oscillation frequency (fmax) of 720 GHz, and production BiCMOS processes offer fmax in the 300 to 500 GHz range. Silicon's advantages are decisive where volume matters: mature manufacturing, low cost, and monolithic integration with the digital signal processing that wideband terahertz links require. Its limitation is output power, since low breakdown voltages restrict per-device power to milliwatts or less near 300 GHz, which pushes designers toward on-chip power combining across many parallel amplifiers.
III-V semiconductors deliver higher speed and higher power density. Indium phosphide (InP) high-electron-mobility transistors with 25-nanometer gates have reached projected fmax values around 1.5 THz, and amplifier circuits built in that process demonstrated measurable gain at 1.0 THz—the first amplification at that frequency. Indium gallium arsenide HEMTs combine low noise figure with useful gain across the band, which makes them the standard choice for sensitive receiver front ends, while InP heterojunction bipolar transistors provide the higher breakdown voltage and output power favored for transmitters. Gallium arsenide remains dominant for Schottky diode multipliers and mixers. The trade-off against silicon is cost, wafer size, and the limited digital integration available in III-V processes, which is why many practical systems combine a III-V front end with a silicon baseband in a heterogeneous module.
Antenna-in-package and system-in-package approaches integrate antennas, RF frontends, and baseband processing in multi-chip modules optimized for terahertz operation. These integrated solutions address the challenge of connecting discrete components at frequencies where even short interconnects introduce significant losses and parasitics. Wafer-scale integration techniques enable complete terahertz transceivers with integrated antennas fabricated in single manufacturing processes, dramatically reducing size and cost while improving performance through elimination of package parasitics and bond wire inductances.
Terahertz Antennas and Arrays
Antenna design for terahertz frequencies combines principles from microwave engineering and optics, resulting in structures that bridge conventional radio antennas and photonic devices. The small wavelengths—ranging from millimeters to tens of micrometers—enable high-gain antennas in extremely compact form factors. Planar antennas fabricated using photolithographic techniques include dipoles, patches, slots, and bow-ties integrated directly with semiconductor devices. These antennas often incorporate matching networks and lens coupling to improve radiation efficiency and beam control.
Horn antennas and reflector systems provide high directivity for point-to-point links, with designs optimized for the quasi-optical nature of terahertz propagation. Silicon and polymer lenses focus terahertz beams and can be integrated with on-chip antenna structures, creating compact transceiver modules. Dielectric resonator antennas offer an alternative approach with good radiation efficiency and relatively simple fabrication. Advanced designs incorporate metamaterials and frequency-selective surfaces to control radiation patterns and bandwidth.
On-chip antennas face a specific difficulty. A planar antenna fabricated on a silicon substrate radiates most of its power into the substrate rather than into free space, because silicon's high dielectric constant of about 11.7 favors that direction, and much of that power is then lost to substrate modes. The standard remedies are a high-resistivity silicon lens attached to the substrate side, which collects the downward radiation and collimates it, or a ground plane and localized backside etching that suppress substrate modes. Both add packaging complexity, and lens alignment tolerances tighten as frequency rises.
Phased arrays and leaky-wave antennas steer beams electronically, which mobile and reconfigurable links require. Half-wavelength element spacing is only 0.5 millimeters at 300 GHz, so hundreds of elements fit within a few square centimeters and chip-scale arrays become feasible. The obstacle is the feed network: microstrip and coplanar lines lose several decibels per centimeter at these frequencies, so distributing a phase-coherent signal across a large array can dissipate more power than the array radiates. Practical designs respond by generating or multiplying the carrier locally at each element, by using low-loss substrate-integrated or dielectric waveguides for distribution, or by adopting hybrid schemes that steer coarsely with a lens or reflector and finely with a small array. Leaky-wave antennas offer a lower-cost alternative in which the radiation angle varies with frequency, so tuning the carrier steers the beam without any phase shifters at all—an elegant trade that couples steering to frequency and therefore suits scanning and sensing better than fixed-channel communication.
Propagation Characteristics at Terahertz Frequencies
Terahertz propagation differs from microwave propagation in ways that dictate system architecture. Molecular absorption is the defining feature: water vapor and, to a lesser degree, oxygen present rotational resonance lines that carve the spectrum into alternating absorption peaks and comparatively transparent windows. Commonly cited ground-level windows lie near 0.3, 0.35, 0.41, 0.67, and 0.85 THz. Absorption rises sharply above 1 THz, which confines outdoor terrestrial links to the lower part of the band and leaves the higher frequencies to short indoor paths, satellite crosslinks, and instrumentation.
Free-space path loss between isotropic antennas grows with the square of frequency, so moving from 2.4 GHz to 300 GHz costs about 42 decibels at the same distance. That comparison overstates the penalty, however, because it holds antenna gain fixed. Hold the physical aperture fixed instead and gain also scales with the square of frequency, so a link with the same dish size at both ends actually improves as frequency rises. A 5-centimeter aperture that yields roughly 19 dBi at 24 GHz delivers about 41 dBi at 300 GHz. The price is directivity itself: that same aperture produces a beam under 1.5 degrees wide, so terahertz links demand precise alignment, and mobile deployments require electronic beam tracking that can reacquire within milliseconds.
Because absorption is wavelength-selective, the usable bandwidth of a terahertz channel shrinks as the link lengthens. A path of a few centimeters sees nearly the whole band as flat; at hundreds of meters, only the window centers remain viable. This distance-dependent bandwidth has no real analogue at microwave frequencies and forces designers to treat link range and modulation bandwidth as coupled parameters rather than independent ones.
Terahertz waves propagate quasi-optically. Diffraction around obstacles is weak, so links are effectively line-of-sight, and a human body crossing the path can attenuate the signal by tens of decibels within milliseconds. Surface roughness comparable to the wavelength scatters energy diffusely, so painted walls and ceiling tiles behave less like mirrors than they do at lower frequencies, and the specular reflections that indoor systems rely on for backup paths lose several decibels more than a microwave designer would expect. Material behavior is also counterintuitive: paper, plastics, ceramics, and most textiles transmit terahertz radiation, while metals reflect it completely and water absorbs it strongly. Weather affects outdoor links, but not as severely as the frequency scaling suggests. Because raindrops are much larger than the wavelength, rain attenuation approaches a frequency-independent plateau above roughly 100 GHz, adding on the order of 10 decibels per kilometer in heavy rain instead of climbing without limit. Fog, high humidity, and airborne dust add further loss, and turbulence-induced beam wander becomes a factor on longer outdoor paths.
Atmospheric Absorption and Propagation Windows
Atmospheric absorption determines which frequency bands and link distances are viable, so it deserves separate quantitative treatment. Water vapor dominates. Its rotational resonance lines are strong and closely spaced across the terahertz range, and at the major lines—among them the transitions near 557 GHz, 752 GHz, and 988 GHz—specific attenuation reaches thousands of decibels per kilometer in humid air. Oxygen contributes a dense complex of lines around 60 GHz and an isolated line near 119 GHz, both below the terahertz range proper but relevant to the sub-terahertz systems that share the same design problems. Between these peaks lie windows where attenuation falls to a few decibels per kilometer or less, still far above the fraction of a decibel per kilometer typical at microwave frequencies.
Outdoor terahertz links target windows where absorption stays below roughly 10 to 20 dB/km in standard conditions. Those windows are not fixed. Water-vapor absorption scales with absolute humidity, so a window that supports a kilometer-long link on a cold dry morning may support only a few hundred meters on a humid afternoon; altitude, temperature, and pressure shift both the depth and the width of each window. Reliable outdoor operation therefore requires link margin, adaptive coding and modulation, or the ability to fall back to a lower band. Above the troposphere the constraint disappears almost entirely, which is why satellite crosslinks and high-altitude platform systems can exploit terahertz frequencies that ground-level systems cannot use at all.
Indoor and very short-range systems face a different balance. Over a few meters, even an absorption peak of 1,000 dB/km contributes only a few decibels, so the entire band becomes usable and the binding constraints become material blockage, multipath, and available transmit power. Dry, climate-controlled environments such as data centers relax the absorption problem further. Some designs turn absorption to advantage, deliberately operating near a water line so that the signal attenuates rapidly beyond the intended range, which limits eavesdropping and enables dense spatial reuse. More generally, the strong frequency dependence rewards adaptive systems that choose an operating window based on measured link distance, humidity, and interference conditions rather than committing to a fixed channel plan.
Ultra-High Data Rates and Modulation
The bandwidth available at terahertz frequencies supports data rates measured in hundreds of gigabits per second, orders of magnitude beyond deployed wireless systems. A widely cited 2013 experiment transmitted 100 Gbit/s over 20 meters on a 237.5 GHz carrier by generating the carrier photonically in a uni-traveling-carrier photodiode and receiving it with an integrated millimeter-wave circuit. Later work has pushed aggregate throughput past 1 Tbit/s over 100 meters near 300 GHz by multiplexing several photonic-wireless channels. These remain laboratory results with bulky optical front ends, but they establish that the channel itself supports terabit-class capacity.
Standardized systems are more modest and more practical. IEEE 802.15.3d, published in 2017, defines a switched point-to-point physical layer spanning roughly 252.72 to 321.84 GHz. It divides the band into sixty-nine overlapping channels built from eight bandwidth options, each an integer multiple of 2.16 GHz and ranging from 2.16 GHz up to 69.12 GHz, with a 4.32 GHz channel as the default. Two physical layer modes are specified: a single-carrier mode supporting modulation from binary phase-shift keying through 64-state quadrature amplitude modulation, and a simpler on-off keying mode intended for low-complexity, low-power transceivers. The single-carrier mode reaches 100 Gbit/s in the widest channel.
The choice between single-carrier and multi-carrier waveforms is contested at these frequencies. Orthogonal frequency-division multiplexing (OFDM) handles frequency-selective channels gracefully but produces a high peak-to-average power ratio, which is punishing when terahertz power amplifiers already operate close to their limits and offer little linear headroom. Single-carrier transmission with frequency-domain equalization keeps the peak-to-average ratio low and is consequently favored for the widest terahertz channels. Whichever waveform is used, multi-gigahertz signals expose dispersion in both the atmosphere and the hardware, and the resulting inter-symbol interference must be equalized digitally—at sampling rates that make the analog-to-digital converter one of the most demanding components in the system.
Higher-order modulation multiplies spectral efficiency but depends on phase stability. Terahertz local oscillators are typically synthesized by multiplying a microwave reference, and multiplication by a factor of N degrades phase noise by 20 log N decibels, so a times-24 chain adds roughly 28 decibels. This penalty, not thermal noise, usually limits how dense a constellation a terahertz link can carry, and it drives the use of pilot-aided carrier recovery and, in photonic systems, of digital or optical phase-locking schemes.
Spatial techniques extend capacity beyond what a single beam provides. Polarization multiplexing doubles throughput by carrying independent streams on orthogonal polarizations. Line-of-sight MIMO exploits the fact that at terahertz wavelengths even modest antenna separations produce the phase differences needed for spatially multiplexed streams over a direct path, a regime that requires impractically large spacings at microwave frequencies. Orbital angular momentum multiplexing, which encodes data on orthogonal helical wavefront modes, has been demonstrated over short terahertz links, though mode divergence and alignment sensitivity have so far kept it in the laboratory. Combining these approaches with wideband channels is the route by which terahertz systems are expected to reach multi-terabit aggregate rates.
Short-Range Terahertz Communication Systems
Short range is not a defect of terahertz communications so much as the condition under which its advantages appear. Every application that suits the band shares three traits: the distance is small enough that path loss and absorption remain manageable, the geometry is controlled enough that a narrow beam can be aimed and held, and the bandwidth requirement is high enough to justify the hardware. The IEEE 802.15.3d standard names the resulting use cases explicitly—intra-device links, kiosk downloading, data center connectivity, and wireless backhaul or fronthaul.
Kiosk downloading is the canonical consumer scenario. A user places a device on or near a terminal, and a link of a few centimeters transfers a feature film or a system image in seconds. The extremely short range removes almost every propagation objection: absorption is negligible over centimeters, alignment is mechanical rather than electronic, and the confined field limits interception. Wireless virtual and augmented reality presents a harder version of the same problem. Untethered headsets need multi-gigabit downlinks at latencies of a few milliseconds to render foveated video without perceptible lag, and a terahertz link can supply the throughput—but the headset moves continuously, so beam tracking and blockage recovery become the design bottleneck rather than raw capacity.
Backhaul and fronthaul for dense radio access networks form the most likely early infrastructure application. Small cells deployed at street level need multi-gigabit connections that fiber cannot always reach economically, and rooftop or lamppost-mounted terahertz links can span a few hundred meters over fixed, professionally aligned paths. Fixed installation solves the tracking problem, and mounting the equipment above pedestrian level avoids most human blockage. Indoor wireless local area networks in the terahertz band remain more speculative: they could serve stationary or slowly moving users at speeds no lower-frequency system can match, but they would complement rather than replace conventional Wi-Fi, which offers the coverage and mobility that terahertz cannot.
Wireless Data Centers
Wireless data centers represent a transformative application of terahertz communications, potentially replacing thousands of fiber-optic cables with reconfigurable wireless links operating at terabit-scale data rates. Traditional data center networks use fixed fiber connections between servers, switches, and storage systems, creating cable management challenges and limiting flexibility when workloads change or equipment is reconfigured. Terahertz wireless links could dynamically establish connections on demand, optimizing network topology for current workload patterns and eliminating physical cable constraints.
The short-range, high-bandwidth characteristics of terahertz communications align well with data center requirements. Rack-to-rack distances typically measure just a few meters, well within terahertz link budgets even accounting for atmospheric absorption and component losses. The controlled indoor environment of a data center eliminates weather-related propagation variability, holds humidity within a narrow range, and permits fixed line-of-sight paths above the racks that people rarely cross. Beamforming and beam steering techniques could create point-to-point links that minimize interference and maximize spatial reuse, allowing many simultaneous terahertz links within the same data center space.
Hybrid architectures combining fiber-optic cables for stable baseline connectivity with terahertz wireless for dynamic traffic demands may emerge as practical deployment models. These systems could maintain critical communications on fiber while using terahertz links to handle traffic bursts, support temporary computing clusters, or provide failover paths for robustness. Management and orchestration systems would need to coordinate radio resource allocation, beam steering, and handover between access points to maintain the ultra-high reliability required for data center operations. Power consumption, thermal management, and electromagnetic compatibility in dense terahertz transceiver deployments present significant engineering challenges for practical implementation.
Chip-to-Chip Communications
Chip-to-chip and board-to-board communications using terahertz frequencies address the growing interconnect bottleneck in high-performance computing systems. As processor speeds increase and multi-chip modules become more prevalent, the bandwidth and latency of inter-chip communications increasingly limit overall system performance. Electrical interconnects face fundamental physical constraints on bandwidth and power consumption, while optical interconnects, though high-bandwidth, require complex interfaces and precise alignment. Terahertz wireless interconnects could provide a middle ground, offering very high bandwidth without the constraints of physical contacts or optical alignment.
On-chip antennas integrated directly in processor or memory chips could transmit and receive terahertz signals across air gaps of just millimeters to centimeters. These short distances enable very high data rates even with modest transmit powers, while the dielectric environment between chips can be carefully controlled to optimize propagation. Beamforming with small antenna arrays ensures that signals reach intended receivers with minimal crosstalk to other chips, enabling simultaneous multi-chip communications. The wireless nature of these links provides flexibility in chip placement and thermal expansion management that rigid electrical connectors cannot match.
Practical implementation faces significant challenges including power consumption of terahertz transceivers, thermal management of high-power-density chips, and co-design of antennas with chip metallization and packaging structures. Maintaining link reliability in the presence of thermal gradients, mechanical vibrations, and electromagnetic interference from other chip functions requires robust system design. Nonetheless, demonstration systems have shown the feasibility of multi-gigabit chip-to-chip terahertz links, and continued advances in integrated circuit technology may make this approach viable for future high-performance computing applications, particularly in specialized systems where the benefits justify the additional complexity.
Terahertz Imaging Applications
The sections that follow turn from communication to sensing. The two domains share sources, detectors, optics, and much of the signal processing, and for most of the field's history sensing has funded the device advances that communication now depends on—so the applications below are not a digression but the commercial base beneath the technology. Terahertz imaging exploits the distinctive interaction of terahertz radiation with materials to form images from transmission, reflection, or emission. Unlike X-rays, terahertz radiation is non-ionizing, so it damages no chemical bonds and poses only a mild surface-heating hazard at the power levels imaging systems use, while still penetrating many optically opaque materials such as clothing, paper, packaging, and some building materials. Time-domain terahertz imaging systems use ultrashort pulses to measure both amplitude and phase, enabling three-dimensional reconstruction and spectroscopic identification of materials based on their terahertz absorption and dispersion signatures.
Continuous-wave terahertz imaging systems offer simpler hardware and faster imaging rates, trading some spectral information for improved real-time performance. Focal plane arrays of terahertz detectors enable video-rate imaging without mechanical scanning, though fabricating large detector arrays at terahertz frequencies remains challenging. Passive imaging systems detect naturally occurring terahertz radiation emitted by objects at different temperatures, similar to infrared thermal imaging but at longer wavelengths that penetrate clothing and some building materials.
Synthetic aperture techniques and computational imaging methods enhance resolution beyond the diffraction limit imposed by terahertz wavelengths. Near-field imaging with subwavelength apertures or structured illumination provides resolution approaching that of optical microscopy in certain configurations. These imaging capabilities complement communication applications, with many technological developments benefiting both domains, particularly in source and detector development, beam steering, and signal processing.
Security Screening Applications
Body screening is the best-known application of this frequency region, and it is also the one most often described imprecisely. The scanners actually deployed in airports are millimeter-wave systems, not terahertz systems: the advanced imaging technology units in United States airports illuminate the subject with roughly 24 to 30 GHz radiation at output powers on the order of tens of microwatts, and other commercial portals operate near 70 to 80 GHz. Those frequencies were chosen deliberately, because they penetrate clothing well, tolerate humidity, and are supported by mature, affordable hardware. True terahertz screening promises finer spatial resolution, since resolution scales with wavelength, but it remains largely a research and niche-deployment technology.
The physical basis is nonetheless the same across the range. Clothing and most packaging are largely transparent, skin reflects strongly, and metal reflects almost completely, so concealed objects appear as contrast against the body outline—without contact and without ionizing radiation. Early expectations that spectroscopy would go further and identify specific explosives or narcotics at a distance have proved much harder to meet than laboratory measurements suggested. Many such materials do exhibit characteristic terahertz absorption features, but at room temperature those features broaden considerably, atmospheric water vapor imposes its own absorption on the return path, and rough surfaces and intervening fabric scatter the signal. Standoff spectroscopic identification of concealed substances therefore remains an open research problem rather than a fielded capability.
Practical portal systems work at distances of a few meters and must balance image quality, throughput, privacy, and detection sensitivity. Privacy concerns drove a significant redesign of deployed systems, which now display a generic body outline with flagged regions rather than an anatomical image. Active illumination gives controlled, repeatable imaging conditions; passive systems that sense the natural thermal emission of the body and its surroundings avoid emitting anything at all, at the cost of far lower signal. Automated threat recognition, increasingly built on machine learning, analyzes the returns and marks suspect regions, which reduces operator workload and improves consistency across shifts and sites.
Package and mail screening systems use terahertz transmission imaging to inspect contents without opening containers. These systems can identify suspicious materials, verify contents match shipping manifests, and detect concealed compartments. The non-destructive nature of terahertz inspection makes it ideal for screening valuable or sensitive items. Combined with automated handling systems, terahertz screening can inspect high volumes of packages with minimal delay, supporting both security and logistics applications.
Medical Imaging Potential
Medical applications of terahertz radiation rest on one dominant contrast mechanism: water. Terahertz absorption in liquid water is extremely strong, so images are governed largely by tissue hydration, and because many pathologies alter local water content, that sensitivity yields useful contrast between healthy and diseased tissue. The same property sets the principal limit, since penetration into hydrated tissue is measured in fractions of a millimeter. Terahertz medicine is therefore a surface and near-surface modality, not a competitor to computed tomography or magnetic resonance imaging. Skin cancer assessment is the most studied application, and clinical studies have reported measurable contrast between normal skin and basal cell carcinoma, though the work remains at the research stage and no terahertz system has entered routine diagnostic practice. Non-ionizing operation does make repeated examination straightforward from a radiation-safety standpoint.
Dental imaging applications use terahertz radiation to detect caries, cracks, and defects in teeth and dental restorations. The dielectric properties of tooth enamel, dentin, and decay differ significantly at terahertz frequencies, enabling early detection of dental problems before they become visible through conventional examination. Burn wound assessment with terahertz imaging can determine burn depth and monitor healing, potentially improving treatment decisions and outcomes for burn patients.
Pharmaceutical applications include tablet coating inspection, where terahertz imaging non-destructively measures coating thickness and uniformity. Time-domain terahertz spectroscopy can identify and quantify active pharmaceutical ingredients, detect counterfeit drugs, and monitor manufacturing quality. Intraoperative guidance systems using terahertz imaging could help surgeons distinguish tumor margins from healthy tissue during cancer surgery, though significant development remains needed to create practical clinical instruments. The translation of terahertz imaging to routine medical use faces challenges including system cost, size, imaging speed, and integration with clinical workflows.
Material Characterization
Terahertz spectroscopy provides powerful capabilities for material characterization, exploiting the fact that many materials exhibit distinctive absorption, transmission, and reflection characteristics at terahertz frequencies. Time-domain terahertz spectroscopy measures the complete electric field of terahertz pulses, both amplitude and phase, allowing extraction of complex refractive index and absorption coefficient as functions of frequency. This spectroscopic information reveals molecular vibrations, crystal structures, and electronic properties that are invisible to other characterization techniques.
Pharmaceutical industry applications include polymorphism detection, where different crystal forms of the same drug molecule exhibit distinct terahertz spectra, ensuring that correct and stable forms are manufactured. Coating thickness measurements with terahertz time-domain systems provide non-contact, non-destructive characterization of thin films on various substrates. Semiconductor industry applications characterize carrier density, mobility, and other electronic properties of materials using terahertz transmission or reflection measurements, complementing electrical and optical characterization methods.
Art conservation and archaeology leverage terahertz imaging to examine paintings, manuscripts, and artifacts, revealing under-drawings, detecting alterations, and identifying pigments without damaging precious works. Food industry applications assess moisture content, fat distribution, and detect foreign objects in packaged foods. Polymer and composite characterization uses terahertz techniques to measure thickness, detect voids and delaminations, and monitor curing processes. The non-destructive, non-contact nature of terahertz measurements makes them particularly valuable for quality control in manufacturing processes where traditional destructive testing is impractical or where contact-based measurements might damage products.
Regulatory Considerations and Spectrum Allocation
Regulatory frameworks for terahertz communications took concrete shape only recently. The 2019 World Radiocommunication Conference, acting under agenda item 1.15, identified four bands within 275 to 450 GHz—275 to 296 GHz, 306 to 313 GHz, 318 to 333 GHz, and 356 to 450 GHz, a total of 137 GHz—for land mobile and fixed service applications with no special conditions required to protect passive users. Three further segments, 296 to 306 GHz, 313 to 318 GHz, and 333 to 356 GHz, may be used by those services only where conditions protecting the Earth exploration-satellite service (passive) have been established. The carve-outs exist because the same water and oxygen lines that attenuate communication signals are precisely what radio astronomers and atmospheric sounding satellites observe, so the strongest absorption features are reserved rather than shared.
National regulators have moved in parallel. In March 2019 the United States Federal Communications Commission adopted its Spectrum Horizons rules, creating a license class for experimentation anywhere between 95 GHz and 3 THz for terms of up to ten years, with provisions for marketing equipment during the experimental period. The same order designated 21.2 GHz of unlicensed spectrum across four bands: 116 to 123 GHz, 174.8 to 182 GHz, 185 to 190 GHz, and 244 to 246 GHz. Before that action, no rules permitted licensed or unlicensed communication above 95 GHz in the United States at all. The light-touch approach reflects the physics: high atmospheric absorption and pencil beams confine interference so effectively that the coordination burden of traditional licensing is hard to justify for short-range systems.
Standardization has advanced furthest at IEEE. IEEE 802.15.3d-2017 was the first published wireless standard to operate above 250 GHz, and its provisions were subsequently folded into the consolidated IEEE 802.15.3 revision. IEEE 802.15 working groups continue to develop channel models and complementary amendments for the band. The Third Generation Partnership Project has not yet standardized a terahertz air interface, but the ITU-R framework for IMT-2030—the requirements basis for what the industry calls 6G—identifies bands above the current millimeter-wave ranges as candidate spectrum under study, and sub-terahertz measurement campaigns feed directly into that process. Safety regulation is comparatively settled: because terahertz radiation is non-ionizing and penetrates only a fraction of a millimeter into skin, international exposure guidelines treat it as a surface heating hazard, and the resulting power density limits sit far above what current sources produce. Export controls remain a live consideration for high-performance terahertz components, given their dual-use value in imaging and sensing.
Challenges and Future Directions
The single most limiting problem is transmit power. Solid-state sources near 300 GHz deliver milliwatts, and power-added efficiency typically falls into the low single-digit percentages, against tens of percent for a well-designed microwave amplifier. Every other constraint follows from this: link budgets stay tight, beams must stay narrow, and range stays short. Progress depends on device-level advances—better III-V and SiGe transistors, efficient on-chip power combining, and higher-power photomixers—rather than on system-level cleverness alone. Digital hardware imposes a second bound. A 69 GHz channel demands analog-to-digital conversion and baseband processing at rates that consume watts of power in current process nodes, which is difficult to reconcile with mobile or embedded deployment.
Blockage and mobility remain unsolved at the system level. A pencil beam that a hand or a passing person can interrupt requires beam management that reacquires within milliseconds, and the beam-search problem grows with the square of the array size. Channel models for the band are still maturing, which complicates the standardization and network planning that commercial deployment presumes. Integration and packaging add practical friction: interconnects that are electrically negligible at microwave frequencies become significant transmission lines above 100 GHz, so waveguide transitions, antenna-in-package structures, and low-loss substrate materials all require careful co-design. The shortage of commercial off-the-shelf components forces custom development, which slows progress and keeps costs high.
Several research directions address these constraints directly. Reconfigurable intelligent surfaces—engineered panels of tunable reflecting elements—aim to create controllable non-line-of-sight paths around obstacles, converting blockage from a link failure into a routing decision. Joint communication and sensing exploits the fact that a terahertz transceiver is already a capable radar, letting the same waveform carry data and map the environment, which in turn supports predictive beam steering. Machine learning is being applied to beam selection, blockage prediction, and adaptive modulation, where the state space is too large for exhaustive search. On the device side, work continues on uncooled high-power sources, large-format detector arrays, and photonic integration that would replace bench-scale optical front ends with chip-scale equivalents.
Convergence with adjacent technologies is likely to shape deployment more than terahertz systems operating alone. Hybrid architectures that pair terahertz links with millimeter-wave or free-space optical fallbacks can preserve availability when a beam is blocked or the weather turns, and photonic integrated circuits offer a path to transmitters that share a fabrication base with optical communications.
Conclusion
Terahertz communications has moved from speculation to demonstrated engineering, but not yet to deployment. The physics is settled and mostly favorable: the band offers contiguous bandwidth measured in tens of gigahertz, links have carried 100 Gbit/s over tens of meters, and a first standard, IEEE 802.15.3d, defines channels up to 69.12 GHz wide. Regulators have cleared the way, with 137 GHz identified internationally between 275 and 450 GHz and unlicensed spectrum opened above 95 GHz in the United States.
What holds the technology back is hardware economics rather than fundamentals. Sources deliver milliwatts at efficiencies of a few percent, wideband baseband processing consumes watts, and narrow beams make blockage a first-order design problem. These constraints do not preclude success; they select the applications. Terahertz systems will arrive first where the range is short, the geometry is controlled, and the bandwidth requirement is extreme—data center interconnects, wireless backhaul, kiosk transfers, and chip-to-chip links—rather than as a general replacement for lower-frequency wireless. The sensing and imaging capabilities that share the same hardware may well reach commercial scale sooner, and the two application domains will continue to pull the underlying device technology forward together.
Further Learning
Three bodies of background knowledge carry the most weight here. The first is electromagnetic propagation at millimeter and submillimeter wavelengths, including atmospheric absorption modeling and quasi-optical beam analysis, since the channel dictates so much of the system design. The second is high-frequency semiconductor device physics—what sets transit frequency and maximum oscillation frequency, and why breakdown voltage and speed trade against one another—because source power is the field's governing constraint. The third is wideband digital communication theory, particularly equalization, carrier and phase recovery, and the practical limits of high-speed analog-to-digital conversion.
Hands-on experience with microwave and millimeter-wave measurement transfers directly, and the habits it builds around calibration, connector loss, and vector network analysis matter more as frequency rises. For primary sources, read the IEEE 802.15.3d standard alongside the ITU Radio Regulations provisions covering 275 to 450 GHz; together they show how physics, spectrum policy, and system specification interlock. Current research appears in the proceedings of the International Conference on Infrared, Millimeter, and Terahertz Waves and in journals covering photonics, microwave theory and techniques, and wireless communications. Because progress depends on both electronic and photonic approaches, following only one of the two literatures gives a distorted picture of where the field stands.