Everything we build rests on a chain that starts with a compass needle twitching in 1820 and ends with a laser that finally runs without liquid helium. These are the people, the experiments and the papers.
Electricity and magnetism turn out to be one thing. That thing turns out to be light. And light, it turns out, can be manufactured.
A dissected frog’s leg twitches when touched with charged metal. Galvani calls it “animal electricity” and is widely disbelieved. He is right: the nervous system is an electrical system, which is the entire premise of every neural interface built since.
During a lecture, Hans Christian Ørsted notices a compass needle deflect when he switches on a nearby current. Electricity and magnetism, assumed separate for centuries, are linked. Within months Ampère has the mathematics of currents acting on currents.
A changing magnetic field induces a current. Faraday, with almost no mathematics, introduces the idea of a field filling space, lines of force that are themselves physical. This is the conceptual leap everything else needs.
Maxwell writes down the equations governing the field and finds they permit travelling waves. He calculates their speed from purely electrical and magnetic constants, and gets the speed of light. His conclusion is one of the great sentences in physics: light is an electromagnetic disturbance.
Crucially, the equations set no upper or lower bound on frequency. Every band we would later use (radio, microwave, terahertz, X-ray) was implied here, decades before anyone could make one.
Heinrich Hertz builds a spark-gap oscillator and a resonant loop detector and generates electromagnetic waves on demand, then reflects, refracts and polarises them to prove they behave exactly as Maxwell’s light should. Asked about applications, he reportedly said there were none.
Two communities march toward each other across an unexplored stretch of spectrum, and meet briefly before the technology of the age abandons the middle.
In Calcutta, Jagadish Chandra Bose builds spark transmitters and galena detectors working down to 5 mm wavelength, or 60 GHz, a band industry would not seriously revisit until 5G. He builds waveguides, horn antennas, polarisers and dielectric lenses, essentially inventing millimetre-wave engineering, and declines to patent almost any of it.
From the optical side, Heinrich Rubens pushes into what he calls “heat rays of great wavelength,” using reststrahlen, residual-ray reflection from ionic crystals, to filter out ever longer wavelengths. It is a brute-force method, and it works.
Rubens and von Baeyer find that a quartz-envelope mercury lamp emits usefully far into the sub-millimetre, reaching wavelengths of several millimetres. The far-infrared is now reachable, barely, with enormous effort, and with almost no signal.
Working from the radio side upward and the optical side downward, Nichols and Tear finally overlap the two, demonstrating experimentally that radio waves and infrared light are the same phenomenon with no gap in nature between them. The spectrum is proven continuous.
And then, for fifty years, almost nobody goes back. The band is continuous in nature but miserable to work in, and both communities have easier problems elsewhere.
Two enormously successful technologies grow away from each other, and the space between them becomes a named problem.
The cavity magnetron and the wartime radar programme pour resources into electronics below ~30 GHz. Klystrons, waveguides and mixers mature into an industry. Electronic sources get very good, then hit a wall as frequency rises, because a transistor can only switch as fast as charge can cross it.
A quantitative model of the action potential in the squid giant axon: ion channels, conductances, a differential equation that reproduces the spike. For the first time the nervous system is not merely electrical but predictable: you can compute what a nerve will do.
Coherent optical sources transform the infrared and visible. But a semiconductor laser needs a bandgap at least as large as the photon it emits, and terahertz photons are tiny, a few millielectronvolts. Photonics can descend toward the band, and then it stops.
Electronics reaches up and fades. Photonics reaches down and fades. Between roughly 0.1 and 10 THz sits a stretch where neither approach delivers useful power, and where room-temperature thermal energy is larger than the photon you are trying to detect. Astronomers and spectroscopists work there heroically, with cryogenics and patience. Everyone else stays away.
The band is finally opened not by building a better oscillator, but by sidestepping oscillators altogether.
David Auston shows that a short laser pulse striking biased silicon creates carriers almost instantly, producing an electrical transient of picosecond duration. The insight that follows is the foundation of the entire modern field: a pulse that short necessarily contains terahertz frequencies. You do not need a terahertz oscillator. You need a fast enough switch.
Auston, Cheung and Smith put that switch at the feed of a small antenna. The transient radiates. A practical, room-temperature, benchtop terahertz emitter exists for the first time. The same structure, run in reverse and gated by the same laser, becomes the detector.
Grischkowsky and colleagues build the technique into a measurement science, sampling the terahertz electric field as a function of time rather than merely its intensity. Because you capture amplitude and phase together, you get the complex refractive index directly, with no Kramers-Kronig inversion, no assumptions about the sample. This remains terahertz’s single most under-exploited advantage.
A silicon microelectrode array with a hundred penetrating shanks, batch-fabricated. It becomes the workhorse of neural interfacing for three decades, and over those decades it teaches the field its hardest lesson: implants degrade. Glial encapsulation and micromotion erode signal quality over months to years.
Spectroscopy becomes imaging, imaging becomes medicine, and the band acquires its first genuine clinical case.
Hu and Nuss raster-scan a sample through a focused terahertz beam and reconstruct an image from the transmitted pulses. They image a packaged integrated circuit and a leaf, and note that the leaf’s contrast comes from water content, drying visibly over hours. The defining property of terahertz biological imaging is present in the very first picture.
Mittleman, Jacobsen and Nuss extend the method to reflection geometry and time-of-flight depth ranging. That is the architecture nearly every practical terahertz imager uses today, because you cannot shine through a person.
Köhler and colleagues build a laser that emits at 4.4 THz by engineering transitions between sub-bands of a superlattice rather than across a material’s bandgap, escaping the constraint that had blocked photonics from the band for forty years. It requires cryogenic cooling, and that limitation defines the next twenty years of research.
Woodward, Wallace and colleagues image basal cell carcinoma and find reproducible contrast between tumour and surrounding tissue, attributed to increased water content and structural change. The work moves from excised tissue to living patients, and terahertz has a real clinical target: surgical margins, determined while the patient is still open.
Taylor and colleagues demonstrate that terahertz reflectivity tracks tissue hydration closely enough to grade burn depth in vivo. That is the quantity that actually determines whether tissue is excised, and which clinicians otherwise assess by eye. It remains one of the clearest cases of terahertz answering a question no other modality answers well.
Sustained, deliberate investment finally attacks the hardware problem from both directions at once.
Under DARPA’s Terahertz Electronics programme, Northrop Grumman demonstrates a monolithic integrated circuit amplifying at 1.0 THz, using indium phosphide transistors operating where nobody had built working electronics before. The electronic side of the gap moves for the first time in decades, and it moves because a government agency decided to push it.
Two groups independently show that stimulating peripheral nerve fibres produces sensation the user localises to the missing hand, and that with feedback, grip force control improves and the limb begins to feel owned rather than worn. Tan’s implants remain stable for years, which is the harder result.
Nanostructuring the electrode geometry of a photoconductive emitter dramatically increases how much of the optical pump becomes terahertz, addressing the field’s oldest complaint, that everything is starved for power. Efficiency, not exotic physics, is what turns a laboratory technique into an instrument.
Khalatpour and Hu report terahertz QCL operation at temperatures reachable with a compact thermoelectric cooler rather than liquid cryogens, around 250 K. A portable terahertz laser stops being a contradiction in terms. This is the single result that most changes what a fielded instrument can be.
Lorach and colleagues restore volitional walking to a man with chronic spinal cord injury by decoding cortical intent and stimulating the spinal cord below the lesion: a digital bridge across the damage. It is proof that the wiring problem, not the biology, is often what stands between injury and recovery.
Two hundred years from a compass needle to a terahertz laser you can carry. The physics has been settled since 1865. What took the century and a half after that was learning to build in a band nature made continuous and engineering made miserable. The gap is not closed. But both sides are moving toward the middle at once, for the first time, and on purpose rather than by luck.
Elpis exists to take the instruments that investment produces and point them at people.
