IEEE Spectrum July, 2007 - 39

it forms degrades, as does its spectroscopic signal. Fortunately,
the terahertz spectrum has a few transmission windows at frequencies that aren't strongly absorbed in air. So one solution is to
generate a continuous wave at one or more of those frequencies.
Researchers are already making such continuous-wave sources,
basically with the same setup of a laser shining on the surface of
an antenna-equipped semiconductor, but with the femtosecond
laser replaced with a continuous one whose amplitude is oscillating at a terahertz frequency [see illustration "T-Ray Scanner"].
You start by focusing two infrared lasers in a device called a
photomixer, with the lasers tuned so that the difference between
their frequencies is a frequency corresponding to one of the
terahertz transmission windows. The photomixer combines
the lasers so that the resulting light "beats" at this terahertzfrequency difference. The beating laser drives a similar
photoconductor-antenna structure to the one used to generate
pulses, causing current to flow through it at the terahertz-beat
frequency, thereby generating many microwatts of T-rays.
The method was demonstrated over a decade ago but became
practical only a few years ago, thanks to pioneering work by
researchers at the imaging start-up TeraView. The key was in
a new type of photomixer, made of indium-gallium-arsenide,
which could efficiently mix lasers of a wavelength easily carried
on optical fibers. Channeling the lasers on optical fibers instead
of having to carefully align laser beams with expensive optics
has greatly simplified terahertz imagers and has also had the
added benefit of driving down their cost.
These optoelectronic methods work well enough, but they are
of limited brightness and are still quite cumbersome. What terahertz researchers really want is to replace these technologies with
a bright, completely solid-state terahertz laser. It's their best hope
of getting imagers smaller, lighter, and cheap enough to massproduce, not only because the light source is smaller but also
because its higher brightness would allow for less expensive and
more compact detector arrays. Unfortunately, the wavelength of
semiconductor lasers is largely determined by the materials that
are used to make them, and none naturally produce T-rays.
A device called a quantum cascade laser, invented in 1994 by
Federico Capasso, among others, at Bell Labs, could be a big part
of the answer. Unlike other semiconductor lasers, QC lasers can
be engineered to emit any of a range of micrometer-wavelength
light, including terahertz wavelengths. The secret to the QC laser
is that its wavelength is determined by the thicknesses of the
layers of semiconductor that make it up-something that can be
carefully controlled.
Here's how it works: lasers emit light when electrons that
have been excited to a particular energy level fall to a lower
energy level. A key difference between a laser and an ordinary light emitter is that there are always more electrons in
the excited state than in the lower energy level. In a QC laser
that aspect is guaranteed by sweeping fallen electrons from the
lower, unexcited state into a third state, at a still lower energy
level. In the QC laser these energy levels exist in three layers
called quantum wells, each nanometers thick . Quantum wells
are structures so thin that, from an electron's perspective, they
are two-dimensional. Confinement in a quantum well makes the
electrons behave as though they were bound to an atom, with
their energy constrained to certain specific levels.
An electron injected into the highest energy level falls to the
lower one, emitting radiation (photons) of a wavelength that is
determined by the thicknesses of the quantum wells. The electron then immediately falls into the still lower third state, emitting a quantum of heat called a phonon. What's really remarkable

www.spectrum.ieee.org

is that this same three-layer structure can be repeated more than
two dozen times. At each structure the electron goes through
exactly the same dance, emitting the same color photon. So
a single electron can emit 24 or more photons on its journey
through the QC laser, as if it were falling down a set of stairs
and emitting a photon at each step.
Last year, researchers at Sandia National Laboratories, in
Albuquerque, used QC lasers to produce 138 milliwatts of terahertz
laser power-a record. The one catch, and it's a pretty big one, is
that QC lasers must be cooled to within 10 degrees of absolute
zero to perform at that level. At liquid nitrogen temperature, 77 K,
QC lasers can't even crack 10 mW. So the aim of much of the
research into terahertz QC lasers is in improving the power output
at higher and higher temperatures. The dream is a terahertz QC
laser that operates at room temperature, or at least at 250 K, which
is in the range of compact, inexpensive thermoelectric coolers.

T-rays can be deTecTed in a number of ways. But one of the
more common detector types is merely an extension of T-ray
generation technology. Recall the picosecond-pulse generators
and the continuous-wave generators. You can easily take the laser
beam, split it, and feed it to another photoconductive antenna
structure. But instead of applying a voltage across the antennas
to push current through them and generate terahertz radiation,
you measure the current through the antennas. As in the pulsegeneration scheme, when the laser pulse hits a photoconductor, it
creates short-lived pairs of electrons and holes. These then flow
through the antenna under the influence of the electric field of
incoming terahertz waves. So the current in the antenna, which
is amplified, acts as a measure of terahertz radiation.
Because the detector is sensing T-rays only during the picosecond or so that the laser pulse allows, it takes several pulses
to get the full waveform of the incoming T-rays. To get the full
waveform, small increments of delay in the form of a longer path
for the laser are added to the detector's optical fiber line. Measuring
the electric field at a number of increments produces slices of the
terahertz wave that can be pasted together in a computer.
The same scheme works when pairs of tuned lasers are used
instead of pulses. Recall that the lasers are tuned so that the
difference in their frequencies is equal to a terahertz frequency.
When T-rays hit the antenna, they mix with the terahertz frequency of the combined lasers to produce a dc signal. These two
schemes are how detectors work in systems built by TeraView,
Picometrix, and others.
Of course, detection is only the start of image making. The
simplest way of producing an image is to scan a single transmitter and detector over an object and record the phase and amplitude of the T-rays that reflect back at each point. State-of-theart terahertz-imaging security systems are capable of such raster
scanning at a rate of 100 pixels per second, certainly not fast
enough for video and only marginal for scanning a bag on a conveyor belt. A briefcase containing a gun, a glass bottle, and a knife
would take half an hour to scan at a resolution of 1.5 millimeters
per pixel using a T-ray pulse-based imager from Picometrix.
Although there are no terahertz camera chips, there are infrared camera chips, and you can tweak those so they can pick up
T-rays. Such chips detect infrared radiation at each pixel because
the radiation reduces the resistance of a minuscule patch of semiconductor there. By themselves, some of these chips are slightly
sensitive to terahertz radiation, but to get a decent image you need
a bright source such as the QC lasers under development.
Another infrared detector concept is electro-optic terahertz imaging. In this scheme, T-rays striking certain types of

July 2007 | IEEE Spectrum | NA

http://www.spectrum.ieee.org

Table of Contents for the Digital Edition of IEEE Spectrum July, 2007