17 April 2012—In the entire electromagnetic spectrum, one of the most conspicuously inaccessible chunks sits smack dab between radio waves and infrared light. Researchers have been trying for decades to come up with better ways to exploit the little-used terahertz band, which could provide ways to find hidden objects and determine an object’s chemical makeup at a distance.
Now a team from IEMN and STMicroelectronics, in France, and the University of Wuppertal, in Germany, has come up with a practical first: a video-rate CMOS camera that’s sensitive to terahertz frequencies.
“I think it’s the hottest thing in terahertz technology at the moment,” says Peter Siegel, who works on terahertz imaging at Caltech and NASA’s Jet Propulsion Laboratory and is not affiliated with the team. “They’ve done a remarkable job of solving a bunch of very pesky problems in working with silicon at high frequencies.”
Up until now, terahertz detectors have tended to be pricey affairs, composed of devices like Schottky diodes or microbolometers. A Schottky diode–based detector usually contains just one or a few pixels, which are raster-scanned across a scene to slowly form an image. Microbolometers can be arranged in arrays, but they must be cooled to boost their sensitivity.
With just 1024 pixels, this new transistor-based camera is unlikely to give a high-resolution window into the unseen terahertz realm. But the advance has researchers excited, because it suggests terahertz technologies may soon get a lot cheaper and more accessible. The single-pixel terahertz detectors in use now can easily cost as much as US $10 000, Siegel says, so developing a detector that could be mass-produced by chip manufacturers represents a significant advance. “I think you’re going to find a lot of applications opening up that didn’t exist before,” Siegel says.
Building terahertz detectors out of silicon is difficult, because even the best transistors don’t operate well at frequencies in excess of a few hundred gigahertz, the lower edge of the terahertz band. This limitation stems mainly from how fast electrons can shoot from one side of the transistor to the other, resulting in an intrinsic cutoff frequency above which a transistor can’t amplify signals sent into it. “Traditionally, people would say that beyond the cutoff frequency, the transistor wouldn’t work anymore,” says Hani Sherry, a doctoral candidate working at STMicroelectronics. But in 1996, device physicists Michel Dyakonov (now at the University of Montpellier, in France) and Michael Shur (of Rensselaer Polytechnic Institute, in Troy, N.Y.) argued in a paper that appeared in IEEE Transactions on Electron Devices that the cutoff frequency can be surpassed. Although they will not be able to amplify signals, some types of field-effect transistors can still respond to frequencies above the cutoff frequency due to electromagnetic oscillations within the transistor’s channel. (The channel is the main body of a transistor through which current flows when the transistor is on. It runs between two electrodes—the source and the drain—and is adjacent to a third, the transistor’s gate.)
The camera’s pixels, which were designed by Ullrich Pfeiffer and his colleagues at the University of Wuppertal, are made of transistors that surpass the cutoff frequency by using only a small patch of the channel. This patch of the channel is close to the transistor’s source electrode, which is connected to a copper ring antenna. The antenna is capable of picking up signals over a wide range of frequencies, from 0.6 to 1 terahertz. A turbulent process of “self-mixing” in the transistor channel turns wildly oscillating terahertz signals into a simple DC output, a voltage at the drain that’s proportional to the square of the amplitude of the incoming terahertz radiation.
The team constructed the camera chip, which was presented in San Francisco in February at the International Solid-State Circuits Conference, using a standard 65-nanometer CMOS chip-fabrication process. Both the transistors and antennas are part of the same chip, with the antennas embedded in wiring layers above the transistors. The ensemble is fixed to the back of a opaque-looking fish-eye lens made of silicon, which unlike glass is transparent to terahertz waves. The camera can capture frame rates of up to 25 frames per second and requires so little power it can be operated by USB.
Despite its finely tuned antennas and optics, the camera’s sensitivity is fairly low. While the photodiodes in a typical optical CMOS sensor can convert nearly all incoming photons into charge, Sherry says, terahertz photons are so low in energy that his team’s camera needs roughly 100 000 of them to produce a single electron.
Sensitivity aside, the development of this video-rate chip, along with other efforts to produce CMOS-compatible detectors, suggests that terahertz detection will soon become more mainstream. “It represents a turning point in terahertz technology,” says Sigfrid Yngvesson, an expert in terahertz detection at the University of Massachusetts Amherst. “We can now look forward to terahertz technology that is less exclusive and less expensive than what we’ve had.”
The applications for this sort of technology are still unclear. A large number of molecules emit and absorb terahertz radiation. And the waves also occupy a Goldilocks-like spot on the electromagnetic spectrum that could be good for body scanning. Because they are lower in frequency than X-rays, terahertz waves don’t have enough energy to ionize human tissue. But the waves are also higher in frequency (shorter in wavelength) than microwave or millimeter radiation, which means they should be able to produce higher-resolution images. As a result, terahertz sensors could lead to new security scanners and medical imagers that can see through clothes, as well as spectrometers with new capabilities.
Terahertz cameras might also be used for a number of niche applications, from food quality control to industrial monitoring of drying processes. “The trouble is, there are a lot of [other] technologies to do this too,” Siegel says. “It will always be a trade-off between cost and efficacy.”
Adding to the cost is the need for an external source of terahertz radiation to go with the camera. The Wuppertal team’s camera isn’t sensitive enough to pick up on ambient terahertz signals; radiation needs to be created and reflected off of objects in order to form an image, like the flash on an optical camera.
Unfortunately, cheap compact terahertz sources that can operate at room temperature have yet to emerge from the lab. To get enough radiation, researchers often resort to tuning the output of expensive optical pump lasers or devices made out of compound semiconductors such as indium gallium arsenide. One common source uses Schottky diodes, but that technique tends to require a lot of energy to convert lower-frequency radiation into terahertz waves. Terahertz radiation can also be generated using fairly small III-V devices called quantum cascade lasers, but those still need to be cooled well below the freezing point of water.
“The primary objective of using CMOS for terahertz is to reduce the cost and lower the power consumption, but I think that objective is lost the moment you put the III-V source into the system,” says Adrian Tang, who works on CMOS-based terahertz imaging at the University of California, Los Angeles. “You [get] an expensive, high-power system immediately.”
Even if terahertz detectors and sources can be made cheaply, there are still some basic physical limitations to contend with, Tang adds. Water vapor and oxygen in the atmosphere readily absorb terahertz photons, limiting an imager’s range. Researchers working on terahertz imaging also struggle with uniform reflectivity; objects over a wide range of depths reflect similar amounts of radiation, producing very cluttered images.
The lack of contrast has led some researchers, like Tang and Siegel, to focus on radar-like systems that measure the time of flight of terahertz waves in order to extract 3-D components from reflected signals. But progress is slow, Tang says: “Terahertz imaging still has a long way to go before significant impact or practical applications are possible.”
Siegel is more optimistic: “Terahertz imaging applications are on the verge of a significant leap forward, in both capability and cost-effectiveness.”
Rachel Courtland, an unabashed astronomy aficionado, is a former senior associate editor at Spectrum. She now works in the editorial department at Nature. At Spectrum, she wrote about a variety of engineering efforts, including the quest for energy-producing fusion at the National Ignition Facility and the hunt for dark matter using an ultraquiet radio receiver. In 2014, she received a Neal Award for her feature on shrinking transistors and how the semiconductor industry talks about the challenge.