MEMS transform infrared imaging

03 Jun 2003

Infrared imaging chips containing thousands of tiny silicon carbide cantilevers are about to rewrite the specifications of uncooled sensors - as well as reducing their price-tag. Rob van den Berg talks to the camera's makers.

From Opto & Laser Europe June 2003

The search for an infrared (IR) sensor that is uncooled, easy to mass-produce and capable of generating high-quality images has driven military-funded research for decades. It has been generally believed that the answer lay with microbolometers - devices that sense IR light by measuring the change in a pixel's electrical resistance - but a new contender has now begun to attract the attention of camera makers.

At the recent SPIE Aerosense Conference in Orlando, Florida, Sarcon Microsystems of Knoxville, Tennessee, and its partner the Sarnoff Corporation of Princeton, New Jersey, presented a prototype IR imaging system that exploits microelectromechanical systems (MEMS) technology. According to the developers, every pixel of the 320 x 240 pixel array can detect IR radiation with a sensitivity 10-20 times better than that of competing uncooled detectors. They also say that their MEMS sensor can be cost-effectively produced in standard complementary metal-oxide silicon (CMOS) foundries using conventional processing techniques.

When the first devices become commercially available in the first quarter of 2004, it is expected that they will find their way into a range of different IR cameras and could be used for anything from generating night images to seeing through smoke.

Sizing up the competition At present the market for IR imaging is dominated by cooled (cryogenic) sensors, which are based on semiconductor or quantum devices. The main attraction of cryogenic sensors is their sensitivity. They typically boast a noise-equivalent differential temperature (the smallest temperature difference that can be sensed between two objects) in the 5-10mK range, which approaches the theoretical limit for IR imaging. Another strong point is their ability to operate at rates of up to 1000 frames per second. The big drawback is that the cooling equipment adds considerably to the complexity, size and cost of the sensor.

This state of affairs may now be about to change, thanks to Sarcon and Sarnoff's MEMS-based sensor technology. Each pixel of their device consists of a miniature cantilever (or microcantilever) made out of amorphous hydrogenated silicon carbide. One end is attached to a fixed support structure, while the other end is allowed to bend.

"Since the cantilever is partly covered by a gold film, it acts just like the bimetal in a thermostat," explained Frank Pantuso, general manager of Sarnoff's imaging business. "By absorbing infrared radiation it heats up and bends towards the surface - approximately 0.1µm for every degree of temperature change. This leads to a change in the capacitance of the device and generates an electrical signal in the sensor that is proportional to the intensity of the infrared radiation absorbed. All unwanted external vibrations are damped using an actively tuned resonant RC-circuit." The cantilever sensors are then stacked in rows and columns to produce a two-dimensional array for generating infrared images.

Getting started The story behind the development of cantilever sensors began in the mid-1990s. "In 1995 we [Sarnoff] were selected by the Defense Advanced Projects Research Agency to start a research effort to improve the performance of uncooled IR detectors," explained Pantuso. However, Sarnoff was not the only firm that was carrying out such research. At around the same time, researchers at the Oak Ridge National Laboratory (ORNL) in Tennessee were busy working with Consultec Scientific, a small R&D company founded by former ORNL employees, to develop similar microcantilever-based IR sensors.

In June 1997 Consultec (which had exclusive rights to the ORNL microcantilever technology) and Sarnoff decided to pool their patents and related know-how, and form a new venture to commercialize the IR sensors. The firm that resulted was Sarcon Microsystems, a technology venture company jointly owned by Sarnoff, Consultec and the Sarcon management.

It was July 1999 when the collaboration's work to develop the sensors first began to pay off, and a prototype with a measured sensitivity of 300mK was produced. Just two months later, Sarcon Microsystems had added "breadboard" electronics and optics and generated the first infrared images from a microcantilever-based sensor array.

The original 16 x 16 pixel array detector had a sensitivity of around 100mK. By July 2000, 32 x 32 pixel sensor arrays had been built, and today's resolution has been significantly increased to 320 x 240 pixels. That figure is likely double before long, while the same fabrication costs will be retained.

Although there could potentially be overlapping patent claims that have not been resolved yet, Sarcon now has the exclusive right to manufacture and market infrared imaging products based on microcantilever sensor technology.

Building a reputation Microcantilever sensors show great promise as imaging chips for use in low-cost, high-performance infrared cameras. "These may be used for making IR measurements - for instance for process control in a factory - or to monitor overheating in equipment," said Pantuso. "But obviously they are also able to generate an image when there is no visible light at all, as may be the case in security surveillance or fire-fighting, for example."

Other uncooled sensor technologies rely on thermoresistive materials such as amorphous silicon or vanadium oxide that change their resistance when heated. "Whereas thermoresistive materials show a response of a few per cent per degree, the sensitivity of our sensor is up to 50% per degree," said Pantuso. "We are confident that by selecting materials with a better thermal coefficient of expansion, it is even possible to match the sensitivity of cooled detectors and approach the theoretical limits of infrared sensing."

The high sensitivity of the cantilevers should help with this, as it allows the individual pixels in the imaging array to be shrunk from their current size of 50µm to 20-30µm while maintaining low-distortion IR imaging. Smaller pixels make it possible to make larger-sized arrays for high spatial resolution imaging.

However, work still remains to be done before MEMS-based sensors start hitting the market. According to David Smith, vice-president of sales and marketing at Sarcon: "Our modules are not yet good enough for potential customers to evaluate. We are busy redesigning the CMOS - the electronic read-out architecture underneath the cantilevers - as we encountered a problem with parasitic capacitances in the sensor. This can, however, easily be solved using shielding layer techniques."

In addition, the detectors will need some further refining. "We also still have to improve the uniformity of the individual detectors across the device. In the first quarter of next year we expect to be able to start handing them out to OEMs [original equipment manufacturers]. They will have to evaluate whether they want to build our modules into their cameras," said Smith.

Both Pantuso and Smith are confident about the future. They say that they have received a tremendous amount of interest from camera manufacturers in the US. The only problems Smith foresees are those faced by every new technology - building up credibility and proving that the sensors can live up to their expected performance.