Seeing Heat in an Uncool Way
Optical probing offers readout of thermal transducers without the need for wiring architectures.
Putting a new twist on an old idea, researchers from Oak Ridge National Laboratory in Tennessee, L-3 Communications Cincinnati Electronics in Mason, Ohio, and The Pennsylvania State University in Freeport have built an infrared imager out of metallized microcantilevers.
An infrared imaging chip has been created using cantilevers as pixels. Each cantilever is formed from a pair of materials with disparate coefficients of thermal expansion — in this case, a layer of gold atop a layer of silicon nitride.
The uncooled imager is an updating and miniaturization of an idea that is more than 50 years old, according to Panos G. Datskos, a senior researcher at the national laboratory as well as a research professor at the University of Tennessee in Knoxville. He noted that his group’s approach provides some benefits compared with other IR imagers. “The biggest advantages are in the performance, cost, spectral tuning and large scalability.”
The novel imager exploits the difference in the expansion coefficients of the materials that comprise the device. That variation makes structures bend because one of the materials expands more than the other for a given temperature change; once the detector is calibrated, users can track temperatures directly.
In their work, the researchers constructed bimaterial devices using manufacturing techniques that are commonly used for microelectromechanical systems. One of the materials was a deposited 600-nm-thick silicon nitride layer that was fashioned into freestanding cantilevers. Each lever represents one pixel. The second material was a 120-nm-thick layer of gold, which sat atop the silicon nitride.
Because silicon nitride has a coefficient of thermal expansion of 0.8 × 10–6 K–1 while that of gold is 14.2 × 10–6 K–1, it absorbed infrared radiation, and the cantilevers bent when exposed to infrared. When that happened, the mirrorlike gold provided another function. “It also served a secondary role of the reflector for the optical readout,” Datskos said.
This approach meant that the imager readout could be done without anything contacting the device, and measurements could be performed without any electronics in the array.
The researchers used the cantilevers to construct 256 × 256-pixel arrays composed of 50-μm square pixels with a 62 percent fill factor. In tests, they found that the arrays were noisier than expected. The main reason for the discrepancy was noise in the optical readout, Datskos explained.
The response time was 6 ms, fast enough for real-time video IR imaging. In addition, they reported that scaling the arrays up to much larger sizes — up to 2000 × 2000 pixels — should be possible.
Some device-to-device variation resulted from differences in the stress that was induced during fabrication. Solving that problem will be part of future research projects, according to Datskos, who also indicated that such variations would not be the only issue tackled.
“The next step is to fabricate larger arrays, [to use] better geometries to amplify the bimaterial effect and reduce stress, and to optimize the optical readout,” he said.
Applied Physics Letters, Aug. 14, 2006, 073118.
MORE FROM PHOTONICS MEDIA