Sofradir backs a two-tone approach
Choosing the correct infrared camera requires some care. Whether it is intended for surveillance, or for identifying a particular object, it is important to select the optimum spectral band to obtain the most informative images.
Unfortunately, the ideal waveband isn t fixed. It depends on factors such as humidity, air temperature and the size of the water droplets in the atmosphere. These all influence the transmission of infrared radiation and can have a significant impact on the quality of images generated by detectors operating over distances of several kilometers or more, such as those used for remote sensing of the earth from satellites and long-range surveillance. On top of this, single-band cameras can struggle in military applications, because flares and decoys can hamper the detection and identification of targets.
However, it s possible to address these issues by switching to infrared detectors and accompanying optics that can deliver high performance in multiple spectral bands. The penalty is greater complexity, as this multi-spectral detector requires pixels to deliver simultaneous detection in at least two wavebands.
Imaging options
Moving to just two bands equips the camera with several modes of operation. One option is to select the most appropriate band for a particular application, but two channels can also be used to provide a pair of images that convey different information. For example, a camera mounted on an orbiting satellite might provide one image that monitors cloud positions, and one that reveals pollution levels. The two images from a dual-band camera can then be merged to generate a single, more informative image.
For an all-weather camera, the best results come from pairing detection in the medium-wave infrared (MWIR) (3–5 µm) and long-wave infrared (LWIR) (8–12 µm). Cameras with this feature should make a big impact on the detector market, as long as the cost premium over a single-band detector is acceptable. However, meeting this criterion demands the deployment of a cost-effective technology that is applicable to high-volume manufacturing.
Today, it s only possible to get dual-band performance with cryogenically cooled designs based on a stack of compound semiconductor layers, which feature separate regions for the absorption of each spectral band. InAsGaSb type-II superlattice designs are one option, alongside quantum-well infrared photodetectors (QWIPs) that are based on layers of GaAlAs and GaAs, and detectors based on the HgCdTe family of compounds.
Although the type-II superlattice structure can in principle cover different wavebands, today s designs can only provide detection in the MWIR and SWIR (1–2.5 µm), while QWIPs are restricted to covering the MWIR and LWIR, or the LWIR and VLWIR (12–20 µm). But detectors based on HgCdTe can detect in all bands, while delivering superior performance to QWIPs.
The development of dual-band HgCdTe detectors has actually been underway for many years – the French national research and development center CEA-Leti started to develop this technology in the early 1990s, and it delivered the first proof-of-concept demonstration at the start of this decade. Since then it has been joined by a domestic partner, our research team at Sofradir, which is headquartered in Chatenay-Malabry and has production and development facilities in Veurey-Voroize.
Multiple architectures
Our partnership has developed and demonstrated two different types of architectures for SW/MW, MW/MW and MW/LW focal plane arrays (FPAs). The first is a conventional "npn" mesa structure that features two back-to-back photodiodes, with one connection per pixel. Commercial FPAs down to 24 µm pitches are available with 640 × 512 pixel arrays operating at a frame rate of 60 Hz, while larger format FPAs are under development.
This approach gives the detector excellent spatial coherence because every pixel detects both bands. However, because the read-out from these two bands is sequential rather than simultaneous, the design reduces temporal coherence.
Our other design addresses this weakness, with simultaneous read-out of the two bands. It is a semi-planar structure that features two n-on-p diodes, which are formed on two different levels of the three-layer architecture (see figure 1 for details). Although spatial coherence is compromised in this design, due to a spatial shift between the two bands equal to a quarter of the length of the pixel, this can be corrected by system-level signal processing.
We describe this structure as semi-planar because the first photodiode is planar and the other is quasi-planar. Using this type of structure is beneficial because it allows us to employ processes that are similar to the standard ones used in the silicon industry. This leads to higher yield, low manufacturing costs and a potentially more straightforward ramp-up to mass production.
Other advantages of the semi-planar structure over the more conventional architecture include a higher manufacturing yield for LWIR pixels, according to our experience. In addition, the two detection bands can be optimized independently and cross-talk between them is minimized. Incorporating avalanche photodiodes and extra-low noise-equivalent temperature-difference (NETD) pixels into this design is also a possibility, which promises increased detector sensitivity.
Our detector development efforts have focused on improving material quality (see box "Growing HgCdTe detectors by MBE"), optimizing the photodiodes and improving the flip-chip bonding processes to reduce pixel pitch.
We have made good progress over the last few years and the performance of each band in our dual-band detectors now offers a very similar level of performance to single-band equivalents.
In 2007 we demonstrated a 320 × 256 pixel detector called Janus at the SPIE Defense, Security and Sensing meeting in Orlando, FL. This dual-band detector operated in the MWIR and featured a 30 µm pitch. During that year we also produced a 256 × 256 pixel MWIR/LWIR dual-band detector (see figure 3 for images).
More recently, we have reduced the pixel pitch to 24 µm and produced a 640 × 480 pixel detector operating at 60 Hz. We aim to produce a dual-band 640 × 512 detector with 20 µm pitch this year. This scaling of the pixel dimensions aims to reduce the cost and weight of the FPA and its associated optics, while still allowing for high-yield manufacture. Progress has come through improvements in etching and deposition, advances in the process used to unite the detector to a CMOS read-out circuit and greater uniformity on large format arrays, which minimizes electrical performance dispersion.
One of our aims for the next few years is to open up new applications for our dual-band detectors by incorporating low NETD structures and avalanche photodiodes, which will improve performance at low infrared light levels. We also plan to address the important issue of cost-effective production.
View pdf of article
