17 Aug 2007
For decades infrared detector development has drawn on many different types of material. Philippe Tribolet explains why Sofradir has selected HgCdTe for its future plans. These include building a new $12 m fab to cut chip manufacturing costs, which should ultimately lower detector prices and drive up sales of these high-specification imagers.
Infrared imaging systems have several advantages over their optical equivalents. They can operate at night, because image contrast is solely provided by temperature differences, and they can work in poor weather conditions by detecting radiation in particular wavelength bands. Image quality is also less affected by reflections from sunlight and emission from bright light sources, such as car headlights, particularly when you select a detection band in the 8–12 µm range.
As a result, these semiconductor-based imagers are suitable for a wide variety of everyday uses. However, cost has limited their deployment to military and space applications, where they are predominantly used for surveillance and targeting of hot objects, such as rockets and vehicle motors.
The military's use of these detectors began in the 1970s with first-generation cameras featuring cooled, small linear arrays of photodetectors coupled to a complex two-dimensional (2D) scanning system. By the 1980s, second-generation versions, which included a read-out integrated circuit for signal pre-processing and multiplexing on the focal plane, had been built in research labs. These were then commercialized in the early 1990s.
The progression to second-generation detectors opened the way for high-resolution scanned arrays featuring time delay and integration, and high-resolution 2D arrays with signal processing on the focal plane. Cooled "staring" 2D arrays operating in the medium- and long-wavelength infrared (IR) bands were also developed, which have been dubbed "2.5-generation" IR detectors. These imagers, which were launched commercially in the late 1990s, have arrays with typically 320 × 256 or 640 × 512 pixels, and are suitable for applications such as missile detection. The improved resolution and performance of all these second-generation detectors have doubled the effective range for target acquisition, and enabled identification of enemy vehicles at night through smoke and dust at distances of 6 km and more.
These various generations of IR imagers have been fabricated with a variety of technologies, and include detector arrays made from HgCdTe, InSb, InGaAs and GaAs quantum wells (see figure 1). Each of these detectors has particular merits, and selection is application specific.
The performance of these detectors can be evaluated by their ability to distinguish a small target from an IR background with a similar temperature. Military, security and surveillance applications use this as a test and rate the imagers' performance in terms of detection or identification range, and its ability to deal with adverse weather.
The highest class of detectors can image objects at distances of 10 km or more, depending on atmospheric conditions and altitude, and are suitable for many different applications. These include missile guidance and tracking, missile warning, and carrying out reconnaissance and climate observations. Comparable detectors with an operating range of 6–10 km can also be used for scientific applications, including analysis of relatively weak signals for gas spectroscopy.
The detectors that operate over these ranges require cooling and are predominantly based on HgCdTe and InSb chips. HgCdTe has several advantages, including a sensitivity over a spectral range stretching from visible wavelengths to 18 µm, a high quantum efficiency coupled to a high signal-to-noise ratio, and a relatively high operating temperature that reduces the cost of the associated cooler. Liquid phase epitaxy (LPE) has been used for many years to produce these chips, but single- and dual-band chips covering short-wave (1–3 µm) and medium-wave (3–5 µm) IR bands can now be produced in large volumes and at low costs using molecular beam epitaxy (MBE).
By comparison, InSb is only sensitive in the medium-wave IR band. Technological limitations also hinder improvements in operating temperature and very small pixel sizes, making this material unsuitable for the most demanding applications. Alternative technologies include quantum-well IR photodetectors (QWIPs), which can be used for long-range detection (8–12 µm) but have slow frame rates and require lower operating temperatures.
Another class of detectors are those with a 2–6 km detection range. These can serve civilian needs such as fire surveillance and security applications such as police surveillance and tracking. Cooling the detector is mandatory in all of these applications. In some cases the selection of a very high-performance detector can actually cut the overall system cost, as this can reduce the size of the optics, simplify the signal processing and ease reliability constraints.
The candidates for providing IR detection over these distances are similar to those for the longer ranges. However, QWIPs operating in the IR long-wavelength band can offer good value for money if their limited efficiency and relatively high dark current can be tolerated. InGaAs cameras operating at shorter wavelengths are also competitive, but they can suffer from read-out circuit noise when the input signal is low, and they cannot detect beyond 1.9 µm.
The final class of detectors are those used for imaging objects at distances between tens of metres and 2 km. They can also serve civilian applications, such as building inspection, industrial process control and automotive driving enhancement. Other uses will emerge as the cost of these detectors fall. Uncooled thermal detectors, such as microbolometers, which include imagers based on amorphous silicon that is fully compatible with CMOS silicon technologies, offer the best value for money and are best suited to these tasks.
The French legacy
Many of these different types of chip-based detectors have been commercialized in France. Currently, over 500 people are employed for the research, advancement and manufacture of these devices in the Grenoble area, making this region a major global player.
The research has been led by the national research and development centre CEA-Leti, which started developing IR detectors in the 1970s. This knowledge has been transferred to Sofradir, which is headquartered in Chatenay-Malabry and has production facilities in Veurey-Voroize. The research from CEA-Leti has been used in our production of second- and 2.5-generation detectors, which we have been manufacturing for over 13 years, and the amorphous silicon microbolometers that we have been building since 2003 through our subsidiary, ULIS. QWIPs developed at Thales Research and Technologies, France, are also being mass-produced in co-operation with us.
Today, we manufacture thousands of cooled IR detectors, alongside tens of thousands of uncooled detectors through ULIS. In addition, we are currently building a new €9 m MBE fab for HgCdTe chip production, which will scale up our production from 2 to 4 inch material. When installation is completed in summer 2008, the increases in manufacturing volumes and efficiencies will reduce our production costs and also enable us to make larger chips. In addition, the lower costs should have an effect on the price of the longer-range detectors that will employ our chips and should drive an increase in sales of this type of imager. One-third of the cost of an IR detector is associated with the focal-plane array (FPA), so lower chip manufacturing costs can have a big impact on the price of the overall system.
Our new fab will be used to mass-produce third-generation detectors that have been developed in-house. These devices provide multicolour operation, have large FPAs with homogenous active layers, can resolve more image detail and operate more effectively in poor weather.
The chips will be produced by MBE, a technique capable of growing the multiple layers required by multicolour detectors on a wide variety of large-diameter substrates. This replaces our current growth technique used for production, LPE, which is carried out on lattice-matched CdZnTe substrates. These substrates are only available in small sizes, so many fabs are now developing silicon or GaAs platforms that offer low cost and a compatibility with the thermo-mechanical characteristics of the read-out circuit. In France, however, germanium has been the preferred alternative, because the lower stability of its oxides makes its surface easier to prepare, both ex situ and in situ, prior to MBE growth.
The MBE process on germanium (211) was demonstrated several years ago at CEA-Leti, and is suitable for making larger arrays. This substrate is already available in 4 inch and larger versions, which can be used to make the high-quality HgCdTe films needed for 2D short-wavelength and medium-wavelength IR arrays. We plan to start mass-producing both of these types of arrays during summer 2008, using two MBE reactors with 1 × 4 and 3 × 4 inch capacities. The switch to a larger wafer size will give us an advantage over InSb and QWIP manufacturers, who are still using 3 inch substrates for mass production (see figure 2). In addition, it will allow us to affordably fabricate 1280 × 1024 pixel IR FPAs with a 15 µm pixel pitch that will provide a greater image identification capability than detectors with fewer pixels.
We are also continuing to improve our third-generation detectors, so that manufacturing costs can be cut and selling prices reduced. In particular, we are focusing on the development of smaller pixel pitches and larger formats, and improving the performance of our avalanche photodiodes (APDs) and multicolour detectors. Our HgCdTe APDs are a unique design based on electron-impact ionization, and deliver exceptional detection. These electron-initiated APDs deliver a thousandfold multiplication gain at an inverse bias of only 10 V. The high gain at low bias, combined with a low noise factor, makes these APDs particularly well suited for integration in the latest FPAs. These are ideal for active laser-based imaging, but they can also serve many passive imaging applications.
• This article originally appeared in the July/August 2007 issue of Optics & Laser Europe magazine.