From Basic Quantum Mechanics to State-of-the-Art Infrared Imaging
July 23, 2013
In this edition of the U.S. Army Research Laboratory Fellows Corner, ARL Fellow Dr. Kwong-Kit Choi discusses quantum well infrared photodetectors, or QWIPs, , and how ARL is working on improving QWIPs fundamentally by inventing new detector structures and working out new ways to understand and predict their performance.
In elementary quantum mechanics, the first example one usually encounters is the particle-in-a-box, a problem in which a particle is confined in a one-dimensional square potential well. The purpose of this example is mainly to introduce the principles of quantum mechanics in the simplest way. Its physical reality is not supposed to be taken seriously. It may then come as a surprise not only that such a system can be realized in the physical world, but it can even be an excellent infrared detector.
Infrared detection serves many important functions in military operations. Its many uses include night vision, surveillance, piloting, automatic navigation, targeting and tracking. It also finds plenty of civilian applications in industrial processes, environmental protection, homeland security, infrared astronomy, medical research, etc. However, there is a lack of suitable materials for long wavelength infrared detection. The incumbent detector material tends to be difficult, expensive, and in short supply. Alternatives are needed. About two decades ago, my colleagues and I at AT&T Bell Lab found a new way to detect infrared using more common materials. We discovered that if one stacks two types of material layers together in an alternate fashion, the difference in their bandgap energies creates a series of square potential wells in their combined conduction band. The material layer that has a smaller bandgap forms the well region, and the layer that has a larger bandgap forms the barrier region. Since the potential along the plane of each layer is still constant, the particle, which is an electron in this case, is free to move in this direction. Therefore, the quantum effects along the plane are negligible, and the system resembles the one-dimensional particle-in-a-box. When the thickness of the well is small enough, the electron energy states in the well region are quantized, just as the example describes. We usually refer to such a potential well as the quantum well.
In a quantum well material, each unit can be treated as one-dimensional atomic specie because its levels are discrete like atoms. Though different from the natural atoms, the energy levels of these artificial atoms are controllable, either by changing the thickness of the well or by changing the height of the barrier through different material compositions. Besides changing the well properties, one can also change the thickness of the barrier, so that the "atomic" wavefunctions in the adjacent wells can overlap in different extents. The degree of overlap determines whether the material stack behaves like isolated atoms, molecules or crystal lattices. By choosing the overall quantum well structure judiciously, one can build up a material with a predetermined energy level structure suitable for an intended application. For infrared detection, we adjust the quantum well parameters such that the energy difference between its first and second quantized states is equal to the incoming infrared photon energy. We also align the second level with the barrier height so that the electron can travel freely across the barrier and generates a photo-response after it absorbs a photon. A quantum well infrared photodetector, or QWIP, typically contains several tens of quantum well units.
The QWIPs are usually made of gallium arsenide, a material already widely used in electronic industries such as smartphones and high-speed communications equipment. With the manufacturing infrastructure already in place, the transition of QWIP technology from the lab to the marketplace went rather smoothly. Within a few years of its invention, relatively low-cost and high-resolution (0.33 megapixels) QWIP focal plane array (FPA) cameras were already widely available. Some of the companies offering QWIP cameras are Thales in France, IRnova in Sweden, Semi-Conducting Devices in Israel, and FLIR in the U.S. This detector technology has been deployed at a large scale by the military in Europe, where it is used for gun-sights in armored vehicles, anti-tank missile systems, forward observer systems, anti-oil pollution monitoring, maritime patrol and surveillance, etc. In the U.S., the technology is more directed to civilian applications such as border security and gas sensing. One of the higher visibility applications is NASA's Landsat Data Continuity Mission, the first application of QWIPs in space. The satellite carrying out this mission was launched in February 2013, and the Thermal Infrared Sensor (TIRS) instrument on board uses three QWIP FPAs to monitor the Earth's well-being in the coming years. ARL was a partner in this mission; we designed and qualified the QWIP material to meet the sensitivity and spectral coverage requirements. The images at the end of this article show the same geological location in the infrared and the visible taken by the Landsat satellite. More information on TIRS is available at http://www.nasa.gov/mission_pages/landsat/main/index.html.
While QWIPs are successful in civilian applications, to adopt the QWIP technology in the U.S. military, the detector still needs improvement. To understand why the detectors are not yet suitable for us, we trace back to the one-dimensional quantization. It turns out that in order to sense this quantization, the light has to travel sideways, in parallel to the material layers. This is readily the case only if the light is shining at the edge of the detector. When the light is shining on the detector surface, as in the usual detection, no light is detected. In the short term, the industry deals with this problem by placing reflection gratings on top of individual detectors. These gratings disperse the incoming light incident from the bottom surface into different angles. A portion is travelling at a large angle and is detected. However, this approach is not very efficient, and the typical quantum efficiency (QE) is only 5 percent. This level of QE cannot provide the military the sensitivity and speed it needs. Furthermore, the gratings are made of very fine periodic posts. They are difficult to produce in large formats to achieve high resolutions. Therefore, we need to devise a better light-coupling scheme for QWIPs.
At ARL, we work on improving QWIPs fundamentally by inventing new detector structures and working out new ways to understand and predict their performance. The goal of our research is to make QWIP FPAs more useful to the military through higher sensitivity, higher speed, larger format, multi-band, higher yield and lower cost. As the first step, several years ago we proposed the corrugated quantum well infrared photodetector (C-QWIP). In this detector structure, instead of relying on grating diffraction, 45 degree inclined detector sidewalls are engineered to reflect light into parallel propagation. Based on reflection, C-QWIP FPAs improve the technology both in performance and manufacturability. Because reflection is more effective than grating diffraction in redirecting the light, QE is larger. Reflection is also independent of wavelength. This means that the detector preserves the natural absorption spectrum of the material, which can be much wider than the grating coupling bandwidth. The same pixel geometry and production process are now equally applicable to all QWIP materials with different detection wavelengths. This allows the simultaneous production of FPAs having a wide range of wavelength bands without jeopardizing QE. In the absence of the fine grating features, it is also feasible to use standard photolithographic techniques to achieve fast, inexpensive and very large format production.
We have since produced a large number of C-QWIP FPAs with resolution from 0.33–4 megapixels. The measured QE was from 10–36 percent, depending on the number of quantum well units used in the detector. Since the larger number of units needs higher bias voltage to operate, the higher QE detectors are less compatible with the current off-the-shelf readout integrated circuits (ROICs). Therefore, the current C-QWIP FPAs were produced with their designed QE substantially less than the 50 percent theoretical limit. Nevertheless, together with L3-Cincinnati Electronics, we have made a number of technology demonstrations with excellent results. These demonstrations include large-area-surveillance, unmanned aerial vehicle detection and tracking, and missile launch and intercept observations, etc. Our present goal is focused on making C-QWIP pixels smaller to increase the resolution of an FPA. To complement this effort, ARL has launched an SBIR program to develop more capable, higher bias ROICs to improve the C-QWIP performance. Besides military applications, ARL also worked with NASA to produce C-QWIP FPAs and used them in applications such as cave hunting in desert environments for Mars mission studies and in environmental studies in Southeast Asia. Recently, under a NASA program, we demonstrated a 4-MP broadband 8–13 μm C-QWIP FPA for a future Landsat mission upgrade. With their spectral versatility and structural simplicity, C-QWIPs will continue to be desirable and competitive in long wavelength infrared detection.
Although C-QWIPs have largely overcome the detection requirement of QWIPs, ARL still strives to further improve the technology. It would be beneficial, in terms of detection speed and ROIC compatibility, to increase QE further while keeping the number of quantum well units small. To this end, one needs to re-examine the cause of the inefficiency of gratings. It turns out that grating diffraction behaves very differently in open space and confined space. Inside a detector, the light can take many different routes to reach a location. It can either take a direct path from the grating to that location or an indirect path via a reflection from one of the detector sidewalls or the bottom surface. This multi-path interference generally reduces the diffraction efficiency because not all the light rays along the diffraction angles are in phase. The grating QE is thus suppressed. To know the grating efficiency, we need to account for all interferences in an enclosed space. In the case of C-QWIPs, their special 45 degree inclined sidewalls happen to make interference effects negligible. In these detectors, the normal incident light is first reflected by a sidewall and makes a parallel pass. Upon reaching the opposite sidewall, the unabsorbed light makes another right angle turn and exits out of the detector perpendicularly. It does not stay inside the detector to interfere with the later incident light. The multi-path interference can thus be avoided in this detector structure. To increase the QE beyond that of a C-QWIP, one has to reverse the role of multi-path interference and make use of it to his advantage.
To yield a strong parallel propagating light, one needs two components: an effective diffractive element (DE) on the top surface and a properly designed detector volume. A proper detector volume allows the unabsorbed light, under certain resonant conditions, to circulate constructively inside the detector and allow the newly diffracted light to reinforce this circulation. As more light enters into the detector, the internal intensity grows until it is balanced out by the stronger absorption induced at the higher intensity level. Therefore, if a QWIP can be made into a resonator, its QE can be significantly larger. To achieve resonance, one needs to design the pixel size and the DE precisely.
To design a resonator-QWIP or R-QWIP, we need to calculate the electromagnetic (EM) fields accurately in complex detector geometry, a feat that has never been demonstrated by the infrared community. After much research, ARL finally succeeded in establishing a finite-element EM model that can apply to any arbitrary three-dimensional detector geometry. The model is able to predict a definitive QE without any empirical parameters. After developing this model, we were able to explain all the yet-to-explain experimental data accumulated in open literature over the course of QWIP research and development. For the ten distinct structures we have examined, including gratings and C-QWIPs, the experimental data were all in close agreement with the EM model. Furthermore, the model is consistent with the classical theory based on ray-tracing technique and agrees with the simpler modal transmission-line model in two-dimensions. Therefore, we have firmly established the validity of our model, and we can use it to design different R-QWIPs. Their predicted QE can be as high as 80 percent, depending on the DE, the detector volume and the intrinsic absorption coefficient. The preliminary experimental results on four less-optimized designs show close agreement with the predictions, with the highest value being 71 percent. Since all these designs are based on a small number of quantum well units, we may have achieved our goal of having high QE and a small number of units at the same time. We are currently producing imaging FPAs to confirm these lab results.
The use of resonance to increase the detector QE should not be confined to QWIPs. Even if the material's absorption is isotropic, the strong light-intensity present in these resonant structures will increase their QE as well. The EM modeling performed on these types of materials, such as Mercury Cadmium Telluride and type-II superlattices, indeed shows the expected benefits. Therefore, our resonant detector approach is generally applicable to other infrared technologies. At the same time, the EM model also works in other parts of electromagnetic spectrum, as expected. For example, we applied it to a multi-layer solar cell and were able to quantitatively explain the observed photocurrent spectrum. Our next step is to design practical structures to enhance the cell efficiency.
As for QWIPs, our new optical design capability enables us to assess the performance and characteristics of a wide range of light-coupling structures without actually making the detectors. This has greatly accelerated the detector development. Combining it with the highly versatile quantum well materials, we have thus created a highly agile infrared technology, in which we can match the detector characteristics with any applications to a high-degree of precision. For example, we can design, under a proper material and light-coupling combination, a detector whose detection spectrum overlaps precisely with the transmission characteristics of a dust cloud. Using such a detector, one will be more able to see through the dust clouds stirred up by a helicopter when it is operated near the desert floor; or it can match the absorption spectrum of a medium for chemical sensing in battlefields and in space. The detector can also be designed to resonate at two wavelengths for dual-band detection; or it can be designed to enhance one optical polarization while suppressing the other for polarization-contrast detection. In the future, we will continue to employ EM modeling to study other frontier optical effects on QWIPs, such as photonic crystals, meta-materials, plasmonic materials and nano-antennas. It will allow us to discover more worthy detector structures and expand their performance and functionality. At the same time, we will also use the new detector characteristics to better understand the underlying optical effects and hence advance the scientific frontier. With the present and future innovations, the prospect of QWIP technology is bright, and the one-dimensional particle-in-a-box will no doubt become more real and practical with each new application serving humanity.