FERNANDO RODRIGUEZ-MORALES, SIGFRID YNGVESSON, AND DAZHEN GU
Much attention has been given to imaging applications at electromagnetic frequencies between 0.3 and 3 THz, mainly because of the unique features of terahertz radiation (T-rays). The noninvasive and nonionizing nature of T-rays makes them well suited for applications in life sciences and medicine, as well as for security screening and surveillance. Furthermore, the components and devices required for terahertz systems are significantly smaller than those used at microwave or millimeter wavelengths, which is advantageous for miniaturization and system integration. These characteristics make further development of terahertz sensors attractive and worthwhile.
Terahertz detectors can operate in the time domain (using broadband pulses) or in the frequency domain (using or detecting what are essentially single-frequency sources). Frequency-domain detectors can be either heterodyne (coherent) or direct (incoherent). The main advantage of direct-detector systems is that they are less complex than their coherent counterparts. Heterodyne receivers (also called mixers or downconverters), however, provide much higher sensitivity than direct detectors, as well as superior spectral resolution.
One of the most attractive applications of terahertz sensors is the implementation of multielement focal-plane arrays (FPAs). An FPA is a two-dimensional array of detectors placed at the focal plane of a reflector (or lens) and used to collect radiation from the target. In contrast to the large arrays available for detection in the visible and infrared (for example, digital cameras with millions of pixels), the majority of the instrumentation available for terahertz heterodyne detection is still based on single-pixel receivers. The implementation of compact terahertz FPAs would improve the detection speed by a factor equal to the number of pixels in the array. Although terahertz heterodyne FPAs cannot compete with the high spatial resolution and low cost offered by visible or IR systems, they could provide other benefits derived from their high frequency selectivity.
Among the technologies used for terahertz downconversion, hot-electron-bolometer (HEB) mixers have attracted the attention of many R&D groups around the world. Their low local-oscillator (LO) power consumption (less than 1 µW), near-quantum-limited noise performance, and ease of fabrication have placed them above competing technologies in the quest for implementation of large-format heterodyne arrays.
Bolometer-mixer arrays for submillimeter wavelengths were first envisioned by researchers at Caltech (Pasadena, CA) in 1979, who presented the concept of an indium antimonide bolometer array with 100 elements.1 For more than three decades, however, the realization of such an ambitious size has remained unreachable, mainly because of the lack of suitable technologies.
FPA design considerations
The constraints for the smallest achievable spacing in a multibeam imaging array have been systematically studied in terms of the geometric separation (Δx) of contiguous elements. To obtain a well-sampled image of the field of view, it is desirable to place the receiving elements at a spacing given by the ideal Nyquist criterion, which is analogous to the rule for sampling band-limited signals in the time domain. It has been shown, however, that it is impossible to construct arrays with the theoretical interelement distance as given by the Nyquist rule without degrading the coupling efficiency.
For lens/antenna arrays, the minimum value of Δx is a function of the coupling scheme used: single lens or multiple lenses. In the first approach, the individual antenna elements are placed near the focus of a single lens. For frequencies between 0.3 and 3 THz, a single-lens array results in typical array periods on the order of microns, which imposes impractical spacing constraints for IF (intermediate frequency) amplifiers or any additional circuitry.
The multiple-lens approach consists of placing multiple integrated lens/antenna cells at the focal plane of the system-it is also known as the “fly’s eye” approach. For the frequencies mentioned above, the lenses can be separated by a distance equal to their diameter, which will produce nearly diffraction-limited angular resolution. A major benefit of the fly’s eye configuration is that it allows ample space for IF amplifiers, transmission lines, and biasing connections. In addition, the required optics for RF (radio-frequency) and LO injection are essentially unchanged when more elements are incorporated. (In the heterodyne-detection scheme, the injected LO input combines with the incoming RF signal, creating an IF signal that is low enough in frequency that electronics can amplify it.)
Recent development efforts
Our group at the University of Massachusetts Amherst recently developed advanced packaging schemes that can make large superconducting HEB-mixer arrays a feasible architecture.2 In a parallel pioneering effort, we have evaluated the suitability of this technology for implementation of single-element scanning-imaging systems.
In a close-fitting array, it is essential that the HEB mixer be closely and directly integrated with its corresponding IF amplifier. With this in mind, we have used two-dimensional (2-D) and three-dimensional (3-D) packaging schemes to achieve the required integration by means of multichip modules. We have progressively decreased the receiver volume by a factor of 20, accompanied by a mass reduction of 15:1 compared to the traditional receiver arrangement in which the IF low-noise amplifier (LNA) and the mixer are placed in two independent modules (see Fig. 1).
In addition, we have developed the first heterodyne FPA ever reported to operate at a frequency above 1 THz. Our three-element assembly uses HEB mixers coupled to microwave-monolithic-integrated-circuit (MMIC) amplifiers (see Fig. 2). The HEB/MMIC detector elements are configured in a fly’s eye array. We have experimental evidence that the optimum array period is indeed close to the diameter of the lens in the integrated antenna, setting up an important precedent for future developments in this direction.
In a second line of research, we have constructed a terahertz passive-imaging system based on a single-element HEB/LNA receiver. Each component in the system is self-functional and interchangeable to accommodate operation at different frequencies. We use electromechanical actuators to accomplish the scanning of the target in one dimension. A thermal sensitivity of 1.5 K was demonstrated and we expect that an optimized HEB detector will provide sensitivities of less than 16 mK, which is better than the best reported imaging apparatus based on direct detectors (equivalent to 125 mK). This imaging prototype was very basic and of limited capabilities, but it served as a successful proof of concept. Other researchers at different institutions have extended this system into a more compact 2-D raster scanner with an-order-of-magnitude sensitivity improvement.
What is next?
We have also learned that HEB mixers are not devoid of shortcomings. These superconducting detectors are indeed extremely sensitive, but they suffer from IF instabilities not present in other technologies such as Schottky-diode mixers. The lack of stability must be a point of consideration, but its stability is more than adequate when fast scanning times (milliseconds) are used. Also, because they will always require cryogenic cooling, their potential for portable systems is limited. Despite these limitations, we believe that a compact FPA with a moderate number of HEB elements is now within reach. Other groups in Europe and the U.S. are exploring alternatives such as membrane-based and “beamlead” HEBs (incorporating so-called beamlead diodes) to construct larger arrays operating at frequencies exceeding 4 THz. Applications such as astronomical observations and remote sensing will greatly benefit from these efforts.
1. A.R. Gillespie and T.G. Phillips, Astronomy and Astrophysics 73(1-2) (March 1979).
2. F. Rodriguez-Morales et al., IEEE Trans. Microwave Theory and Techniques54,(6) Part 1 (June 2006).
FERNANDO RODRIGUEZ-MORALES is a senior RF engineer at the Center for Remote Sensing of Ice Sheets at the University of Kansas, Lawrence, KS 66045; e-mail: [email protected]. SIGFRID YNGVESSON is an emeritus professor at the University of Massachusetts (U. Mass.), Amherst, MA 01003. DAZHEN GU is a research fellow at the National Institute of Standards and Technology, 325 Broadway St., Boulder, CO 80305. Rodriguez-Morales and Gu were research associates at U. Mass. when this research was done.