For Phase I, we propose to investigate the scientific, technical, and commercial merit and feasibility of IR sensors based on nanostructures that are compatible with standard silicon IC technology. In Phases II and III we will integrate these sensors with future generations of cellular neural network (CNN) chips. Specifically, we propose to investigate and develop nanoantennas composed of lithographically-defined metal-line dipoles separated by metal-oxide-metal rectifiers that will be integrated into each pixel of a CNN processor array. The operation of such nanoantennas as IR detectors has been reported in the literature, and they appear to be promising candidates since they offer CMOS compatibility, good prospects for achieving multispectral sensing, small size, and high speed. Conventional high-resolution imaging array sensor technology does not readily allow detection of multiple IR wavelengths at each pixel, and the resulting separation of sensing and computing functions creates a bottleneck for image processing throughput. Our proposed approach has the potential to eliminate this bottleneck by developing CMOS-compatible, multispectral nanoantennas as IR sensors that will enable systems to readily combine sensing and computing functions. Our goal is to investigate the feasibility of CNN chips with integrated MWIR- and LWIR-band nanoantenna sensors operating at 10,000 frames per second
