キコンボ アンドリュー キリンガ2009 年度 卒 ／博士（工学）
This thesis aims at establishing novel signal processing architectures for single-electron devices.
The present trend in improving the performance of silicon LSIs has been primarily as a result of the continuous scaling of CMOS devices. The need to scale transistors has led to improvement of fabrication technologies. With the advamced LSI fabrication technologies, research on fabrication of minute nano scale structures (devices) has attracted a lot of attention. Such devices include nano wires, quantum nano dots and single-electron devices.
Single-electron devices utilize quantum-mechanical effects to control transport of electrons at the single level. Thus single-electron devices inherently operate with minimum low power dissipation. Additionally, owing to the minute physical sizes of single-electron devices, they are considered as potential devices in implementing parallel-based information paradigms that would require high device densities.
Single-electron devices operate on different principles as compared to the conventional MOSFET devices. Therefore to employ them in signal processing systems, there in need to establish new circuit architecture frame works that fully utilize their properties. This research aims at exploiting both dynamical and structural properties of single-electron devices toward establishing LSI platforms for nano devices.
This research starts with investigating non-linear characteristics of coupled single-electron devices. Single-electron devices portray interesting non-linear dynamics: a single-electron device shows relaxation oscillations, while a double-oscillator system (two single-electron oscillators coupled through a capacitor) have attractors of oscillation that are independent of initial node voltage conditions. A quadruple oscillator system (two capacitively coupled double-oscillator systems) show multi periodic oscillations. Furthermore, by coupling single-electron devices, one can control the flow of tunneling events within the device network.
By combining the above non-linear dynamics with the structural properties, we proposed a two-dimensional photon position detecting circuit, and evaluated its performance.Secondly, by obtaining hints from neuronal systems, we proposed two bio-inspired LSI circuits: an edge detection circuit and a motion detector circuit. The thesis also discusses the implications of device fabrication mismatches and environmental noises in fabricating the two bio-inspired circuits. Instead of getting rid of such noises, we propose a novel method where such noises are actively utilized to improve the performance of LSI circuits.