Computing with Liquid-State Nanomechanical Oscillator Networks

点击翻译按钮获取中文摘要

Our research program explores the use of nanoscale mechanical elements to replace electronic transistors as the elemental building block of advanced computing devices. It is well-known that it is increasingly difficult and expensive to reduce the dimensions of electronic transistors to create better computing applications, due to fundamental physical limitations in their fabrication process. Instead, we look for inspiration in extraordinarily powerful and energy-efficient computing systems such as the human brain, to propose a computing machine based on mechanical units interconnected in a complex network. Mechanical units are chosen because they can be fabricated at small dimensions with unique capabilities (high speed, low energy consumption, non-linear behavior). These units are organized in a network, like neurons in the brain, because this type of organization can lead to much more interesting computing devices with characteristics typical of living organisms (massive parallelism, robustness, adaptability). We suggest, in particular, the novel use of very small oscillating droplets of liquid metal (diameters down to 100 nm) as the mechanical units for our demonstrators. These nano-droplets are reasonably simple to fabricate and, being liquid, have very interesting properties for our purpose. Thousands of droplets will be assembled in networks to study the emergence of computing capabilities. It is expected that these complex mechanisms will perform far better than standard electronic computers on difficult and important problems, such as the detection of patterns in sounds or in images.      Our multi-disciplinary research is important scientifically because it introduces a new tool, liquid metal nano-droplets, to explore the motion of fluids at very small scales; because it creates what is probably the most sophisticated computing machine ever fabricated from mechanical elements, with dimensions and power-efficiency better than those of microelectronic circuits; and because it deepens our understanding of the emergence of computing capabilities in complex systems, a theme of central importance in computer science and neuroscience. Our research is also important technologically, as it may lead to miniature, energy-efficient and robust computing devices which can be reconfigured on the fly to rapidly solve challenging problems. These devices would be especially useful in applications with limited energy resources, such as mobile phones or distributed sensor networks.