Establishing Electrically Programmable Reaction Arrays as Universal Chemical Computers

本研究提出创建一种新型的混合电可编程化学阵列（EPCA）化学硅计算架构，可以解决数据密集型计算，如随机优化（如模拟退火）和复杂的计算问题。目标是为化学计算开辟一条新的路线图，目标是针对数学难题的通用计算。这项工作的愿景是产生一种硅-化学-混合处理器。可编程化学计算机的开发将通过构建硅-化学混合信息处理体系结构来实现，其中计算以化学方式进行，输入来自电极阵列，建立在储层计算领域的新兴工作基础上。该系统将由电极阵列编程来驱动电化学过程，输出将由离子敏感场效应晶体管（isfet）阵列读取。这种双电极传感器系统将用于编程和读取嵌入在电活性凝胶中的氧化还原活性化学振荡器。数字信息处理将在化学介质中进行，使用氧化还原可配置凝胶封装和导航一个巨大的可编程状态空间。这种导航将由一个化学时钟引导，它也将用于纠错。凝胶网络将有助于稳定电编程化学阵列。化学并行处理阵列将作为数字计算机和机器学习系统的基准，从冯·诺伊曼瓶颈中解放出来。

This research proposes to create new type of hybrid electrically programmable chemical array (EPCA) chemical-silicon computing architecture that can tackle data intensive computations such as stochastic optimizations, (e.g. simulated annealing), and complex computational problems. The objective is to start a new roadmap for chemical computing targeting universal computation aiming at hard mathematical problems. The vision of this work is to generate a hybrid silicon-chemical-processor. This development of a programmable chemical computer will be achieved by building a hybrid silicon-chemical information processing architecture, where the computations occur chemically, and the inputs come from an array of electrodes, building on nascent work in the area of reservoir computing. The system will be programmed by an electrode array to drive electrochemical processes, and the outputs will be read by arrays of ion sensitive field effect transistors (ISFETs). This dual-electrode-sensor system will be used to both program and read out from a redox active chemical oscillator, embedded in an electroactive gel. The digital information processing will occur in the chemical medium, using the redox-configurable gel to encapsulate and navigate a vast programmable state space. This navigation will be guided by a chemical clock, that will also be used for error correction. The gel network will help stabilize the electrically programmed chemical array. The chemical parallel processing array will be benchmarked as both a digital computer and a machine learning system, liberated from the von Neumann bottleneck.