Defect-Tolerant Computing Based on an Asynchronous Cellular Automaton Teijiro Isokawa, Fukutaro Abo, Ferdinand Peper, Naotake Kamiura, and Nobuyuki Matsui Abstract Research towards the development of logic devices and simple circuits on molec- ular scales has attracted much efforts in recent years. Though opinions differ widely as to how actual computers can be built by such technology, it is agreed upon that computers with a regular structure of simple elements are very suit- able with respect to the ease of manufacturing. It is also recognized that what- ever computer architecture is employed, methods need to be found to cope with the less than 100 percent reliability of nanometer-scale manufacturing processes, as a result, of which faulty molecular structures should be expected. It is im- practicable to discard defective manufactures as in the current semiconductor technology, so a different strategy is required, one in which only those parts which are working are used and the defective parts are ignored. For nanocom- puters with a CA-based architecture this means that all work needs to be divided over the non-defective cells, and that a strategy needs to be found to identify the defective cells and to circumvent them without this giving rise to a drastic decrease of efficiency. In this paper, we focus on the problem of coping with defects by the use of a novel CA model, called Defect Tolerant Asynchronous Cellular Automata (DTACA). We assume that defective cells in DTACAs are unable to update their state by transition rules. The method we advance consists of two stages. First, defective cells are marked by a special state of the cells surrounding them, effectively isolating the defective cells. This is realized by a specifically designed set of transition rules, according to which the non-defective cells behave. Next, signals that move around on the cellular array are redirected when they run into an area of marked cells. In other words, each signal moves along a marked area and after circumventing it leaves it and continues in the same direction it moved into before it was redirected. An additional set of transition rules is required to make the cells behave in this way. By simulations we show that our proposed DTACA can work around the defects thus achive reliable computation.