The next step to even smaller production processes would be quite simple thanks to silicon memory. Researchers have developed simple memory cells consisting only of silicon and connections.
The new type of memory became developed after observations of thin strips of graphite, about ten nanometers wide. Under the influence of a current pulse, the strip could be broken and reconnected, forming the equivalent of a memory bit. However, further investigation revealed that this effect does not depend on graphite; silicon also exhibits this phenomenon. Since silicon is the main component of almost all processors and can be incorporated well into semiconductor products, silicon memory could be quickly integrated into manufacturing processes.
The silicon memory works by sandwiching a layer of silicon oxide, an electrical insulator, between two layers of conductive polycrystalline silicon. When a current is applied, oxide atoms are stripped from the silicon, resulting in the formation of conductive silicon crystals. Once formed, these crystal trails, like the graphite before it, can be connected and broken again and again. The ‘poly’ layers serve as electrodes. Flash memory is more complex and requires three electrodes.
The memory bits formed in this way can be produced particularly easily and, given their composition, they can easily be processed in processors, for example. In addition, the structures are very small. The crystal wires are between five and ten nanometers in size, much smaller than the feature size that current memory production techniques are used to with their 20 to 30nm. The silicon oxide memory bits can also be stacked into three-dimensional structures and switched in less than 100ns. The Rice University researchers are collaborating with several companies and already have a working prototype memory chip with a thousand elements. This is now being tested.