Transistorless Computing - EMusician

Transistorless Computing

A lot of information about computational nanotechnology has appeared on my desktop in recent months, some of which could be relevant to electronic musicians in the future.
Author:
Publish date:
Image placeholder title

A lot of information about computational nanotechnology has appeared on my desktop in recent months, some of which could be relevant to electronic musicians in the future. As I've said many times, any advance in computer technology has the potential to benefit those who use computers and other digital devices to make music. For example, see "Tech Page: NanoRAM" in the July 2005 issue of EM and "Tech Page: Blast from the Past" in the January 2006 issue.

Image placeholder title

FIG. 1: In this artist's conception, criss-crossing nanowires are connected at each intersection by a single molecule that can be switched from one state to another, facilitating data storage and processing logic.

HP Labs (www.hpl.hp.com) is working on some interesting nanotech projects, including a technology called crossbar computing. The ultimate goal of this project is nothing less than the replacement of the transistor, the most basic element of computing for the past 50 years. During that time, the number of transistors on a silicon chip has risen from one to nearly one billion, but Moore's Law can't continue to be upheld forever with conventional semiconductors. It is generally believed that transistors within a silicon chip won't function once they are reduced in size below about 10 nanometers (nm; billionths of a meter), which corresponds to approximately 30 atoms. So alternatives must be found if we want more powerful processors.

Three scientists working in HP Labs' Quantum Science Research (QRS) group in Palo Alto, California, recently published a paper in the Journal of Applied Physics describing a transistor alternative called a crossbar latch. Nanowires measuring only 30 nm (100 atoms) in diameter are arranged in a crosshatch pattern, with one layer of parallel wires on top of and perpendicular to another layer of parallel wires, forming a rectilinear grid of intersections.

Sandwiched between the wires is a layer of material that is only a few atoms thick. This bistable material (meaning it has two stable states, allowing it to conduct more or less electricity) is electrically switchable, allowing its polarity at each intersection to be reversed independently with the appropriate application of voltages to the wires. As a result, the material at each intersection can store a data bit.

The crossbar-latch concept was patented in 2003, but the QRS team published the results of an actual demonstration in 2005. The demo consisted of a single latch, with one signal wire crossing two control wires and molecular-scale, electronically switchable "devices" at the intersections (see Fig. 1). The latch was able to perform the NOT operation and restore signal levels to their ideal voltages, which will allow many latches to be chained together to perform complex computational tasks. Previous experiments had demonstrated molecular-scale data storage as well as the AND and OR operations, but the addition of the NOT operation completes the basic operational palette for general-purpose computing at the nanoscale.

The simple crosshatch pattern makes manufacturing relatively straightforward, especially compared with conventional microelectronic devices, and it can be constructed using a wide variety of materials and processes, providing great flexibility. In addition, a single geometry can be used for memory, processing logic, and interconnection, making the crossbar concept highly adaptable. Manufacturing costs can be kept relatively low using chemical self-assembly, but that inevitably produces defects and irregularities in the size of nanoscale components. Another concern is the presence of random fluctuations in such small structures, especially at room temperature and above. Fortunately, the grid structure allows chip architects to easily design around those flaws using massive redundancy and other techniques. It will be years before crossbar latches find their way into mainstream devices, but they could form the basis of true molecular computing, allowing Moore's Law to survive another 50 years. This technology offers the potential for processors thousands of times more powerful than today's, which should be sweet harmony to any electronic musician's ears.