I try not to put too much weight on very early advances in technology. This is particularly true of computer technology because there are so many relatively new avenues of research, all clamoring for attention: Quantum computing, DNA computing, optical computing…etc. On the other hand, computing has become so vital, especially for science and business, that it’s important to keep a hawk’s eye view of the entries into the field. So here’s another – phase-change computing.

The basis of this approach, phase-change materials (PCM) is not new at all. Solid, liquid and gas are the phases of most materials, for example, H2O – ice, liquid water, water vapor. As the materials pass from one phase to another (solid-solid, solid-liquid, solid-gas, liquid-gas) they give up or store heat. Some materials exchange more heat energy than others and those are the ones identified as phase-change materials. For example, salt hydrates, fatty acids and various paraffins are PCMs and have been used to store heat since the late 1800’s.

It’s the property of dramatically changing energy level that interests computer scientists. A phase-change material that starts at one low energy state (0) and after an electrical charge has a detectably high energy state (1) can be the basis for a memory storage device, or a calculation register. These are the basic components of a digital computer, and that’s why PCM materials are on the way to use in commercial memory devices. What is relatively ‘new’ is the attempt to use PCM materials that in some rudimentary way emulate neurons (brain cells).

Specifically, a paper from a team at the University of Exeter (Cornwall, UK) published in Advanced Materials [22 June 2011, paywalled, Arithmetic and Biologically-Inspired Computing Using Phase-Change Materials] gives credit to being inspired by biological systems (read: neural synapses) in their search to find an electronic equivalent. There are two key points to their approach: One is that the device they built can do both memory storage and digital processing simultaneously. Second, they are using materials that react photonically, that is, their light reflecting properties change.

The device they built as proof of concept is called a memflector because it retains a direction of optical reflectance (the direction in which the light refracts). The memflector is often compared to another recent development, the memristor, which retains a level of electrical resistance depending on the direction of current flow. The Exeter memflector uses a special phase change material known as a chalcogenide (one of 16 elements in the periodic table) in this case an alloy of germanium, antimony and tellurium.

The memflector is where the ‘synaptic-like’ behavior occurs. Somewhat like a synapse it is responsive to aggregating electrical impulses, an ‘integrate and fire’ response pattern. This also sets the optical reflectance of the memflector. The memflector package, according to the researchers, constitutes a ‘hardware neuron.’ In their own words:

Lead author Professor David Wright of the University of Exeter said: “Our findings have major implications for the development of entirely new forms of computing, including ‘brain-like’ computers. We have uncovered a technique for potentially developing new forms of ‘brain-like’ computer systems that could learn, adapt and change over time. This is something that researchers have been striving for over many years.”

[Source: Exeter University]

This does sound like PR gush. The skeptic in me wants to say, “could be.” In any case, the work here is rudimentary, a proof of concept, and a very long way from demonstrating ‘brain-like’ capability, much less a ‘new form of computing.’ That doesn’t mean a phase-change computer won’t happen. What’s important now is how many others pick up the line of research. I don’t doubt the approach is in some scientific respects ‘reasonable’ but other scientists, particularly ones with a commercial eye, will quickly evaluate this approach for problems in manufacturing and scaling.

Whether this particular approach works or not, I think the take-away on it is that many scientists are not satisfied with the mainstream notion that digital computing as we now know it is sufficient for duplicating or emulating human brain function. Many have said that throwing yottaflops (that’s 10 to the 24th floating point operations per second) of processing speed at ‘brain-like’ calculations won’t add up to ‘brain-like’ functionality. Some other kind of computer, perhaps the phase-change computer, will be necessary.