A global team of researchers including a
Michigan Tech University physicist announced in late April they've succeeded in building a molecular computer that has the potential to solve complex problems in a way that resembles the workings of the human brain.
Physicist
Ranjit Pati at Michigan Technological University in Houghton is part of a team of researchers that has published its work under the title "Massively Parallel Computing on an Organic Molecule Layer," in the April 25 issue of the journal Nature Physics.
"This started five years back, and it's really the brainchild of my colleague
Anirban Bandyopadhyay from Japan," says Pati.
The problem they are trying to address is the limitations of standard digital computers when it comes to multi-tasking and self-organization. While standard processors are extremely fast at solving sequential problems, they don't deal as efficiently with more complex problems as a person can, Pati explains.
"This is important for complex problems, not simple problems, because digital computers can solve those better," he says.
Molecular computers have been approached from many angles before, and Pati says several research projects are also trying to address the same limitations.
"What is happening is a lot of people are working on this type of thing. Computers keep getting faster and faster, because we can put more transistors on a circuit, and keep making them smaller and smaller, but there is a limit," he says. "So people are trying to use molecules to increase the computing power."
Digital computers currently work at enormous speeds -- completing up to 10 trillion instructions in one second -- but must do them one after the other, Pati explains. In contrast, neurons in the human brain only fire about 1,000 times in a second, but they connect to and communicate with up to 10,000 neighboring neurons at the same time, making them much more efficient at doing multiple tasks at one time.
"Human neurons are much slower than digital computers, but our brain is much more intelligent, because the neurons work in connectivity with each other," Pati says.
To simulate this characteristic, the researchers used an organic-based molecule, made of nitrogen, oxygen, chlorine and carbon, called DDQ that assembles itself into two layers on the gold substrate. It can switch among four conducting states: 0, 1, 2 and 3, instead of the usual 0 and 1 binary states used in standard digital computing.
"For computing, you must be able to do three things: store information, process information, and communicating information. All of these things we have been able to do with this molecule," Pati says.
The team used a scanning tunneling microscope to switch the states of about 300 DDQ molecules in a grid, to test out its computational problem-solving abilities. A scanning tunneling microscope can use electrical charges to turn the conducting state of the molecule from one option to another, and also "reads" the molecular surface to tell researchers what is happening with the molecules. The change in state for one molecule goes on to affect the states of other molecules.
The scientists set up simple logic gates, and showed transfer of information from one molecule to several others, before trying more complex problems.
One of these was a representation of how heat diffuses, which digital computing has a hard time creating accurately. Pati explains that the molecular computer can solve this problem by using several switches at once to change the states of surrounding molecules, since heat diffusion doesn't originate from a single point the way a digital computer would attempt a solution. The team was able to simulate not only heat diffusion but also the development of normal cells into cancer cells, both processes that are usually strictly the domain of biological processors, according to the team's published findings.
Even more helpfully, the molecular computer is self-healing; if one molecule stops working, its functions are switched over to a new molecule. "This is something that happens in the brain; one neuron fails, and another takes over," Pati says.
But the applications for this molecular-level computer are still a long way off.
"This is a very slow and very costly process," he says. "This is a conceptual breakthrough, but are we going to see a computer using these in the next year? No."
Pati and his colleagues are now looking at other ways to improve on the molecular computing process, such as increasing the number of states available for switching, and connecting many more molecules together.
"We have started to try to build a bigger processor. If we can have 10,000 molecules, we can do much more with that," he says. "There are other possibilities too -- we are looking into more conducting states, perhaps six or even eight. We'll see what happens."
Kim Hoyum is a freelance writer based in Michigan's Upper Peninsula. Her credits include contributor to Geek Girl on the Street as well as a regular writer for Marquette Monthly. Hoyum is a graduate of Northern Michigan University where she obtained a Bachelor's Degree in writing.