IBM researchers said they overcame stubborn challenges in building computer memory out of carbon instead of silicon, which could allow for systems that are faster and sturdier than conventional flash memory or RAM.
Carbon electronics research began at the end of the 20th century, but there were at least two nagging problems. The material must survive at the high temperatures needed in the circuit manufacturing process and in the data read/write process. Also, it must be pliable enough to quickly change states — from a zero to a one and back — yet stable enough to remain non-volatile when needed.
Workers from IBM Zurich found a way to ensure those qualities by adding oxygen into the mixture, they wrote in Nature last week.
The goal, IBM scientist and paper co-author Abu Sebastian told TechRepublic, is a commercial application of resistive memory. It works by measuring the electrical resistance across a circuit instead of using the conventional approach of simply having a circuit toggle on-off by applying voltage. Resistive methods can be exponentially faster and allow for many times more read/write cycles than modern flash or RAM, he explained. It also allows circuits to be physically smaller than conventional methods, which could lead to the continued progress of Moore’s Law, he noted.
Currently, memory companies are pushing up their product specifications with methods such as three-dimensional circuits instead of traditional two-dimensional versions, Sebastian said. “They have a lot of workarounds to boost the endurance. If you have resistive memory, then hopefully you don’t need those,” he observed.
IBM calls its invention oxygen-amorphous memory. “We just create a mesh of carbon atoms. It doesn’t have any shape like a diamond. That is why we call it amorphous,” Sebastian said.
However, Sebastian is realistic about the future. He predicted that flash memory and RAM will continue to dominate their markets in the present decade, while resistive memory could make inroads for high-end applications in the next few years. Now that the science is settling, engineers will have to determine how to manufacture the new kind of memory at large scales while keeping it affordable — no small challenge in itself.
Resistive memory is also being examined by other top companies such as Samsung, Toshiba, Hynix, and an Intel/Micron partnership, Envisioneering Group consultant Richard Doherty explained, in Seaford, New York.
What IBM didn’t reveal is exactly how they made the oxygen stick to carbon, nor how they’ll handle the increased amount of addressable memory locations that results from having smaller circuits. That information will probably remain secret until Big Blue publishes a patent application, Doherty said.
Doherty said he thinks Intel/Micron is the most notable rival because Hewlett-Packard’s similar research, called memristors, is going slowly.
Regardless of which company makes it, resistive memory will probably follow the usual pattern of how new technology trickles downward. It will first be used in military-specification products, then in critical commercial servers, and eventually in consumer products, Doherty continued. He said the cycle will take at least a decade. Conventional memory will still power your iPhone in the foreseeable future.