06-07-2018 | | By Rob Coppinger
Memristors are simpler than transistors, smaller, use less energy, can alter their resistance and remember previous states which could lead to computers that use less power, never forget, have improved distributed memory and processing and can switch on and off far faster than today’s device.
The transistor is the foundation of all microchip technology and it has been miniaturised to nano-scale levels to fit billions on to a chip, but it is a technology that is reaching its limits. Memristors present a new opportunity for microelectronic design. They could pack more computational power in a smaller area, with less power consumption that will enable applications including Internet-of-Things (IoT) computing, smart medical implants, and radiation-resistant electronics, say researchers.
Under an £11 million, six-year project, that began in April, UK universities and companies plan to create memristor circuit design tools, build primitive memristor-based modules for future larger electronic systems, find ways for memristors to work with existing transistor-based microelectronics, and commercialise their findings. A key advance will be to find ways that memristor circuits can integrate with existing transistor microchip technology.
Memristor array manufactured at the Southampton Nanofabrication Centre
“How do we add the memristors on top of commercially manufactured CMOS as no one is going to build a new factory [for memristors only],” Dr Christos Papavassiliou explained, adding that a protocol is needed for, “how we add memristors on CMOS into circuits, so we can use some memristor functionality.” The complementary metal–oxide–semiconductor, or CMOS, technology controls the integrated transistor circuit. Papavassiliou is a senior lecturer at Imperial College London’s electrical and electronic engineering department. He is leading the analogue applications team, looking at possible uses such as potentiometers and reconfiguring antennae.
The project’s researchers expect to be able to distribute their core technology and design tools within three years to the research community leading to more application demonstrators. “We would be very disappointed if could not spin out anything [commercially] from this project,” Papavassiliou said. As well as IoT sensors, implants and radiation resistance, applications include, random number generators, radio frequency switches to change the behaviour of antennae, microwave filters, and fuzzy logic gates. Fuzzy logic is where values between zero and one are used to calculate outcomes, not just one and zero, which is the convention.
Microprocessors are not the only key components of a computer this project is advancing. These memristive microelectronics will use resistance random access memory, ReRAM. In traditional memory, data is stored as an electric charge, while in ReRAM a voltage is applied to a material stack that creates a change in the resistance and that records the data. ReRAM also has a lower read latency and a faster write performance than existing memory.
Papavassiliou and his colleagues have received £6.2 million in funding from the UK government’s Engineering and Physical Sciences Research Council. The project is led by the University of Southampton, its academic partners are Imperial College London and the University of Manchester. A further £5 million from the project partners brings the total funding to above £11 million. The companies involved include microchip designer ARM, Galvani Bioelectronics, National Microelectronics Institute, and NXP Semiconductors UK.