Image Credit: Bilkent University UNAM
Memristors, resistors of which conductance is a function of the history of voltage applied to them, have attracted great attention in the present decade as potential components of memory and computing platforms. These long-forgotten device components were predicted over fifty years ago by Leon Chua, who described them as the fourth fundamental circuit element alongside resistors, capacitors and inductors – although the true origins of the memristor are even older than its name, as the term applies to such a broad range of electronic phenomena that the original observations of memristive behavior date over a century ago.
In the modern world, however, memristors are making a comeback: They are expected to play a major role in the development of novel computing platforms, particularly neuromorphic systems, where brain-like (cortical) computational schemes can be efficiently implemented using semiconductor technology. Recent reports on memristive switches demonstrate improvements in the uniformity and controllability of nanoionics-based memristors, though there is still a long way to go before memristors can compete with other circuit elements in forming neuromorphic systems with billions of synapses and millions of solid state neurons. On the other hand, Flash memory -another non-volatile memory technology- has reached very high densities and is readily compatible with CMOS technology, which makes it a particularly suitable model for the implementation of memristor-based applications.
In recent work at the National Nanotechnology Research Center at Bilkent University (Turkey), the Dâna group has demonstrated that junctionless flash memory cells can be operated like a memristor: The write and erase operations commonly performed through voltage pulses applied to the gate can be effected through the application of voltages through the source and drain terminals of the transistor. In fact, the equations that relate the transistor operation and charge/discharge of the floating gate show that the flash memory, when operated in this way, behaves nearly like an ideal memristor. The significance of the demonstration is that it connects the two non-volatile memory device families, the memristor and flash, and may facilitate future applications of flash memory devices in neuromorphic computing. Considering that the density of flash drives has improved to such an extent that multigigabit chips can be mass produced and have entered into virtually every cell-phone, flash technology already has the infrastructure that can enable the implemention of billions of synthetic synapses. Moreover, the flashristor mode -the new operation mode is referred to in the article- can be implemented with high uniformity and repeatability.
The group’s findings are also highlighted in the IEEE Spectrum website, which can be found at: