Printed electronics has recently moved from a focus on the production of individual components towards the design and initial commercialization of integrated systems. To create such systems, the use of non-volatile random access memory is often essential. This article describes the function and design of Thinfilm memories based on a reversible ferro-electric capacitor structure, and their application to novel products and markets.
Basic memory cell
A Thinfilm memory cell consists of a ferroelectric polymer sandwiched between two electrodes, denoted arbitrarily as a bit line (BL) and a word line (WL). On the application of a sufficient voltage, the dielectric dipoles within the polymer layer align, and because of hysteresis when the voltage is removed, the dipoles remain pinned in the state they had during the voltage pulse.
This is depicted in the curve in figure 1a. As the voltage can either be polled in a positive or negative direction, the dipoles can align in two separate directions. The ferro-electric capacitor cell is thus bistable and by applying a proper voltage pulse it can be ‘written’ into two different stable states. As is indicated in figure 1, the 0-state is defined to be on the top of the hysteresis curve, when the x-axis is the BL voltage minus the WL voltage, while the 1-state is at the bottom.
Figure 1 a) Hysteresis curve, b) ferro-electric connected to a charge integrator and c). 0-signals and 1-signals from 5500 R2R produced cells.
The electrode chosen as the BL is connected to a sense amplifier which detects the cell state. Using charge integrator circuitry (Fig.1b), the sense amp outputs a voltage proportional to the amount of charge released from the cell when the read voltage pulse is applied. As can be seen in the hysteresis curve, reading a cell in the 0-state gives a larger charge release compared to reading a 1-state cell. Figure 1c shows sampled charge integrator outputs, converted to polarization from 5500 memory cells. The red dots show the polarization (uC/cm2) measured when the cell is written into a 0-state before measurement, while blue dots correspond to readings from the same cells written with 1-state signals. As can be seen, there is a large difference in output signals, and use of a suitable threshold allows for identification of the data content of the memory cell.
The green curves show the resistive leakage of the cells, and contribute insignificantly to the polarization measured from the 0-state. That means that the signal-to-noise ratio is very good, and this holds for a large range of environmental conditions and is not affected by using printing as the manufacturing process.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.