Introducing the first ReRAM-Forum movie!! In part 2 of their recently published papers in the TED of the IEEE, Professor Ielmini’s group describe the modeling of resistive switching in bipolar metal oxide ReRAM. Like part 1, the paper is a collaboration with David Gilmer of Sematech who provided the Hafnium Oxide based ReRAM samples. Key components included in the model are ion migration through drift and diffusion with Arrhenius activated diffusivity and mobility along with local temperature and electric field being derived self-consistently. The numerical model solves the drift/diffusion equations for ion migration allowing the evolution of the conductive filament (see figure) to be viewed in ‘real time’. The equations in are solved in Comsol, which is largely accessible to all industry/university labs. The model reproduces the abrupt set and gradual reset transitions along with the kinetics of cell behaviour observed experimentally.
The movie shows the I/V characteristic (left hand side) and the doping density (right hand side) during a reset sweep of a device set to a resistance of ~400ohms. (The movie is posted on Youtube). Otherwise, it is hopefully self-explanatory. Many thanks to Professor Ielmini and Stefano Larentis (who is now at UT Austin) for providing the movie!
The paper goes on to describe various types of measurements where the model and experimental results are compared. As with Professor Wong’s recent modeling paper, reproduction of key characteristics of an ReRAM cell are impressive and you will have to seek out the paper for further details. However, one aspect that I was particularly struck by, was replication of the kinetic behavior of the reset process. This has been done in two ways. First I-V characteristics are compared under the application of a triangular voltage waveform with sweep rates varying by two orders of magnitude. Second the device resistance is shown as a function of total biasing time during a sequence of rectangular reset pulses. In both cases there is clear agreement between model and experiment. The faster sweep rate resulting in a larger reset voltage. Similarly the larger voltage reset pulses require a shorter total bias time. However, in the latter (second) case, the shorter time is particularly marked. For example, reading from the figures, increasing the amplitude of the reset pulses from 0.7 to 1.2 V results in a ~300x decrease in the time required for the resistance of the cell to increase by a given amount! (Another way of putting this would be to say that the device resistance is not just related to the time integral of the reset voltage alone.) The authors use these measurements to project that a read disturb (at 0.1V) is unlikely for total read times less than 10^6 seconds.
While, I should add my usual caveat that just because a model replicates an experimental series of observations, one should never be 100% certain that the model is complete (as happened in a project in which I was involved in an earlier life), the authors have clearly exercised and validated their model extensively. As such their conclusions extrapolated to timescales inaccessible in the Research Lab seem justified. And of course there is the Blog’s first movie!
Christie Marrian, www.ReRAM-Forum.com Moderator
IEEE Transactions On Electron Devices, Vol. 59, No. 9, September 2012 p2468, Resistive Switching by Voltage-Driven Ion Migration in Bipolar RRAM—Part II: Modeling, Stefano Larentis, Federico Nardi, Simone Balatti, David C. Gilmer and Daniele Ielmini