«

»

Jan 15

The View from HP: Part 2

Hybrid CMOS/memristor circuits

Three authors with strong links to HP (they have all worked at HP but two have moved on) have recently published a thorough review of ReRAM and CBRAM materials and mechanisms*. The article appeared in Nature Nanotechnology and a colleague was kind enough to alert me to the paper and supplementary material (which alone has over 100 references). The authors have done an admirable job in trying to make sense of the enormous number of different materials in which reversible switching phenomena have been observed with around 70 materials being referenced! They divide the materials into Anion (where the mobile species is a positive ion, usually oxygen) and Cation where the mobile species is a negative, usually metal, ion). Broadly speaking the metal oxide based devices we have discussed here at ReRAM-Forum can be categorized as Anion devices whereas CBRAM devices are Cation. (Of course there are exceptions, notably metal oxide based Cation devices which are duly noted in the paper.)

There is a wealth of information in the paper but I am going to focus on a couple of points. First, ReRAM cell material selection. Two distinct phases in the metal-oxygen phase diagram are required. A conductive (metallic phase) with a large solubility of oxygen which is the mobile element during switching. A separate insulating oxide phase provides the high resistance state of the cell. These phases should be separate even at high temperature to eliminate purely thermal switching. In the lead author’s NCCAVS presentation discussed in the DECXX Blog, several phase diagrams were included. For example, Ta, Er and Y all have much simpler oxide phase diagrams than, for example, Ti and Zr which have multiple stable oxide phases which could reduce oxygen mobility after multiple switching cycles.

Second, ReRAM switching mechanisms. The authors point to four types defined with varying degrees of field and thermal driven switching. Field driven Ionic motion is perpendicular to the cell electrodes, i.e. in the direction of the field whereas thermal driven motion is driven by concentration gradients so can be lateral (i.e. parallel to the electrodes). Non-polar devices are identified as being more thermally driven whereas bipolar devices where SET and RESET occur under different voltage polarities are more field driven. The conduction filament model is identified as requiring a degree of thermal induced diffusion as lateral (i.e. perpendicular to the filament and parallel the electrodes) diffusion is needed to rupture the filament. I certainly admire the authors’ courage in attempting to categorize the 50 or so Anion based devices in this way. However, I expect the intent was to point out that there is a continuum of switching possibilities with four chosen to illustrate this.


Finally, the authors present a spider web chart (an example shown above) to describe the readiness of resistive switching devices to meet four specific applications. I am always suspicious of this type of data presentation as it tends to give a far more optimistic view of a technology’s readiness than is in fact the case. The problem is that each parameter (direction in the web) tends to be a ‘best of’ value taken from different devices from different authors and almost certainly not in combination with any of the other parameters. To be fair to the authors, they do point this out and the supplemental data contains the source and more detail on the actual values used in the plot. More interesting in my view is the identification of required values for each of the parameters for the four different applications (Storage (i.e. NAND replacement), Memory (i.e. DRAM replacement), Programmable Logic (i.e FPGA like circuits) and Neuromorphic systems). Of these Memory is the most demanding and furthest from today’s ‘best of’ performance but Logic is the least challenging and apparently within reach. Is this an insight into HP’s focus for bringing this technology to market?

Christie Marrian, www.ReRAM-Forum.com Moderator

*Memristive devices for computing, J. Joshua Yang, Dmitri B. Strukov and Duncan R. Stewart, p13, NATURE NANOTECHNOLOGY, VOL 8, JANUARY 2013, www.nature.com/naturenanotechnology

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>