(Invited) Conductive Bridging RAM (CBRAM): Then, Now, and Tomorrow

Applications of electrochemistry are ubiquitous in modern engineering, but digital computer memory is not typically included in the list. This is because the mainstays of the industry for the past 40 years – volatile DRAM and SRAM and nonvolatile NOR and NAND flash – are all based on the movement and storage of electrons alone, without any accompanying atomic motions. With the advent of nonvolatile Conductive Bridging RAM (CBRAM)*, which operates by the electrically induced making and breaking of atomic-scale conductive filaments inside a solid-state capacitor, electrochemistry has become relevant in the field of digital memory. Here, we present Te-based “subquantum” CBRAM cells as an important step forward in the development of a nonvolatile memory technology with power and performance advantages over flash memory. This new class of CBRAM cells has satisfied technological requirements which CBRAM cells of the past could not meet, such as high-temperature retention, integration into a standard CMOS process with no alterations, full JEDEC qualification, and yields suitable for high-volume manufacturing. The present talk summarizes similarities and differences between subquantum CBRAM cells and CBRAM cells of the past, including their switching mechanisms, materials, electrical characteristics, and commercial applications. For the switching mechanism, a view of the prototypical CBRAM cell as being a tiny electroplating “bath” (typically Cu or Ag) is shown to be poorly suited to subquantum CBRAM cells, whose switching mechanism is suggested to rely on a replacement reaction involving two mobile species (Te and O), neither of which is a metal. A Te filament is formed by this replacement reaction, as inferred from electrical analogies with the Ag- and Cu-based CBRAM cells of the past. Such nonmetal filaments allow subquantum CBRAM cells to reach ON-state conductance levels well below the quantum G 0 =2e 2 /h characteristic of a 1-atom filament of a metal (hence the term “subquantum”). New insight into the chemistry of this replacement reaction is deduced here from the influence of the anode stoichiometry. The difference in mechanisms between subquantum CBRAM cells and CBRAM cells of the past drives a difference in materials. For the anode, the necessity of using an amorphous alloy is discussed, and the thermal anneals encountered in standard CMOS processing are shown to place a limitation on which alloys may be used for Te-based cells. Other fundamental and practical requirements for the anode are discussed, as are requirements for the switching layer and anode capping layer. The success of a new memory technology is determined by its acceptance in the market. This depends in part on the qualification, manufacture, and assembly of the new technology being compatible with existing infrastructure. Even with such compatibility, though, the new memory must still offer advantages over the incumbent technologies. Distinguishing features of subquantum CBRAM, as compared to NOR flash, include a low-energy write operation, an ultra-fast write speed, inherently low-voltage operation, a low-power read operation, scalability to CMOS technology nodes below 65nm, and low integration cost. The Internet of Things (IoT) is a market having applications for which these attributes are particularly well-suited. The present talk will discuss select aspects of system and circuit design for the IoT, and will cite examples of IoT-friendly designs which capitalize on the distinguishing features of CBRAM. It is hoped that the present talk is suitable as an introduction to CBRAM, and also as a re-introduction to CBRAM for those familiar with the field as it stood prior to the innovations of the past several years. * Adesto and CBRAM are Trademarks of Adesto Technologies Corp.