Conductive bridging random access memory—materials, devices and applications

Abstract

We present a review and primer on the subject of conductive bridging random access memory (CBRAM), a metal ion-based resistive switching technology, in the context of current research and the near-term requirements of the electronics industry in ultra-low energy devices and new computing paradigms. We include extensive discussions of the materials involved, the underlying physics and electrochemistry, the critical roles of ion transport and electrode reactions in conducting filament formation and device switching, and the electrical characteristics of the devices. Two general cation material systems are given-a fast ion chacogenide electrolyte and a lower ion mobility oxide ion conductor, and numerical examples are offered to enhance understanding of the operation of devices based on these. The effect of device conditioning on the activation energy for ion transport and consequent switching speed is discussed, as well as the mechanisms involved in the removal of the conducting bridge. The morphology of the filament and how this could be influenced by the solid electrolyte structure is described, and the electrical characteristics of filaments with atomic-scale constrictions are discussed. Consideration is also given to the thermal and mechanical environments within the devices. Finite element and compact modelling illustrations are given and aspects of CBRAM storage elements in memory circuits and arrays are included. Considerable emphasis is placed on the effects of ionizing radiation on CBRAM since this is important in various high reliability applications, and the potential uses of the devices in reconfigurable logic and neuromorphic systems is also discussed.

Interface Engineering via MoS2 Insertion Layer for Improving Resistive Switching of Conductive‐Bridging Random Access Memory

“…Since the filament rupture is actually a field‐assisted thermal diffusion process, temperature influences the rate of redox reaction. The I RESET can be expressed aswhere E a , β, and q are activation energy for cation hopping, dimensionless correction factor, and elementary charge, respectively. k b is Boltzmann constant, T is temperature, c n is cation concentration out of CF, and c 0 is cation concentration within CF.…”

Lithium‐Battery Anode Gains Additional Functionality for Neuromorphic Computing through Metal–Insulator Phase Separation

“…Under bias, mobile cations, such as Ag + or Cu 2+ , are transported through the interstitials of a solid electrolyte (usually a chalcogenide) forming and dissolving metallic filaments via redox reactions at the electrodes. Typical diffusivities range around 10 −4 –10 −7 cm 2 s −1 in Cu 2–α S electrolyte films, which result in low switching currents under applied voltages . Both classes of memristive devices rely on filament formation, a highly complex process still under investigation.…”

“…Typical diffusivities range around 10 −4 -10 −7 cm 2 s −1 [26] in Cu 2-α S electrolyte films, which result in low switching currents under applied voltages. [27][28][29][30] Both classes of memristive devices rely on filament formation, a highly complex process still under investigation. Although empirical operation protocols have been devised for voltage and current versus time, it is difficult to predict the size and number of formed filaments, which hinders reversibility, accuracy and symmetry.…”

Memristive and CMOS Devices for Neuromorphic Computing

“…Two resistance transitions can be noted by the current-voltage characteristics shown in Figure 11b, which evidence both the abrupt nature of the set process due to the positive feedback loop involving the two driving forces for ion migration, namely the electric field and temperature, and the more gradual dynamics of the reset process due to the negative feedback occurring within the device as a negative pulse is applied [66]. Similar to the bipolar RRAM described in Figure 11b, which typically relies on switching layers, including HfO x [67], TaO x [68], TiO x [69], SiO x [70], and WO x [71], the conductive-bridge random access memory (CBRAM), where metallic CFs are created/disrupted between active Cu/Ag electrodes, has also received strong interest in recent years [72]. In addition to bipolar RRAM concepts, another type of filamentary RRAM called unipolar RRAM, typically based on NiO [73][74][75], has been widely investigated, evidencing that pulses with the same polarity can induce both set and reset processes as a result of the key role played by Joule heating for the creation/disruption of CF [73,75].…”