Nanoionic redox-based resistive switches - challenges and prospects

48 downloads 83 Views 15KB Size Report
In terms of information theory, a ReRAM cell operates on the configuration of ... R. Waser (Ed.), Nanoelectronics and Information Technology (3rd edition), 2012.
Nanoionic redox-based resistive switches - challenges and prospects Rainer Waser1,2*, Vikas Rana1, Stephan Menzel1, Eike Linn2 1

PGI-7 & JARA-FIT, Forschungszentrum Jülich, 52425 Jülich, IWE2 & JARA-FIT, RWTH Aachen University, 52056 Aachen, Germany

2

Redox-Based Resistive Switching Memories (ReRAM), also called nanoionic memories or memristive elements, are widely considered to provide a potential leap beyond the limits of Flash (with respect to write speed, write energies) and DRAM (with respect to scalability, retention times) as well as energyefficient approaches to neuromorphic concepts. In terms of information theory, a ReRAM cell operates on the configuration of ions as the state variable. Such a cell is generally built by a MIM structure, composed of an ion or mixed ion-electron conducting (i. e. resistive) material I sandwiched between two (typically different) electron conductors M as electrodes. The materials I are oxides, higher chalcogenides, or other ionic solids. Often, an initial electroforming cycle is required before these MIM cells can be electrically switched between at least two different resistance states. For bipolar resistive switching elements, a range of material systems exist in which ionic transport and redox reactions on the nanoscale provide the essential mechanisms for the switching [1]. One class relies on mobile cations which are easily created by electrochemical oxidation of the corresponding electrode metal, transported in the insulating layer, and reduced at the inert counterelectrode (so-called electrochemical metallization memories, ECM). Another important class (so-called valence change memories, VCM) operates through the migration of ions which constitute the I phase, typically oxygen ions, between a metal electrode and a reduced part of the oxide. Here, the reduction of the cation valences in the cation sublattice provides a local (typically filamentary) metallic or semiconducting phase. In both, ECM and VCM, the electrochemical nature of the memristive effects triggers a bipolar memory operation. To evaluate the energy-relevant parameters, the analysis must be extended from a discrete single cell to a complete circuit such as arrays [2]. In particular, the impact of parasitics such as the bit line capacitance per cell and technical boundary conditions such as supply voltages or minimum sense voltage levels have to be considered [3]. The analysis must also comprise the select device issue, i. e. an inherent nonlinearity of the cell (A), a complementary resistive switch (B), a bipolar diode (C) or a transistor in a 1T1R concept (D). It is an essential requirement for all non-volatile memories that the switching kinetics shows an extremely high non-linearity to allow for a very fast switching in the ns range at high (voltage) stimuli and long retention times in the range of years at low stimuli. In the case of filamentary VCM-type cells, the non-linearity originates from a combination of a field-enhanced ion mobility and a local temperature effect as shown by FEM simulations [4]. If we consider a typical ReRAM stack based on a VCM cell and scale the feature size F in vertical and lateral dimensions, the non-linearity improves and the switching energy decreases. For ECM cells, we have been using a STM technique to demonstrated that the Ag critical nucleus formation is rate limiting the kinetic process [5]. In a subsequent study, we were able to identify the sequence of rate limiting steps over 12 orders of magnitude in time [6]. In the area of the application of redox-based memristive elements for neuromorphic computing the aim is to realize computational tasks in brain-like multi-parallel manner. The use of ultimately scaled nanocrossbar array or dedicated circuits could pave the path to energy-efficient solutions, either by realizing implication logic or implementing associative networks. Acknowledgement: SFB 917, Intel for financial support, and IMEC of 1T substrates.

References: [1] [2] [3] [4] [5] [6] *

R. Waser, M. Aono, Nat. Mater. 6 (2007) 833. V. V. Zhirnov, R. K. Cavin, S. Menzel, E. Linn, S. Schmelzer, D. Bräuhaus, C. Schindler, R. Waser, Proc. IEEE 98 (2010) 2185. R. Waser (Ed.), Nanoelectronics and Information Technology (3rd edition), 2012 S. Menzel, M. Waters, A. Marchewka, U. Böttger, R: Dittmann, R. Waser, Adv. Funct. Mater. 21 (2011) 4487. I. Valov, I. Sapezanskaia, A. Nayak, T. Tsuruoka, T. Bredow, T. Hasegawa, G. Staikov, M. Aono, R. Waser, Nat. Mater. 11 (2012) 530. S. Menzel, S. Tappertzhofen, R. Waser, I. Valov, PCCP 15 (2013) 6945.

E-mail: [email protected]