(Invited) Improving Memory Performance of Cu/HfO2/Pt Conducting-Bridge RAM by Solvent Substitution

Research and development of resistive memories advanced from the standpoint of circuit [1,2] and material design technologies [3]. However, controlling memory characteristics, such as switching voltages, by selecting materials used for electrodes and memory layers and by controlling their crystallinity seems not to be very effective for conducting-bridge memory (CB-RAM). Under the circumstance, it was reported that both switching voltages and reset current were decreased by providing water to the HfO₂ memory layer in CB-RAM with a Cu/polycrystalline HfO₂ (poly-HfO₂)/Pt structure [4]. These results suggest that water absorbed in grain boundary due to capillary condensation causes the migration of Cu ions through electrochemical dissolution and diffusion. Therefore, we consider that the poly-HfO₂ layer should not be regarded as a diffusive medium for the migration of Cu ions as ever but should be regarded as a natural nano-porous body that absorbs water, where the pore corresponds to grain boundary.

In this paper, we propose the concept of ‘pore-engineering’ as a method for controlling device performance, through (i) the kind and property of solvents, (ii) the shape and size of the pores, and (iii) the physical and chemical properties of the pore surface. In particular, the stable switching operation with reduced switching voltages that is robust against environmental change was achieved by providing ionic liquid (IL) to the poly-HfO₂ memory layer of a Cu/poly-HfO₂/Pt structure, evidencing the item (i).

A poly-HfO₂ film with the thickness of 25 nm was deposited by using RF reactive sputtering method after the deposition of Pt bottom electrode with the thickness of 50 nm. A Cu-probe was used as a top electrode by contacting the surface of the poly-HfO₂ film with the probe. The effect of supplying solvents to the poly-HfO₂ film on I-V characteristics was measured by dropping a solvent on the surface of the poly-HfO₂ film using a dropper and contacting the poly-HfO₂ film surface with the Cu-probe through the droplet.

First, forming voltage (V_form) was confirmed to be drastically decreased from the value of V_form in solvent-non-supplied Cu/poly-HfO₂/Pt structures by supplying water. Then, we investigated the temperature dependence of V_form in water-supplied samples.V_form increased with decreasing temperature, suggesting that the cooling of water hinders the ionization and migration of Cu ions. Nevertheless, V_form at -30 °C was still smaller than V_form of the solvent-non-supplied samples in the atmosphere at RT. These results suggest that the HfO₂ layer plays a role as a nano-porous body and water that is absorbed in the grain boundary plays a key role on the resistive switching.

Figures 1(a) and (b) respectively show cumulative probabilities of V_form and V_set for water-supplied (squares) and solvent-non-supplied (circles) Cu/HfO₂/Pt structures in the atmosphere. Both V_form and V_set are reduced by supplying water. However, we also confirmed that RS of water-supplied samples was unstable against environmental changes such as changes in temperature, voltage, and ambient pressure, due to the instability of the liquid phase of water. Therefore, stable solvent that can reduce switching voltages and can be replaced with water is necessary for the practical use of CB-RAM. Cumulative probabilities of V_form and V_set for IL-supplied samples (triangles) show that switching voltages can be reduced by supplying IL more effectively than by supplying water. In addition, thanks to the wide potential window and low vapor pressure of IL, high robustness against voltage stress and stable RS in vacuum were achieved respectively in IL-supplied samples. The high effectiveness of supplying solvents in the improvement of memory characteristics strongly supports the validity of the pore-engineering.

 [1] K. Kinoshita et al., APL 93, 033506 (2008). [2] S.-G. Koh et al., APL 104, 083518 (2014). [3] K. Tsunoda et al., Tech. Digest IEDM 2007, p.767. [4] S. Hasegawa et al., ECS Transactions 50, 61 (2013).