(Invited) Resistive RAM Memories from a Material Perspective: Exploration of the Switching for Several Oxides Using Ab Initio Simulations

Abstract

The Resistive RAM memories are among the most promising technologies opening new perspectives beyond the current FLASH memory characteristics [1]. Several options compatible with a CMOS integration exist imposing constraints in term of electrode materials and dielectric layers. Indeed, many options are tested since a few years and the actual literature shows a wide dispersion of very encouraging results. In the same time the theoretical effort to simulate and to model the RRAM essential features has rapidly grown. By sustaining the technological efforts, we need to provide sound basis in order to pursue the RRAM development by decreasing the trial error period. Nevertheless, the origin of the mechanisms at the heart of a RRAM is sufficiently complex to resist to both semi-classical and empirical approaches inherited from the microelectronics. One of the main reason is that these devices mix two types of motions: the electronic current and the ionic rearrangement mainly for the dielectric, which can also interfere in turn with the electrode. Therefore, the atomistic simulation appears as the main alternative to analyze the situation without doing too strong assumptions. Our contribution is based onto our previous works for a Ti/HfO2 OxRRAM [2-4] and a Cu/Al2O3 CBRAM [5-6] plus a new and original mechanism for the HfO2 dielectric breakdown. Depending onto the electrode work function and its oxygen solubility, we show that the Forming step is mainly governed by the available charge states for the oxygen vacancy and the oxygen interstitial. For the Ti electrode and HfO2 dielectric, this leads to only one scenario where several oxygen rails leave the dielectric toward the reactive electrode. Thereafter we propose a Forming scenario for a Ti/HfO2 OxRRAM up to obtaining a conductive filament through the dielectric, see Fig. 1. This filamentary mode is obtained by basing our thinking onto equilibrium thermodynamics in between Ti oxides and Hf oxides. This model is then used to evaluate the activation energies implied in the SET and RESET steps: in order to move an oxygen inward and outward the conductive filament. In parallel, the density of states is then analyzed to evaluate relative change of filament conductivity. Then in a second part, we show that a Cu/Al2O3 CBRAM shares the same principles for the oxygen movements but in a more complex way. By analyzing the activation energies for Cu, O and Al, we show that oxygen movements need to be combined with the copper interstitial. A copper based electrode reacts with the dielectric by a combined motion: copper ions eventually end their run in substitutional oxygen sites. This movement can also be accompanied by a slight reorganization of some aluminum atoms which is usually not possible with HfO2 where Hf atoms stay at rest due to their low diffusivity. In conclusion we draw a possible track to follow in order to combine both types of mechanisms in between OxRRAM and CBRAM. Bibliography [1] M. Ueki, K. Takeuchi, T. Yamamoto, A. Tanabe, N. Ikarashi, M. Saitoh, T. Nagumo, H. Sunamura, M. Narihiro, K. Uejima, K. Masuzaki, N. Furutake, S. Saito, Y. Yabe, A. Mitsuiki, K. Takeda, T. Hase, and Y. Hayashi, in VLSI Technology (VLSI Technology), 2015 Symposium on (2015) pp. T108-T109[2] K.-H. Xue, B. Traore, P. Blaise, L. R. C. Fonseca, E. Vianello, G. Molas, B. D. Salvo, G. Ghibaudo, B. Magyari-Köpe, and Y. Nishi, IEEE Transactions on Electron Devices 61, 1394 (2014). [3] B. Traore, P. Blaise, E. Vianello, E. Jalaguier, G. Molas, J. Nodin, L. Perniola, B. De Salvo, and Y. Nishi, in IEEE IRPS (2014) pp. 5E.2.1-5E.2.5.[4] B. Traore, P. Blaise, E. Vianello, L. Perniola, B. De Salvo, and Y. Nishi, Electron Devices, IEEE Transactions on Electron Devices 63, 360 (2016). [5] J. Guy, G. Molas, P. Blaise, A. Roule. G. LeCarval, V. Delaye, A. Toffoli, G. Ghibaudo, F. Clermidy, B. de Salvo, L. Perniola, in IEEE Transactions on Electron Devices 62, pp. 3482-3489, (2015). [6] C. Nail, P. Blaise, G. Molas, SISC proceedings (2015) Figure 1