Simulation of Phase Transformations Occurring in Ge-Rich GeSbTe Phase Change Memories—Coupling of Multi-Phase Field and Thermal Models

Phase change memories (PCM) rely on phase-change materials to store information. During memory operations (write and erase), the PCM cell undergoes Joule heating during the application of appropriate electrical pulses. Thanks to a highly resistive electrode in contact with the PCM layer, phase change occurs in a dome-shaped active zone of the phase change material located above this electrode [1]. Depending on heating and cooling profiles, the dome ends up in the amorphous or in the crystalline state, both states having different electrical resistivities.<br/>The stoichiometric ternary compound Ge2Sb2Te5 (GST) is widely used in PCM technologies thanks to its interesting physical properties (ovonic switching, fast and congruent crystallization, large resistivity difference between the two states) [2]. However, because of its relatively low crystallization temperature, at which the amorphous state spontaneously crystallizes (effectively losing the stored information), GST is not suited for some industrial applications that require high operating temperatures, such as automotive microcontrollers [3]. To increase the crystallization temperature, GST can be enriched with germanium [4]. However, this leads to segregation during crystallization: two types of crystalline phases appear, germanium-rich phases and phases close to pure GST. The crystallization of GST is congruent, but that of Ge-rich GST is not.<br/><br/>In order to simulate the crystallization of the material after the microfabrication process and during memory operations, we have developed a multi-phase field model coupled with an orientation model [5]. This model is able to reproduce the segregation effect and to perform memory operations. However, it makes substantial approximations for the couplings to the thermal field and the electrical current that accompany memory operations.<br/>In this work, we couple the multi-phase field approach with a thermal model that takes into account the multi-phase character of the material. Physical parameters such as thermal conductivities and specific heat have been determined with available experimental data, they are phase and temperature dependent when it is relevant. Such dependence is especially important for thermal conductivities because there are more than two orders of magnitude between their values in the different phases. Latent heats that were not taken into account by the previous model are also included to account for the heat produced during crystallization and absorbed during melting.<br/>Chemical species diffusion and thermal diffusion are phenomena occurring at different time scales. This leads to a numerically stiff problem when they are coupled in simulations. We combine multiple numerical approaches and optimizations to keep our results accurate and simulation times low.<br/>With this thermal multi-phase field model and with a heat source provided by an electrical model, first simulations are proposed and discussed. We investigate different electrical pulses and their effects on crystallization scenarios. While our model remains qualitative in nature, it can give insight on the crystallization mechanism occurring during memory operations.<br/><br/>[1] V. Sousa, G. Navarro, N. Castellani, M. Coué, O. Cueto, C. Sabbione, P. Noé, L. Perniola, S. Blonkowski, P. Zuliani, R. Annunziata, in 2015 Symposium on VLSI Technology (VLSI Technology) (2015), pp. T98–T99.<br/><br/>[2] P. Guo, A. M. Sarangan, I. Agha, Applied Sciences. 9 (2019)<br/><br/>[3] F. Arnaud et al., in 2018 IEEE International Electron Devices Meeting (IEDM) (2018), p. 18.4.1-18.4.4.<br/><br/>[4] P. Zuliani, E. Varesi, E. Palumbo, M. Borghi, I. Tortorelli, D. Erbetta, G. D. Libera, N. Pessina, A. Gandolfo, C. Prelini, L. Ravazzi, R. Annunziata, IEEE Transactions on Electron Devices. 60, 4020–4026 (2013).<br/><br/>[5] R. Bayle, O. Cueto, S. Blonkowski, T. Philippe, H. Henry, M. Plapp, Journal of Applied Physics. 128, 185101 (2020).