(Invited) Atomistic Simulations for Understanding Microscopic Mechanism of  Amorphous-Tantalum-Oxide-Based Resistive Switching Devices

Wednesday, 4 October 2017: 09:30
Camellia 4 (Gaylord National Resort and Convention Center)
S. Watanabe (Dept. Materials Eng., UTokyo, and CMI2-NIMS), B. Xiao (School of Chem. and Chem. Eng., Yantai Univ.), and W. Li (Dept. Materials Eng., UTokyo)
Among oxide-based resistive switching devices, amorphous-TaOx based ones show good endurance and switching performances. Their switching behaviors are understood based on the formation/rupture of conductive filaments (CFs). However, the atomistic structures of CFs and their microscopic formation/rupture processes have not been sufficiently clarified yet. In this talk, we present the results of our atomistic simulations (most of which are with the density functional theory (DFT)) and discuss several aspects related to the nature and formation processes of CFs as follows.
1) CFs in the Cu/amorphous-Ta2O5/Pt heterostructure: Our simulations reveals that single Cu chains in Ta2O5 cannot work as conductive filaments (CFs), while Cu nanowires with a diameter of three atoms or larger can work as CFs. The stability of the Cu nanowires has been checked by ab initio molecular dynamics [1].
2) CFs in the Pt/TaOx/Pt heterostructure with x < 2.5: Our results suggest that not O vacancy chain but the Ta-Ta bonding mainly contributes to the CFs [2].
3) Difference in the atomistic features between Cu/ Ta2O5 and Pt/Ta2O5 interfaces: In the former, considerable portion of interface Cu atoms tend to migrate to the amorphous Ta2O5 layer, while similar behavior is not seen in the latter [3].
4) Ion migration behaviors in TaOx: Our results of ab initio molecular dynamics and nudged elastic band calculations show that the diffusion coefficient in TaOx strongly depends on x [3].
5) Construction of neural network interatomic potentials: We constructed interatomic potential combining the DFT calculations and neural network for simulations of Cu migration behavior in amorphous-Ta2O5 to achieve computation speed and reliability simultaneously. The pathways and barrier energies for Cu diffusion calculated using the NN potential agree well with those obtained from DFT calculations [5].
 The final part of the present works was partly supported by the Support Program for Starting up Innovation hub from Japan Science and Technology Agency (JST), and CREST-JST, Japan.
[1] B. Xiao, T. Gu, T. Tada and S. Watanabe, J. Appl. Phys. 115, 034503 (2014).
[2] B. Xiao and S. Watanabe, Nanoscale 6, 10169 (2014).
[3] B. Xiao and S. Watanabe, ACS Appl. Mater. Interfaces 7, 519 (2015).
[4] B. Xiao and S. Watanabe, Extended abstract of SSDM2014 (September 8-11, 2014, Tsukuba, Japan).
[5] W. Li, Y. Ando and S. Watanabe, submitted.