Resistive Switching and Synaptic Properties of Ni/SiO_xN_y/Si Devices

In the present study, the resistive switching phenomenon in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{Ni} / \text{SiO}_{\mathrm{x}} \mathrm{N}_{\mathrm{y}} / \text{Si}$</tex> structures was investigated. Non-stoichiometric silicon oxynitrides <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\text{SiO}_{\mathrm{x}} \mathrm{N}_{\mathrm{y}})$</tex> obtained by plasma-enhanced chemical vapor deposition on a setup with inductively excited remote plasma and a wide-aperture source were used as active layers in memristor structures. A memristor effect with bipolar switching mechanism was observed in measured I-V characteristics of memristors involving <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{S i O}_{\mathbf{0. 8 9}} \mathbf{N}_{\mathbf{0. 5 2}}$</tex> films as active layers. Stable intermediate states were observed in between the open and closed states. The transition from off- to on-state and back occurs not abruptly (i.e. in a jump-like manner), but rather smoothly over a fairly large range of voltage sweeps. For studying synaptic responses, measurements of electric-current changes observed on increasing the number of pulses applied to the structures in “excitation potentiation” and “inhibition depression” modes were performed. Up to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0 0 0}$</tex> voltage pulses of various durations (from 10 to 60 ms) were applied to the structures under study. In the pulse duration range of 50 to 60 ms, the current changes were found to be roughly linear up to the saturation current. The multi-level nature of resistive states in silicon-based dielectrics and the ability to control their switching parameters characterize the plasticity and prospects for use of such dielectrics in memory devices for neuromorphic applications involving spiking neural networks.