Search

Devices with volatile memristive switching and self‐sustained relaxation oscillations have received significant attention for their use in neuromorphic computing and solving optimization problems. However, obtaining devices with stable response and reliable oscillation remains challenging. This work describes the utility of metal/oxide interlayers in achieving reliable device performance with tuneable characteristics using Nb‐Nb2O5‐Pt structures. Detailed physical characterization of the Nb/Nb2O5 interlayer region shows that it consists of a thin, near‐stoichiometric Nb2O5 and an NbOx layer with a graded oxygen content. The impact of these interlayers on device performance is assessed by investigating their switching characteristics and oscillation dynamics. It is shown that the presence of the interlayer has a direct impact on the memristive switching mode and parameters threshold switching characteristics parameters. These dependencies are explained with reference to a lumped‐element circuit model of the device and shown to be dominated by changes in device resistance caused by interlayer formation. Significantly, stable oscillation with tuneable frequency is obtained by increasing the thickness of the reactive Nb electrode. Then, by simulating an array of coupled oscillators, a unique Hamiltonian solver is demonstrated. These results provide pathways for optimizing volatile memristive switching characteristics and guidance on how such devices can be employed to solve combinatorial optimization problems. [ABSTRACT FROM AUTHOR]

Devices exploiting negative differential resistance (NDR) are of particular interest for analog computing applications, including oscillator‐based neural networks. These devices typically exploit the continuous S‐shaped current–voltage characteristic produced by materials with a strong temperature‐dependent electrical conductivity, but recent studies have also highlighted the existence of a second, discontinuous (snap‐back) characteristic that has the potential to provide additional functionality. The development of devices based on this characteristic is currently limited by uncertainty over the underlying physical mechanism, which remains the subject of active debate. In situ thermoreflectance imaging and a simple model are used to finally resolve this issue. Specifically, it is shown that the snap‐back response is a direct consequence of current localization and redistribution within the oxide film, and that material and device dependencies are consistent with model predictions. These results conclusively demonstrate that the snap‐back characteristic is a generic response of materials with a strong temperature‐dependent conductivity and therefore has the same physical origin as the S‐type characteristic. This is a significant outcome that resolves a long‐standing controversy and provides a solid foundation for engineering functional devices with specific NDR characteristics. [ABSTRACT FROM AUTHOR]

Current‐controlled negative differential resistance has significant potential as a fundamental building block in brain‐inspired neuromorphic computing. However, achieving the desired negative differential resistance characteristics, which is crucial for practical implementation, remains challenging due to a lack of consensus on the underlying mechanism and design criteria. Here, a material‐independent model of current‐controlled negative differential resistance is reported to explain a broad range of characteristics, including the origin of the discontinuous snap‐back response observed in many transition metal oxides. This is achieved by explicitly accounting for a non‐uniform current distribution in the oxide film and its impact on the effective circuit of the device rather than a material‐specific phase transition. The predictions of the model are then compared with experimental observations to show that the continuous S‐type and discontinuous snap‐back characteristics serve as fundamental building blocks for composite behavior with higher complexity. Finally, the potential of our approach is demonstrated for predicting and engineering unconventional compound behavior with novel functionality for emerging electronic and neuromorphic computing applications. [ABSTRACT FROM AUTHOR]