Finn Zahari

This study examines the development of TiN/SiO_x/Cu/SiO_x/TiN memristive devices for neuromorphic applications using wedge‐type deposition and Monte Carlo simulations. Identifying critical parameters for the desired device characteristics can be challenging with conventional trial‐and‐error methods, which often obscure the effects of varying layer compositions. By employing an off‐center thermal evaporation method, a thickness gradient of SiO_x and Cu on a 4 inch wafer is created, facilitating detailed resistance map analysis through semiautomatic measurements. This approach allows for investigating the influence of layer composition and thickness while keeping other process conditions constant. Combining experimental data with simulations provides a precise understanding of layer thickness distribution and its impact on device performance. Optimizing the SiO_x layers to be below 12 nm, coupled with a discontinuous Cu layer with a nominal thickness under 0.6 nm, exhibits analog switching properties with an R_on/R_off ratio of >100, suitable for neuromorphic applications, while R × A and power exponent γ analysis show signs of multiple conduction mechanisms. The findings highlight the importance of SiO_x and Cu thickness in determining switching behavior, offering insights for developing high‐performance analog switching components for bioinspired computing systems.

Memristors are promising candidates for new memory technologies and are capable to mimic synapses in artificial neural networks. The switching in memristive devices occurs typically in few nanometer thin dielectric layers. The direct observation of the switching mechanism is crucial for better comprehension and improvements of memristors. Therefore, in situ experiments are conducted in a transmission electron microscope (TEM). However, sample preparation processes and electron beam irradiation can lead to a chemical and structural modification of the active layers. Moreover, devices may show significant device-to-device variability due to the details of processing parameters. Thus, it is essential to characterize the identical device electrically before microstructural analysis.

Memristive devices are under intense development as non‐volatile memory elements for extending the computing capabilities of traditional silicon technology by enabling novel computing primitives. In this respect, interface‐based memristive devices are promising candidates to emulate synaptic functionalities in neuromorphic circuits aiming to replicate the information processing of nervous systems. A device composed of Nb/NbO_x/Al₂O₃/HfO₂/Au that shows promising features like analog switching, no electro‐forming, and high current‐voltage non‐linearity is reported. Synchrotron‐based X‐ray photoelectron spectroscopy and depth‐dependent hard X‐ray photoelectron spectroscopy are used to probe in situ different resistance states and thus the origin of memristive switching. Spectroscopic evidence for memristive switching based on the charge state of electron traps within HfO₂ is found. Electron energy loss spectroscopy and transmission electron microscopy support the analysis. A device model is proposed that considers a two‐terminal metal–insulator–semiconductor structure in which traps within the insulator (HfO₂/Al₂O₃) modulate the space charge region within the semiconductor (NbO_x) and, thereby, the overall resistance. The experimental findings are in line with impedance spectroscopy data reported in the companion paper (Marquardt et al). Both works complement one another to derive a detailed device model, which helps to engineer device performance and integrate devices into silicon technology.

Memristive devices for neuromorphic computing have been attracting ever growing attention over the last couple of years. In neuromorphic electronics, memristive devices with multi‐level resistance states are required to accurately reproduce synaptic weights. Here, a memristive device based on a multilayer oxide system (Nb/NbO_x/Al₂O₃/HfO₂/Au), which features a filamentary‐free, homogenous interfacial resistive switching mechanism, is investigated. To gain a deeper insight into the switching mechanism, impedance spectroscopy (ImpSpec), X‐ray photoelectron spectroscopy (XPS), and transmission electron microscopy (TEM) are exploited. While this work focuses on the analysis of impedance and current‐voltage characteristics, XPS and TEM investigations can be found in a companion paper (Zahari et al.). In the course of this investigation, potentiodynamic impedance spectroscopy (PD‐ImpSpec) and time resolved impedance spectroscopy (TR‐ImpSpec) in combination with transient analysis are used. Evidence is presented of switching kinetics at voltages above 2.1 V directly related to changes in Schottky barrier resistance. These switching kinetics can in turn be interpreted by the charging and discharging of double positively charged oxygen vacancies ≈ 0.9 eV. The results of the impedance analysis are translated into a more general model for memristive devices to map the physical processes during switching.

A large number of simulation models have been proposed over the years to mimic the electrical behavior of memristive devices. The models are based either on sophisticated mathematical formulations that do not account for physical and chemical processes responsible for the actual switching dynamics or on multi-physical spatially resolved approaches that include the inherent stochastic behavior of real-world memristive devices but are computationally very expensive. In contrast to the available models, we present a computationally inexpensive and robust spatially 1D model for simulating interface-type memristive devices. The model efficiently incorporates the stochastic behavior observed in experiments and can be easily transferred to circuit simulation frameworks. The ion transport, responsible for the resistive switching behavior, is modeled using the kinetic cloud-in-a-cell scheme. The calculated current–voltage characteristics obtained using the proposed model show excellent agreement with the experimental findings.

A large number of simulation models have been proposed over the years to mimic the electrical behaviour of memristive devices. The models are based either on sophisticated mathematical formulations that do not account for physical and chemical processes responsible for the actual switching dynamics or on multi-physical spatially resolved approaches that include the inherent stochastic behaviour of real-world memristive devices but are computationally very expensive. In contrast to the available models, we present a computationally inexpensive and robust spatially 1D model for simulating interface-type memristive devices. The model efficiently incorporates the stochastic behaviour observed in experiments and can be easily transferred to circuit simulation frameworks. The ion transport, responsible for the resistive switching behaviour, is modelled using the kinetic Cloud-In-a-Cell scheme. The calculated current-voltage characteristics obtained using the proposed model show excellent agreement with the experimental findings.

Biological neural networks outperform current computer technology in terms of power consumption and computing speed while performing associative tasks, such as pattern recognition. The analogue and massive parallel in-memory computing in biology differs strongly from conventional transistor electronics that rely on the von Neumann architecture. Therefore, novel bio-inspired computing architectures have been attracting a lot of attention in the field of neuromorphic computing. Here, memristive devices, which serve as non-volatile resistive memory, are employed to emulate the plastic behaviour of biological synapses. In particular, CMOS integrated resistive random access memory (RRAM) devices are promising candidates to extend conventional CMOS technology to neuromorphic systems. However, dealing with the inherent stochasticity of resistive switching can be challenging for network performance. In this work, the probabilistic switching is exploited to emulate stochastic plasticity with fully CMOS integrated binary RRAM devices. Two different RRAM technologies with different device variabilities are investigated in detail, and their potential applications in stochastic artificial neural networks (StochANNs) capable of solving MNIST pattern recognition tasks is examined. A mixed-signal implementation with hardware synapses and software neurons combined with numerical simulations shows that the proposed concept of stochastic computing is able to process analogue data with binary memory cells.

Memristive devices have been the object of intensive studies for non-volatile memories, neuromorphic engineering and image processing algorithms. The intrinsic properties of these devices are determined by its I-V characteristics influenced by different process parameters. The double-barrier memristive devices investigated in this work are based on the motion of charged species, i.e. oxygen vacancies or ions, within a NbO_x layer. Since the layers are deposited by magnetron sputtering, it is important to understand the physics of the discharge and its effect on the film properties. For plasma diagnostics we used a calorimetric probe, which can be operated simultaneously as a passive thermal probe for energy flux measurement and as a planar Langmuir probe for measuring the ion current, the floating and plasma potentials and the electron temperature. In particular, we investigated the reactive sputter deposition of the NbO_x layer by a floating and a biased probe. The parameters were determined in dependence on the radial position of the probe across the substrate region. The results allowed us to find correlations between the plasma parameters and the electrical properties of the memristive devices produced on one 100-mm wafer. Furthermore, we could point out the dominating factors affecting strongly the properties of these thin films.