Enver Solan

An enhanced lumped element electrical model of a double barrier memristive device

The massive parallel approach of neuromorphic circuits leads to effective methods for solving complex problems. It has turned out that resistive switching devices with a continuous resistance range are potential candidates for such applications. These devices are memristive systems - nonlinear resistors with memory. They are fabricated in nanotechnology and hence parameter spread during fabrication may aggravate reproducible analyses. This issue makes simulation models of memristive devices worthwhile. Kinetic Monte-Carlo simulations based on a distributed model of the device can be used to understand the underlying physical and chemical phenomena. However, such simulations are very time-consuming and neither convenient for investigations of whole circuits nor for real-time applications, e.g. emulation purposes. Instead, a concentrated model of the device can be used for both fast simulations and real-time applications, respectively. We introduce an enhanced electrical model of a valence change mechanism (VCM) based double barrier memristive device (DBMD) with a continuous resistance range. This device consists of an ultra-thin memristive layer sandwiched between a tunnel barrier and a Schottky-contact. The introduced model leads to very fast simulations by using usual circuit simulation tools while maintaining physically meaningful parameters. Kinetic Monte-Carlo simulations based on a distributed model and experimental data have been utilized as references to verify the concentrated model.

Wave digital emulation of charge-or flux-controlled memristors

Memristors are passive elementary circuit elements, which essentially are resistors with memory. This property makes the memristor well suited for applications in neuromorphic computation circuits. In such circuits the functionality mainly depends on the memristor. Hence, different desired functionalities for the resulting circuit need appropriate devices. Normally, memristors are not commercially available. This makes the use of emulators, which mimic the memristive behavior, expedient. For this purpose, we introduce a method based on wave digital virtualization for emulating real memristors after identifying them from i-u-measurements. The so virtualized memristor can be directly integrated in a real circuit.

Sensitivity analysis of memristors based on emulation techniques

Memristors are novel elementary passive circuit elements, which behaves like a resistor with memory. Because of their similar functionalities with synapses in human brain, they are well suited for neuromorphic computation circuits. Noisy signals and parameter spreading are common in biological systems. Electronic circuits, including memristors for modeling biological systems, are equivalently exposed with these distortions. Both, noisy signals and parameter spreading can harm the functionality of such memristive circuits. A careful investigation needs reproducible sensitivity analysis with memristors. For this purpose, we use memristor emulators, based on the wave digital method, for sensitivity analysis. With this, we get a reproducible investigation method, which can be used in real circuits, before fabricating the real device.

Wave digital emulation of a double barrier memristive device

A memristor is a novel elementary passive device which is essentially a resistor with memory. This new device is capable of different technical applications like non-volatile memory, reconfigurable logic, and neuromorphic computation. Unfortunately, the commercial use of memristors is not common, which complicates the fabrication of devices with an arbitrarily desired functionality. This makes an investigation and development of real memristive circuits dedicated to a specific utilization very cumbersome. In order to circumvent these problems, we propose a wave digital memristor emulator, which is inherently passive like its analog counterpart. Due to its passivity even stable neuromorphic computation circuits with unpredictable interconnection structure can reliably be emulated.

Anticipation of digital patterns

A memristive device is a novel passive device, which is essentially a resistor with memory. This device can be utilized for novel technical applications like neuromorphic computation. In this paper, we focus on anticipation - a capability of a system to decide how to react in an environment by predicting future states. Especially, we have designed an elementary memristive circuit for the anticipation of digital patterns, where this circuit is based on the capability of an amoeba to anticipate periodically occurring unipolar pulses. The resulting circuit has been verified by digital simulations and has been realized in hardware as well. For the practical realization, we have used an Ag-doped TiO2-x-based memristive device, which has been fabricated in planar capacitor structures on a silicon wafer. The functionality of the circuit is shown by simulations and measurements. Finally, the anticipation of information is demonstrated by using images, where the robustness of this anticipatory circuit against noise and faulty intermediate information is visualized.

Wave digital emulation of spike-timing dependent plasticity

Neuromorphic circuits are potential candidates for solving costly computations in an efficient manner. Such circuits, mimicking partial functionalities of the brain, need a large number of components. Simulation models are appropriate for first investigations, but they are very time-consuming regarding complex systems. Hardware realizations of specific components, like memristive devices, in such circuits are not possible for a desired functionality. We propose a methodology based on emulation techniques to overcome such problems. Therefore, the wave digital method has been utilized to get a digital replica of a memristive circuit mimicking long-term changes in the strength of synaptic coupling. In this work, both long-term potentiation as well as long-term depression are successfully emulated. This approach can also be used for a partial emulation of the subsystems, e.g. only memristive devices, of a more complex overall system.

Wave digital information anticipator

Pioneering developments in electrical engineering are based on inspirations from biology. They exhibit naturally an efficient information processing. For example, hardware realizations of memristive circuits mimicking the anticipatory behavior of unicellular organisms like amoebas have been developed in this context. Unfortunately, circuits are not appropriate for algorithms dedicated in the area of digital signal processing. We intended to get an algorithmic model of the anticipation circuit for utilizations in digital signal processing applications. In our approach we have applied the wave digital method to get a digital replica of the analog circuit. This offers several benefits, like the ability for preserving the passivity of the analog counterpart, the possibility for a parallel processing approach, efficiency, and robustness of the resulting algorithmic model. The memristive device of the original circuit has been replaced by a novel multilevel memristor model with suitable features for anticipation of digital patterns. The resulting algorithmic model offers innovative applications in e.g. robotics or artificial neural networks.

Parameter identification of a double barrier memristive device

Memristive systems are nonlinear resistors with memory. Most of them are realized as resistive switching devices in nanotechnology. One example, with appropriate properties especially in neuromorphic applications, is the double barrier memristive device (DBMD). A continuous resistance range makes the DBMD suitable for replacing the synapses in neuromorphic circuits. Structural and functional descriptions based on physical insights can help to obtain a parametric concentrated model of the device for both reproducible investigations as well as emulations. Achieving physically meaningful values for model parameters in order to fit the measured data is not trivial. We propose a parameter identification method based on an optimization problem. Because of very fast and efficient algorithms, the wave digital method has been utilized in the objective function. As an example, a reduced model of the DBMD with optimized parameters for fitting the measured data is shown.

A consistent modeling of passive memcapacitive systems

Memelements — circuit elements with memory — serve novel applications in several technical disciplines. Due to similar functionalities to synapses, they are especially suitable for neuromorphic circuits. Physical and chemical phenomena in the nano-scale lead to the unique information storage characteristic of these elements. Fabrication of devices with memory considering a particular desired functionality is still difficult to achieve. Therefore, simulation models based on a consistent modeling approach are needed. A consistent model in this context should consider important energetic properties of the real device, e.g. passivity. We propose a novel circuit theoretic approach for a consistent modeling of lossless memcapacitive devices. They can be interpreted as nonlinear capacitances with memory. A comparison with existing modeling approaches of such elements underlines benefits as well as the necessity of a consistent model. The strategy introduced here is more general and independent of the underlying model. Beside simulations, it can also be utilized in emulations regarding real-time capable implementations.

Wave digital emulation of general memristors

Present Address Department of Electrical Engineering and Information Technology/Communications Engineering, Ruhr-University, D-44780 Bochum, Germany. Summary Memristive devices are nonlinear resistors with memory. Due to the memory effect, those devices are potential candidates for self-organizing circuits capable of learning from environmental influences in the past. The complexity of single devices with memory in combination with the required huge number of these devices in circuits including them make pre-investigations based on simulations very inefficient and time-consuming. Flexible and real-time capable memristive emulators, which can directly be incorporated into real circuits, can overcome this problem. In our approach, we introduce a general memristor emulator based on wave digital principles. The proposed emulator is flexible, robust, efficient, and it preserves the passivity of the real device in a digital signal processing sense. All these properties result in a reusable emulator, independent of a particular application. This work lists the wave digital emulations of different models from ideal to extended memristors. As an example for an extended memristor, the wave digital emulation of a double barrier memristive device is demonstrated.