Modelling and Applications of Memristive Devices

• Autores: Mohamad Moner Al Chawa
• Directores de la Tesis: Rodrigo Picos Gayà (dir. tes.)
• Lectura: En la Universitat de les Illes Balears ( España ) en 2019
• Idioma: inglés
• Tribunal Calificador de la Tesis: Stavros Stavrinides (presid.), Miquel Roca Adrover (secret.), Mireia Bargalló González (voc.)
• Programa de doctorado: Programa de Doctorado en Ingeniería Electrónica por la Universidad de las Illes Balears y la Universidad Politécnica de Catalunya
• Materias:
  • Física
    • Electrónica
  • Ciencias tecnológicas
    • Tecnología electrónica
• Enlaces
  • Tesis en acceso abierto en: RepositoriUIB
• Resumen

  • Introduction In this work, a new approach in place of the traditional Voltage- Current (V − I) method has been used to successfully model unipolar and bipolar ReRAM in Flux-Charge (φ − Q) space by considering the devices as memristors. This approach considers a dynamical system and enables capture of the complex dynamic behavior of ReRAM devices through the use of a simple ordinary differential equation system that can be implemented in circuit simulations. Thus, it is possible to simulate the device behavior for any input signal, regardless the input signal shape. This modeling approach overcomes the drawbacks that appear in traditional complex models in V − I domain which make simulation of large circuits rather difficult or even impractical.

    . Research content A simple model in φ − Q space with a reduced set of parameters was employed to fit successfully the different experimental reset processes for unipolar ReRAM memristive devices. The intrinsic variability in these devices has been analysed by using the φ − Q approach. Physical simulations of devices with different conductive filament sizes were employed to fit the model introduced.

    Afterwards, the relations between the model parameters and the conductive filament geometrical features were characterized indepth.

    The obtained relations were then used to generate a new ensemble of parameters that shows similar statistical properties to the experimental one, thus proving the validity of the approach. In addition, a model to obtain the energy employed in the reset process was presented.

    The reset transition of a bipolar ReRAM memristive device has been analysed based on energy considerations. An analysis of the experimental results has been done in the φ − Q space beside the usual V − I one. The effect of changing the slope of the input signal in the reset point has been considered, and a set of equations to estimate the new parameters has been found. These equations, based on a quasi-static energy analysis, allow the characterization of the reset transition of a bipolar ReRAM memristive device by using only three parameters in addition to the signal slope, a thermal resistance and the reset temperature of the conductive filament. A piecewise model for the reset and set transitions of a bipolar ReRAM memristive device in the φ − Q space has been developed and tested. The model used is very simple and provides accurate simulation results. It also allows the development of simple expressions for the conductance and power consumption, as well as the characterization of the ReRAM memristive device in V − I domain by using two points for any reset or set cycle. The case of a ramp input signal with different slopes has been considered to obtain the model parameters, and the predictions of the model with experimental results have been compared. Finally, a quasi-stationary compact model for bipolar ReRAM memristive device has been implemented and tested for different frequencies.

    . Conclusion A non-linear relation between flux and charge has been used to model the reset transitions of unipolar devices. This non-linear relation is extended to a piecewise one to model the reset/set transitions of bipolar devices. Current and voltage can be obtained by deriving with respect to time the main model magnitudes, flux and charge. For both unipolar and bipolar ReRAM memristive devices, simulated and experimental data have been employed to develop a model that can be easily included in circuit simulators in order to describe resistive switching operation