20 May 2016 in MRS Advances
7 August 2017 in Applied Physics Letters
Mutli-level switching in resistive memory devices enables a wide range of computational paradigms, including neuromorphic and cognitive computing. To this end, we have developed a bi-layer tantalum oxide based resistive random access memory device using Hf as the oxygen exchange layer. Multiple, discrete resistance levels were achieved by modulating the RESET pulse width and height, ranging from 2 kΩ to several MΩ. For a fixed pulse height, OFF state resistance was found to increase gradually with the increase in the pulse width, whereas for a fixed pulse width, the increase in the pulse height resulted in drastic changes in resistance. Resistive switching in these devices transitioned from Schottky emission in the OFF state to tunneling based conduction in the ON state, based on I-V curve fitting and temperature dependent current measurements. These devices also demonstrated endurance of more than 108 cycles with a satisfactory Roff/Ron ratio and retention greater than 104 s.
2 June 2016 in IEEE Transactions on Emerging Topics in Computing
Currently the semiconductor industry is in search of a Physically-Unclonable-Function (PUF) implementation, which combines high reliability and uniqueness with low area and power consumption. The characteristics of emerging nanoscale Resistive Random Access Memory (ReRAM) devices fulfill most of these properties, as they exhibit inherent variability with low area consumption. Of particular interest is that the resistive states of ReRAM devices show a strong dependence on the distribution of grain boundaries within the device, which leads to variability in total device resistance. In this work we transform the classic CMOS time-delay PUF (TD-PUF) utilizing integrated nanoscale ReRAM devices to achieve better performance metrics including uniqueness and reliabilitiiy. The enhanced design exploits the property of high resistance variability of ReRAMs for the design of a ReRAM based delay stage that exhibits excellent uniqueness. Accurate simulation and characterization of the proposed PUF was achieved by extracting resistance values, temperature dependence and usage stress of ReRAM devices fabricated in-house and their application in the proposed TD-PUF are discussed. A 24 stage time-delay PUF utilizing 48 ReRAM devices was simulated and results show excellent reliability with respect to environmental parameters. A temperature range of 0 to 125°C was simulated and an optimum reliability was observed at 0.79 V. A supply voltage noise of ±30 mV had no impact on the uniqueness and reliability. The proposed design was compared against two pure CMOS implementations of a TD-PUF. The comparison was performed with respect to the aforementioned metrics and under the same environmental conditions, showing up to 5 times increase in performance.
31 October 2022 in IEEE Access
The predictive capability of existing physical descriptions of multi-state devices ( e.g ., oxide memristors, ferroelectrics, antiferroelectric, etc.) cannot be fully leveraged in circuit simulations due to practical limitations regarding the complexity of compact models. We attempt to circumvent this issue by adopting a machine-learning (ML) - based approach to develop a compact model that retains the full physical description of these devices. ML-based modeling approaches have garnered immense interest in recent years and have already been successfully utilized in making models for several novel devices. A known hurdle for ML-based compact modeling is the need for large amount of experimental data to properly train the model. We propose a method to simulate additional data by duplicating the data and adding Gaussian Noise to the duplicates. We propose a generalized framework to - (i) facilitate efficient training of ML-based device models, (ii) conduct seamless conversion to Verilog-A model, and (iii) interface with industry-standard circuit simulators (HSPICE, SPECTRE, etc.). We demonstrate the capabilities of our framework using the hafnium oxide (HfOx) memristor as a test device. As the source of the training data, we use a physical model that unifies detailed atomic-level descriptions with self-consistent evaluation of electronic transport. In addition, we test our model with experimental data for multiple memristor samples and repeated cycles of the same sample. Our ML-based framework prepares a circuit-compatible compact model to facilitate system-level simulations. With our model, we achieve a root mean squared error (RMSE) of 0.000863 and an R2 of 0.977371 on our testing data.