Abhishek A. Sharma

Oscillatory Neural Networks Based on TMO Nano-Oscillators and Multi-Level RRAM Cells

In massively parallel computational tasks, such as pattern recognition, conventional computing architectures have insufficient power efficiency for energy constrained environments. This has made alternative architectures, such as neuromorphic computing, increasingly attractive. Oscillatory neural networks (ONNs) are one promising architecture, but efficient hardware implementations have been limited by shortcomings in CMOS technology, specifically in the efficient implementation of oscillators and synaptic weights. The authors have recently demonstrated that metal-oxide based resistive switching (RRAM) structures can be engineered to create low-power, scalable, voltage-controlled oscillators that utilize inherent meta-stability in the device. This work proposes an RRAM-based ONN that couples oscillatory “neurons” through weighted “synapses” using oscillator phase as the state-variable. This paper demonstrates a robust architecture using only a few logic gates per neuron to implement phase initialization and locking of these oscillators, and demonstrate their capability to identify stored patterns from noisy inputs. Using measured characteristics of RRAMs as oscillators and programmable resistors, compact models are derived and used to simulate both an 8-neuron and 20-neuron network.

Transient characterization of the electroforming process in TiO2 based resistive switching devices

The transient electroforming process of TiO2-based resistive switching devices is investigated using a pulsed voltage method, and the electroforming time is found to vary from 10−8 s to 10−1 s as function of pulse magnitude (3–8 V) and ambient temperature (25–100 °C). Pulsed experiments and thermal simulations reveal that Joule self-heating has a significant effect on the electroforming dynamics, specially for electroforming voltages above 5.5 V where there is little dependence on ambient temperature and the electroforming time (10–100 ns) is much shorter than the device thermal time constant (≈2 μs).

Phase Coupling and Control of Oxide-Based Oscillators for Neuromorphic Computing

Neuromorphic computing using neural network hardware has attracted significant interest as it promises improved performance at low power for data-intensive error-resilient graphical signal processing. Oscillatory neural networks (ONNs) use either frequency or phase as state variables to implement frequency-shift keying (FSK)- and phase-shift keying (PSK)-based neural networks, respectively. To make these ONNs power and area efficient, back-end-of-the-line compatible, and capable of processing multilevel information, we explore an emerging class of oscillators that show fine-grain frequency-tuning and phase-coupling. We examine TaOx- and TiOx-based oscillators (resistive random access memory-type) as elements of a neuromorphic compute block and experimentally demonstrate: 1) frequency control over four decades using a ballast MOSFET; 2) variable phase coupling between oscillators; and 3) variable phase programming between oscillators coupled with a MOSFET. Such fine-grain control over both frequency and relative phase serve as the desirable characteristics of oscillators required for multilevel information processing in star-type directly coupled FSK- and PSK-based neuromorphic systems that find applications in gray-scale image processing and other graphical compute paradigms. These attributes combined with the small size (<;1 μm2) and simplicity, make these devices attractive candidates for realizing large-scale neuromorphic systems at reasonable size and power.

Joule Heating-Induced Metal–Insulator Transition in Epitaxial VO2/TiO2 Devices

DC and pulse voltage-induced metal-insulator transition (MIT) in epitaxial VO2 two terminal devices were measured at various stage temperatures. The power needed to switch the device to the ON-state decrease linearly with increasing stage temperature, which can be explained by the Joule heating effect. During transient voltage induced MIT measurement, the incubation time varied across 6 orders of magnitude. Both DC I-V characteristic and incubation times calculated from the electrothermal simulations show good agreement with measured values, indicating Joule heating effect is the cause of MIT with no evidence of electronic effects. The width of the metallic filament in the ON-state of the device was extracted and simulated within the thermal model.

In Situ TEM Imaging of Defect Dynamics under Electrical Bias in Resistive Switching Rutile-TiO 2

In this study, in situ electrical biasing was combined with transmission electron microscopy (TEM) in order to study the formation and evolution of Wadsley defects and Magnéli phases during electrical biasing and resistive switching in titanium dioxide (TiO2). Resistive switching devices were fabricated from single-crystal rutile TiO2 substrates through focused ion beam milling and lift-out techniques. Defect evolution and phase transformations in rutile TiO2 were monitored by diffraction contrast imaging inside the TEM during electrical biasing. Reversible bipolar resistive switching behavior was observed in these single-crystal TiO2 devices. Biased induced reduction reactions created increased oxygen vacancy concentrations to such an extent that shear faults (Wadsley defects) and oxygen-deficient phases (Magnéli phases) formed over large volumes within the TiO2 TEM specimen. Nevertheless, the observed reversible formation/dissociation of Wadsley defects does not appear to correlate to resistive switching phenomena at these length scales. These defect zones were found to reversibly reconfigure in a manner consistent with charged oxygen vacancy migration responding to the applied bias polarity.

Thermometry of Filamentary RRAM Devices

Since thermal effects play a major role in filamentary RRAM devices, we compare the two localized thermometry methods developed for such devices. One method is based on short-pulsed measurements and the other on the measurement of minority-carrier injection from the filament into a semiconductor electrode by thermionic emission. We carried out and compared the measurements on the same functional oxide layer. Both methods indicate that the filament temperature is at least ~550 K during device operation. Furthermore, comparison between the measured thermal resistance and the thermal simulations of both techniques shows that under the conditions of low forming current compliance (~10 μA), the filament dimensions are below ~5 nm. We show that the thermionic emission method is useful for high-resistance (>100 kQ) devices operating at low-power conditions (<;10 μW), whereas the pulsed thermometry is more suitable for lower resistance devices (<;500 kQ) operated above 1 μW. The average thermal resistance measured by the pulse technique decreases with applied power. Our simulations indicate that the expansion of the heated zone surrounding the filament can explain the observed reduction in thermal resistance with applied power. The underlying physics of the two methods is discussed.

Mechanism of localized electrical conduction at the onset of electroforming in TiO2 based resistive switching devices

The onset of localized current conduction during electroforming of TiO2-based resistive switching devices is investigated using a pulsed voltage method. The temperature rise at electroforming onset is found to vary from 25 to 300 °C as the pulse amplitude and the width are varied between 3–8 V and 10 ns–100 ms, respectively. The effective activation energy of the forming event is strongly electric field dependent and decreases from 0.7 eV at 3 V to almost zero at 8 V. The functional form of this dependence points toward charge trapping as the mechanism rather than oxygen vacancy motion.

Transient Thermometry and High-Resolution Transmission Electron Microscopy Analysis of Filamentary Resistive Switches

We present data on the filament size and temperature distribution in Hf0.82Al0.18Ox-based Resistive Random Access Memory (RRAM) devices obtained by transient thermometry and high-resolution transmission electron microscopy (HRTEM). The thermometry shows that the temperature of the nonvolatile conducting filament can reach temperatures as high as 1600 K at the onset of RESET at voltage of 0.8 V and power of 40 μW. The size of the filament was estimated at about 1 nm in diameter. Hot filament increases the temperature of the surrounding high resistivity oxide, causing it to conduct and carry a significant fraction of the total current. The current spreading results in slowing down the filament temperature increase at higher power. The results of thermometry have been corroborated by HRTEM analysis of the as-fabricated and switched RRAM devices. The functional HfAlOx layer in as-fabricated devices is amorphous. In devices that were switched, we detected a small crystalline region of 10-15 nm in size. The crystallization temperature of the HfAlOx was determined to be 850 K in an independent annealing experiment. The size of the crystalline region agrees with thermal modeling based on the thermometry data. Scanning transmission electron microscopy (TEM) coordinated with electron energy loss spectroscopy could not detect changes in the chemical makeup of the filament.

Electro-Thermal Model of Threshold Switching in TaOx-Based Devices

Pulsed and quasi-static current-voltage (I-V) characteristics of threshold switching in TiN/TaOx/TiN crossbar devices were measured as a function of stage temperature (200-495 K) and oxygen flow during the deposition of TaOx. A comparison of the pulsed and quasi-static characteristics in the high resistance part of the I-V revealed that Joule self-heating significantly affected the current and was a likely source of negative differential resistance (NDR) and thermal runaway. The experimental quasi-static I-Vs were simulated using a finite element electro-thermal model that coupled current and heat flow and incorporated an external circuit with an appropriate load resistor. The simulation reproduced the experimental I-V including the OFF-state at low currents and the volatile NDR region. In the NDR region, the simulation predicted spontaneous current constriction forming a small-diameter hot conducting filament with a radius of 250 nm in a 6 μm diameter device.

Dynamics of electroforming in binary metal oxide-based resistive switching memory

The onset of localized conduction in TaOx- and TiOx-based resistive switching devices during forming has been characterized. The novel temperature and voltage dependencies of forming times were extracted with pulsed forming experiments that spanned five orders of magnitude in time and showed three different regimes of electroforming. A universal field-induced-nucleation theory which included self-heating effects was used to explain a strong reduction in forming voltage with increasing forming time over all observed regimes of electroforming. It was shown that the effective activation energy for the incubation time changes inversely proportional with the electric field. A diameter of the volatile filament that precedes forming was estimated at ∼1 nm.