Hysteresis and time delay effects find important applications in devices that are explored for resistive switching and neuromorphic computation, such as halide perovskite and organic memristors and transistors for synapses and neurons. Impedance spectroscopy consists of the measurement of small signal ac impedance at fixed points of the operation curve. The frequency domain analysis of memristors and more generally, of conducting systems with memory features of some kind, provides essential information about the dynamic behaviour of the system. The impedance response of a memristor can be represented as a linear circuit made of resistances, capacitors, and inductors, with voltage-dependent elements. Here we show a classification of various manifestations of hysteresis by identifying common elements. The circuit enables a determination of the type of hysteresis in current-voltage curves under dynamic scans, and the transient response to a voltage step that causes the synaptic function in brain-like systems for neuromorphic computation.^1,2 The equivalent circuit properties also establish the criteria for a Hopf bifurcation that produces spiking of artificial neurons.

Memristors are two-terminal devices, where the resistance depends on previous current flow. This feature unites storage and computing capabilities in a single device, which might help address the von Neumann bottleneck of today’s computers. Furthermore, such in-memory computing and features like plasticity might enable simple realizations of neuromorphic computing.

Perovskites became interesting for memristors due to the hysteresis, they show in their current-voltage curve. Additionally, filamentary switching has been observed. Here, conductive nanofilaments are created, which can be reversibly ruptured and closed, turning the memristor on and off. These filaments are created either by defect ions in the perovskite or metal ions.

In this work, we report highest-performance and highly stable perovskite memristive switches (millions of cycles), whose switching behavior is further analyzed. This includes voltage-scan-rate and temperature dependent measurements to understand which parameters govern the values of SET and RESET voltage, as well as the switching dynamics. The effect of heat generation is measured by photoluminescence and thermography imaging and analyzed using a combined electrical and thermal model.

Neuromorphic devices can help perform memory-heavy tasks more efficientlydue to the co-localization of memory and computing. In biological systems,fast dynamics are necessary for rapid communication, while slow dynamicsaid in the amplification of signals over noise and regulatory processes such asadaptation- such dual dynamics are key for neuromorphic control systems.Halide perovskites exhibit much more complex phenomena than conventionalsemiconductors due to their coupled ionic, electronic, and optical propertieswhich result in modulatable drift, diffusion of ions, carriers, and radiativerecombination dynamics. This is exploited to engineer a dual-emitter tandemdevice with the requisite dual slow-fast dynamics. Here, a perovskite-organictandem light-emitting diode (LED) capable of modulating its emissionspectrum and intensity owing to the ion-mediated recombination zonemodulation between the green-emitting quasi-2D perovskite layer and thered-emitting organic layer is introduced. Frequency-dependent response andhigh dynamic range memory of emission intensity and spectra in a LED aredemonstrated. Utilizing the emissive read-out, image contrast enhancementas a neuromorphic pre-processing step to improve pattern recognitioncapabilities is illustrated. As proof of concept using the device’s slow-fastdynamics, an inhibition of the return mechanism is physically emulated

In order to reduce energy consumption, we aim to explore electronic materials beyond silicon for memristor and neuromorphic devices. Halide perovskites have emerged as a choice semiconductor due to their defect tolerance, low defect formation energies, compositional and optical bandgap tunability, and light/electric pulse-induced stimulation possibilities. [1,2] The resistance depends on the applied electrical signals, and this makes them potential candidates for future data storage and neuromorphic computing. The optoelectronic and carrier transport merits allow mimicking the characteristics of neurons and synaptic functions in the human brain. Methylammonium lead triiodide showed synaptic behaviors of photonic and electronic stimulations, and due to the intrinsic phase transition, limited efforts are made to study neuromorphic properties. We developed CsFAPbI3 microcrystals in grams quantity and studied CsFAPbI3-based memristive neuromorphic devices that can switch at low power and show larger endurance. [1] The fabricated memristors also showed an ultra-high paired-pulse facilitation index with applied electric stimuli pulse, and the short-term to long-term memory transition consumes ultra-low energy with long relaxation times. Our results suggest low-power neuromorphic devices that are synchronic to the human brain's performance with faster learning and memorization processes.

Keywords: Perovskite, Passivation, Surface modification, defect density.

Ion migration causes the degradation of perovskite solar cells. Ions are moved easily, with only a few hundred meV activation energy. Still, ions move many orders of magnitude more slowly than charges in metal halide perovskites. We use this difference in timescales to imprint memory in a resistive device. Because ions take very little energy to move, switching a memristive state in a perovskite device can also be very energy efficient. We show an artificial synapse that takes only a few hundred femtojoules to switch its resistive state[1]. This is achieved by downscaling the device to the micrometer scale. We use a novel back-contact architecture for these devices to avoid damage to the perovskite during lithography. We further discuss the working mechanism of these devices. Probably, the switching is achieved by filamentary formation. This mechanism would also allow the building of artificial neurons. With a memristive device and an artificial neuron, full hardware neural networks could be built. If time allows, I will also briefly discuss the implications of such filament formation on solar cell stability. We observe these filaments in lateral devices, and we see evidence for permanent, dramatic voltage-bias induced damage.

References: [1] Preprint: http://dx.doi.org/10.2139/ssrn.4592586