Emergent phenomena exploited at manganite heterointerfaces for unconventional computing

Classical computing based on von Neumann architecture are limited by a memory bottleneck, high power consumption and heat dissipation. This is primarily due to the non colocation of memory and processing units, making the hardware sequential and deterministic. These challenges can be efficiently tackled by processing the information on the signal, similar to what the human brain does using memristive devices, collocating memory and processing in the same functional unit. For this, a popular choice of devices are based on magnetic materials, particularly those possessing competing magnetic order and ground states. Complex oxide ferromagnets, offering a rich phase space are ideal candidates, where emergent phenomena arising at their heterointerfaces can be tailored by strain and doping and tuned by external stimuli such as temperature, electric field and magnetic field.

In this talk, I will discuss how we uniquely exploit the entire network of strained films of La0.67Sr0.33MnO3 by engineering the octahedron tilt at the atomic scale, to demonstrate complex biologically plausible brain functionalities such as self-oscillation and leaky-integrate and fire neurons. The combination of an intrinsic coupled phase transition in La0.67Sr0.33MnO3 and octahedral distortion due to the textured surface of the LAlO3 substrate leads to multiple negative differential resistance (NDR) regimes in the network, when electrically driven out of thermodynamic equilibrium, at room temperature. By sweeping either voltage or current at room temperature, we find that thermal effects trigger the film toward non-linear transport regimes. A physics-based Mott resistor network modeling accounts for these local phase distributions by incorporating a Gaussian distribution of critical temperatures. I will also discuss the demonstration of biologically plausible neuronal functionalities by engineering the time dynamics in the metastable phases in La0.67Sr0.33MnO3 . We design a voltage-tunable oscillator that dynamically oscillates at variable frequencies, relevant for application based on edge computing. Finally, I will discuss how the network responds as a leaky, integrate, and fire neuron and how its time dynamics leads to the development of probabilistic bits, useful for ultra-low power stochastic hardware