Abstract
Phase change memories have shown a greater impact on brain-inspired computing as we move closer to the artificial intelligence world. By mimicking the biological neurons and synapses, phase change memories emulated them efficiently. In this thesis, we investigated a Ge2Sb2Te5 phase-change memory for the artificial synapse behaviours in terms of efficacy and plasticity in milliseconds time interval and microseconds time interval. For the first time, we explored the synaptic efficacy in microseconds inter-pulse interval with paired-pulse facilitation and paired-pulse depression to show the faster synapse with reduced energy consumption (-50%). Synaptic plasticity was demonstrated by utilising the intermediate resistance states with different pulse trains in long-term plasticity. Further, we emulated a synaptic tunability with the application of various potentials and pulse durations. On synaptic learning, we delivered unsupervised learning based on Hebbian learning rules ranging from milliseconds to microseconds inter-spike interval with a slight increase applied potential (+0.2 V). These synaptic plasticity characteristics would pave the way for improvements in learning and computation. Additionally, electro-thermal simulations were studied by Joule-heating induced crystallization and amorphization to correlate our experimental data.