Neuromorphic networks that emulate brain plasticity and adaptive learning will revolutionize artificial intelligence and computing. The key element for neuromorphic networks is the artificial synapse, which emulates the function of its biological counterpart using strictly solid-state components. Necessary attributes of an artificial synapse are non-volatile memory and history-dependent analogue states. These synapse characteristics can be realized using two-terminal memristive devices. However, two-terminal memristors cannot readily utilize existing computational algorithms designed for CMOS circuits, thus requiring development of new algorithms. Recently, three-terminal, transistor-like memristive devices utilizing an ionic liquid gate to control source-drain resistance by injection or extraction of oxygen vacancies have been demonstrated at Harvard and IBM. This functionality enables the implementation of spike time dependent plasticity (STDP) by converting the time difference between source (pre-neuron) and drain (post-neuron) spikes into a gate voltage. The exact channel conductance state achieved through a set of voltage pulses determines the synapse weight. The Harvard and IBM approach, however, has two principal drawbacks: 1) it requires a liquid gate to effectively inject or withdraw oxygen vacancies 2) switching is slow (~seconds) even at 160 ^oC due to the sluggish motion of oxygen ions in solid lattices at or near room temperature. To address these shortcomings, we describe an alternative synaptic transistor based on insertion/extraction of Li-ions from a Li_xCoO₂ channel. Similarly to oxygen vacancies, Li-ions act as dopants, contributing a mobile electron or hole every Li-ion insertion or extraction, depending on the specific material. Compared to O-vacancies, however, Li-ions move as interstitials with much lower activation energy. Since Li diffusivity at room temperature is relatively high, the ionic liquid gate is replaced in our device by a solid Li electrolyte to realize all-solid state synaptic transistors.

The last decade has seen nearly exponential growth in the science and technology of two-dimensional materials. Beyond graphene, there is a huge variety of layered materials that range in properties from insulating to superconducting. Furthermore, heterogeneous stacking of 2D materials also allows for additional “dimensionality” for band structure engineering. In this talk, I will discuss recent breakthroughs in two-dimensional atomic layer synthesis and properties, including novel 2D heterostructures and novel 2D nitrides. Our recent works demonstrate that the properties of 2D materials, especially those grown via CVD, are extremely sensitive to the substrate choice. I will discuss substrate impact on 2D layer growth and properties, doping of 2D materials with magentic elements, selective area synthesis of 2D materials, and the first demonstration of 2D gallium nitride (2D-GaN). Our work and the work of our collaborators has lead to a better understanding of how substrate not only impacts 2D crystal quality, but also doping efficiency in 2D materials, and stabalization of nitrides at their quantum limit.