Artificial Intelligence (AI) places specific demands on the information processing and as the applications multiply, they require a dedicated hardware specifically designed for the task. Current transistor-based digital circuits are not well suited for AI and finding new materials with novel functionalities, enabling new device principles is critical to the progress in this field. This project aims to bring advanced knowledge of material properties evolution upon working (cycling) conditions for a specific type of devices, namely for switches based on tantalum oxide that can repeatedly change the electrical resistance between high and low values as the result of the applied bias. The applications of such structures include non-volatile solid state (no moving parts) memories and artificial synapses mimicking the functions of the human brain. The research emphasizes understanding of material properties and the processes controlling the resistance of such device. The parallel goal is to educate the next generation of engineers for the electronics industry workforce. The components of the plan are (i) reaching out to high school students through 'Engineering@CMU' program, (ii) involvement of undergraduates in research through Semiconductor Research Corporation - Undergraduate Research Opportunity program, and (iii) holding regular conference calls with companies such as Intel and IBM.
Resistive switching devices have been demonstrated more than a decade ago but many of the desired performance criteria have not been met. This is, in part, due to our poor understanding of the processes involved in resistance switching and, in particular, the formation and evolution of a conducting filament within highly resistive oxide layer. Resistive switches encode information by redistribution of ions both vertically between the functional oxide and the electrodes and laterally within the oxide film. The research aims to map out the ion motion in the oxide film as a function of device bias history semiconductor by recording two- and three-dimensional elemental maps by X-ray energy dispersive spectroscopy and electron and atom probe tomographies. Of particular interest is the filament evolution toward the end of device lifetime. The analysis of end-of-endurance material properties leads to identification and prediction of the device failure mechanisms. The experimental distribution maps are being simulated by the finite element model based on diffusion equations including motion of ions and transport of heat and electric charge with additional restrictions imposed by the electrical biasing circuit. The model is intended to be universal for the entire class of resistive switching devices.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
