CMStatistics 2018: Start Registration
View Submission - CMStatistics
B0723
Title: Functional data analysis of resistive switching processes Authors:  Ana Maria Aguilera - University of Granada (Spain)
M Carmen Aguilera-Morillo - Universidad Carlos III de Madrid (Spain)
Juan Bautista Roldan - University of Granada (Spain)
Francisco Jimenez-Molinos - University of Granada (Spain)
Christian Acal - University of Granada (Spain) [presenting]
Abstract: The current flash memories have several problems that prevent a further reduction so that it is necessary a technological substitute. One of the strong candidates for future nonvolatile applications are Resistive Random Access Memories (RRAMs) due to their excellent properties of good scalability, long endurance, fast switching speed, and ease of integration in the CMOS processing. These devices have a simple physical structure: two metal plates acting as electrodes with a dielectric in between. The physical mechanisms behind resistive switching are stochastic and present important numerical problems for the correct modelling. The conduction takes place trough conductive filaments and a reasonable electric current is obtained under an applied voltage. The filaments are formed (set) and destroyed (reset) within the resistive switching device operation. The device resistance switches from a High Resistance State (HRS) to a Low Resistance State (LRS) so that the result is a sample of current-voltage curves corresponding to the reset/set cycles. Because of this, functional data analysis (FDA) methodologies are applied for modelling and explaining the variability associated with the stochastic process generating these curves. Data registration, P-spline approximation and Functional Principal Component Analysis (FPCA) are considered to provide a simple model that explains most variability in terms of one scalar variable highly correlated with the voltage to reset/set.