A new prediction method for ReRAM data retention statistics based on 3D filament structures

Instead of wide distributed resistance for a single bit, we introduce non-fluctuating physical parameters, filament diameter and packing factor (corresponding to oxygen vacancy concentration) to describe ReRAM bit. The quantitative 3D percolation model is developed based on direct observation of the filament structure and hopping conduction, which is confirmed with ultra-low temperature (30 K) measurement. Moreover, we provide a simulation method to obtain quantitative filament diameter, packing factor and to do the prediction of the resistance distribution after retention, which is verified with experiment.