Facts and artifacts in the blinking statistics of semiconductor nanocrystals

Since its initial discovery just over a decade ago, blinking of semiconductor nanocrystals has typically been described in terms of probability distributions for durations of bright, or "on," states and dark, or "off," states. These distributions are obtained by binning photon counts in order to construct a time series for emission intensity and then applying a threshold to distinguish on states from off states. By examining experimental data from CdSe/ZnS core/shell nanocrystals and by simulating this data according to a simple, two-state blinking model, we find that the apparent truncated power-law distributions of on times can depend significantly on the choices of binning time and threshold. For example, increasing the binning time by a factor of 10 can double the apparent truncation time and change the apparent power-law exponent by 30%, even though the binning time is only 3% of the truncation time. Our findings indicate that stringent experimental conditions are needed to accurately determine blinking-time probability distributions. Similar considerations should apply to any phenomenon characterized by time series data that displays telegraph noise.