Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

ferent samples while keeping the same overall experimental time. We calculateaverage on times of 312, 283, and 256 ms for the same CdSe/ZnS sample under10 K and 175 W/cm 2 , 10 K and 700 W/cm 2 and RT and 175 W/cm 2 excitationintensity, respectively. The effective truncation times (1.5, 4.6, and 71 s, respectively)can be extrapolated by determining the end point within thepower-law distribution that corresponds to the measured average on time.In Figs. 9c and 9d, the vertical lines correspond to this calculated averagetruncation point, indicating the crossover in the time domain from oneblinking mechanism to the other.Furthermore, we can understand the consequence of this secondarymechanism in terms of single-QD quantum efficiency. For ensemble systems,quantum efficiency is defined as the rate of photons emitted versus the photonsabsorbed. Figure 10a shows the changes to the single-QD time tracewith increasing excitation intensity: The intensity values at peak heightsincrease linearly with excitation power, but the frequency of the on–offtransitions also increases. Moreover, the measured time-averaged single-QD emission photon flux at different excitation intensities, marked by emptytriangles in Fig. 10b, clearly shows a saturation effect. This saturationbehavior is due to the secondary blinking-off process shown in Fig. 9. Thefilled triangles in Fig. 10b plot the expected time-averaged emitted photonflux at different excitation intensities calculated from a power-law blinkingdistribution and truncation values similar to those in Fig. 9. The similarity ofthe two saturation plots (triangles) in Fig. 10b demonstrates the significanceof the secondary mechanism to the overall fluorescence of the QD system.The filled circles represent the peak intensity of the single QD at each of theexcitation intensities.Modification of the surface morphology or excitation intensity showedno difference in the statistical nature of the off times or blinking-on process.The statistical data are consistent with previous work [12,22]; however, theseparation of the power-law statistics from truncation effects clearly demon-Figure 9 (a) Three single-QD on-time probability distributions at 10 K, 700 W/cm 2 . The arrows indicate the truncation point for the probability distribution foreach QD. (b) Four single-QD on-time probability distributions for CdSe(ZnS) QDsat RT, 100 W/cm 2 . (c) Average on-time probability distribution for 25-Å-radiusCdSe(ZnS) QD at 300 K and 175 W/cm 2 (E), 10 K and 700 W/cm 2 (5), and 10 Kand 175 W/cm 2 (n). The straight line is a best-fit line with exponent f 1.6. (d)Average on-time probability distribution for 15-Å-radius CdSe(ZnS) QD (E), 25-A˚ -radius CdSe(ZnS) QD (5) and 25-A˚ -radius CdSe QD (x) at RT, 100 W/cm 2 . Thestraight line here is a guide for the eye. The vertical lines correspond to truncationpoints where the power-law behavior is estimated to end.Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.