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

transition energies. In fact, these emission dynamics suggest a QD intimatelycoupled to and reacting to a fluctuating environment. Through the concurrentmeasure of spectral diffusion and fluorescence intermittency, we examine theextent of this influence from the QD environment and observed an unexpectedrelationship between spectral diffusion and blinking. Zooming into the timetraces of Figs. 4b and 4c reveals a surprising correlation between blinking andspectral shifting. As shown in Fig. 4d, magnifying the marked region in thetime trace of Fig. 4b reveals a pronounced correlation between individualspectral jumps and blinking: following a blink-off period, the blink-on event isaccompanied by a shift in the emission energy. Furthermore, as shown inFig. 4e, zooming into the time trace of Fig. 4c reveals a similar correlation. Asin Fig. 4d, the trace shows dark periods that are accompanied by discontinuousjumps in the emission frequency. The periods between shifts in Figs. 4dand 4e, however, differ by nearly an order of magnitude in timescales. Due toour limited time resolution, no blinking events shorter than 100 ms can bedetected. Any fluorescence change that is faster than the ‘blink-and-shift’event shown in Fig. 4e is not resolved by our apparatus and appears in astatistical analysis as a large frequency shift during an apparent on-timeperiod. This limitation weakens the experimentally observed correlationbetween blinking and frequency shifts. Nevertheless, a statistically measurabledifference between shifts following on and off times can be extracted fromour results.Because changes in the emitting state cannot be observed when the QDis off, we compare the net shifts in the spectral positions between the initialand final emission frequency of each on and off event. The histogram of netspectral shifts during the on times, shown in Fig. 5a, reveals a nearlyGaussian distribution (dark line) with a 3.8-meV full width at half-maximum.However, the histogram for the off-time spectral shifts in Fig. 5b showsa distribution better described as a sum of two distributions: a Gaussiandistribution of small shifts and a distribution of large spectral shifts locatedin the tails of the Gaussian profile. To illustrate the difference between thedistributions of on- and off-time spectral shifts, the on-time spectral distributionis subtracted from the off-time spectral distribution shown in Fig. 5c.Even with our limited time resolution, this difference histogram shows thatlarge spectral shifts occur significantly more often during off times (longerthan 100 ms) than during on times; hence, large spectral shifts are more likelyto accompany a blink-off event than during the time the QD is on. Thisstatistical treatment does not try to assess the distribution of QDs that showthis correlation but rather confirms the strong correlation between spectralshifting and blinking events in the QDs observed. The off-time histogram,plotted on a logarithmic scale in Fig. 5d, shows that a single GaussianCopyright 2004 by Marcel Dekker, Inc. All Rights Reserved.