Microsecond Blinking Events in the Fluorescence of Colloidal Quantum Dots Revealed by Correlation Analysis on Preselected Photons

Abstract

Nearly all colloidal quantum dots, when measured at the single-emitter level, exhibit fluorescence "blinking". However, despite over 20 years of research on this phenomenon, its microscopic origins are still debated. One reason is a gap in available experimental information, specifically for dynamics at short (submillisecond) time scales. Here, we use photon-correlation analysis to investigate microsecond blinking events in individual quantum dots. While the strongly distributed kinetics of blinking normally makes such events difficult to study, we show that they can be analyzed by excluding photons emitted during long bright or dark periods. Moreover, we find that submillisecond blinking events are more common than one might expect from extrapolating the power-law blinking statistics observed on longer (millisecond) time scales. This result provides important experimental data for developing a microscopic understanding of blinking. More generally, our method offers a simple strategy for analyzing microsecond switching dynamics in the fluorescence of quantum emitters.

Figure 1

(a) Emission intensity from an individual CdSe/CdS/ZnS core/shell/shell QD binned with a time resolution of 10 ms under pulsed excitation (405 nm; 10 MHz; 1 μJ cm^–2). The QD core diameter is 3.2 nm, and the shell has nominally 8 (2) monolayers of CdS (ZnS). (b) Corresponding “fluorescence lifetime intensity distribution” (FLID), a two-dimensional histogram of the emission intensities and fitted fluorescence lifetimes. The diagram is based on a 300 s experiment divided into 30 000 10 ms time bins (see Figure S1 in the Supporting Information for details). The effect of flickering is highlighted. (c) Corresponding one-dimensional intensity histogram, with the threshold used for statistical analysis indicated as a red line. (d) Blinking statistics for the ON periods (blue) and OFF periods (red), extracted from 10 ms binning and thresholding for the data in panel a. (e–h) Same as panels a–d but obtained using 3 ms time bins on the same single-photon data.

Figure 2

(a) Intensity correlation function g⁽²⁾ constructed from the same experiment as analyzed in Figure 1. Antibunching with g⁽²⁾(t = 0) < 0.5 proves that we are studying a single emitter. (b) The same intensity correlation function rebinned to integer numbers of 100 ns laser repetition periods (red data points), plotted with a logarithmic time axis. Photon pairs with positive (detector 1 clicks before detector 2) and negative (vice versa) delay times have been averaged, and both are plotted at positive delay times. The blue line shows the first 500 ns at high resolution (same as in panel a). (c) Schematic (without Poisson noise) of the situation of a typical blinking experiment, in which the duration of ON periods varies widely. The correlation function is basically a histogram of time differences between photon pairs, so it is dominated by the long periods. In this example, the 100 ms ON period contributes more photon pairs than the 100 μs ON period by a factor of 10⁶.

Figure 3

(a) Emission intensity trace of the QD (reproduced from Figure 1a) with a range of intensities highlighted in yellow (120–160 counts/10 ms) that are likely due to flickering, i.e. ON–OFF switching on time scales faster than the 10 ms binning used. (b) Intensity histogram with different intensity ranges highlighted that are used to preselect photons for correlation analysis. (c–g) Correlation functions g⁽²⁾ constructed from preselected photons emitted during 10 ms time bins in which 76–124 (red; c), 124–160 (yellow; d), 160–220 (green; e), 220–280 (blue; f), or 348–372 (purple; g) photons were recorded. (h) Maximum bunching amplitude, i.e., g⁽²⁾(1–20 μs), as a function of the condition for emission intensity that was used to preselect photons for the analysis.

Figure 4

(a) Schematic of an emission intensity trace from an individual QD without Poisson noise. Vertical dashed lines indicate the time binning used. A flickering event is highlighted in yellow, i.e. blinking to an ON period shorter than the experimental binning. (b) Calculated correlation function g⁽²⁾ for time bins that exhibit flickering such as that schematically shown in panel a (black line) compared to the experiment (data points, reproduced from Figure 3d). The correlation function is averaged over the timing of the short ON period with respect to the beginning of the time bin (details in Figure S4). The arrow highlights a deviation in the slope of the correlation function between the calculation and the experiment. (c) Blinking statistics assumed in our Monte Carlo model for single-QD emission for ON (blue) and OFF (red dashed line) periods, extrapolating the power-law experimental statistics obtained from binning and thresholding (data points) and cutting off at t = 1 μs at the short-time-scale end and at t = 100 ms (OFF) or t = 1 s (ON) at the long-time-scale end. See the Supporting Information for the Monte Carlo modeling procedure and more simulated correlation functions (Figure S5). (d) Part of a simulated intensity trace. (e) Correlation function of preselected photons from the simulation (black line), treated in the same way as the experimental data (points). The arrow highlights a deviation in the slope of the correlation function between the calculation and the experiment. (f–h) Same as panels c–e, but assuming that the ON and OFF statistics are described by a power-law dependence with steeper slope for 1 μs < t < 1 ms than for t > 1 ms. More precisely, the power-law exponents are p_ON1 = 1.7 and p_OFF1 = 1.6 for 1 μs < t < 1 ms and p_ON2 = 1.2 and p_OFF2 = 1.1 for t > 1 ms.