FLUORESCENCE IMAGING: Understanding fluorescence blinking is the first path to an imaging solution
FELIX VIETMEYER, SANDOR VOLKAN-KACSO, PAVEL FRANTSUZOV, MASARU KUNO, and BOLDIZSAR JANKO
The past 50 years have seen immense growth in the area of fluorescence spectroscopy. Starting with biochemistry and biophysics, the use of fluorescence techniques has spread rapidly and is now a versatile tool in a wide range of disciplines, ranging from bioimaging to materials characterization.
Fluorescence microscopy is a powerful tool for characterizing structural and dynamic processes on sub-micrometer scales. For instance, the localization of fluorophores with specific biological affinities can be used to visualize the internal structure of living cells. A major advantage of quantum-dot (QD) labels for long imaging sessions is their photostability.1 While the common dye AlexaFluor 488 (green) shows significant photobleaching over the course of 180 s, the QD 630-streptavidin (red) retains its full brightness under continuous illumination (see Fig. 1). Furthermore, the tunable optical properties of QDs as well as the variety of materials from which they can be composed allow for easy multiplexing along with complex labeling strategies where emission wavelengths can span the visible to the infrared.
In parallel, single-particle tracking experiments can characterize time-dependent diffusion constants of labeled species within cells to elucidate cellular transport mechanisms. Although the high brightness and photostability of QDs allows for long single-particle trajectories, these experiments also illustrate an intriguing single-molecule phenomenon: fluorescence intermittency or "blinking." Blinking occurs when a fluorophore's emission intensity shows seemingly random fluctuations between "on" and "off" states. This phenomenon is universal and has been seen in a wide variety of nanoscale emitters, including semiconductor nanowires, nanorods, QDs, and organic dye molecules. Despite being studied for the last 15 years, it remains a mystery to this day.
Fluorescence blinking—Why?
Blinking is an intriguing phenomenon, especially since it occurs during continuous excitation of the molecule. Even though quantum mechanics can explain some intensity fluctuations on short timescales (a phenomenon predicted by Niels Bohr in 1913 called quantum jumps), experimental observations reveal that fluorophores often remain off for periods far beyond quantum-mechanical timescales (seconds or minutes). Our group at the University of Notre Dame has been studying the experimental and theoretical successes that have recently begun to unravel the origin of blinking in single fluorophores such as colloidal QDs with an aim toward developing experimental techniques that can minimize this imaging obstacle.
Single-molecule fluorescence microscopy is a powerful tool that allows us to study blinking on individual emitters. While early experiments used cryogenic temperatures to improve experimental signal-to-noise ratios, advances in detector technology now enable room-temperature studies of even low-quantum-yield systems. In a general experimental scheme, laser radiation is focused through a high-numerical-aperture microscope objective to a diffraction-limited spot with typical diameters of 150–500 nm (see Fig. 2). The analyte, placed in this excitation region, absorbs the excitation illumination and re-emits light at a different frequency. The same microscope objective collects a fraction of this emission with typical system-wide collection efficiencies of approximately 1%. The detectors most commonly used in these measurements are single-photon-counting avalanche photodiodes (SPADs) and allow researchers to obtain single-fluorophore emission intensities as a function of time.
Blinking: What is known
First, the most striking feature of fluorescence intermittency is the absence of a unique timescale. There exists no characteristic rate for turning the fluorophore on or off. This is corroborated by threshold analyses that distinguish on versus off states in single QD emission trajectories. From histograms of resulting on or off times plotted on a semilogarithmic graph, it is immediately apparent that associated on/off probability densities are distributed over a wide range of timescales. In fact, the broad distributions of on and off times are better captured on log-log plots and exhibit linear behavior over 5 decades in time and 7 decades in probability density. This is remarkable given that they stem from the photophysics of a single fluorophore. Both on-times and off-times in QDs are therefore power-law distributed. Power laws are common in everyday life; for example, they describe things like the frequency with which words are used, the size of power outages, or the distribution of wealth.
While there is common agreement on the power-law nature of blinking, its physical origin is still in dispute. One clear conclusion, though, is that it is surface and environment related. In particular, recent studies have shown that blinking can (empirically) be stopped in QDs coated with a sufficiently thick layer of another semiconductor.2 Neuhauser and colleagues have also shown that a correlation exists between blinking and spectral diffusion—another ubiquitous phenomenon at the single-molecule level.3 This is important because common explanations for QD spectral diffusion involve its surface.
A number of models have been suggested to explain blinking.4 The first involves a QD charging and discharging mechanism. Namely, a charged (ionized) QD is thought to be nonemissive and emits only after returning to its neutral state. As a consequence, blinking can be recast as the visual observation of electron transfer events at the single-molecule level. While appealing, the model suffers from a few drawbacks. Specifically, while it readily explains switching events between on and off states, it does not predict power law behavior for either on-time or off-time probability densities. As a consequence, the model has been modified in various ways to yield power-law distributions. This involves introducing multiple acceptor states, their spectral diffusion, or even fluctuations in a tunneling barrier height or width. Alternate power law-predicting models exist that invoke the diffusion of ejected carriers or simply entail intrinsic fluctuations of nonradiative carrier recombination rates.
All models mentioned above rely on continuous space diffusion concepts to explain power-law blinking. More recent progress, however, shows that subsequent switching events are not uncorrelated (see Fig. 3).5 Namely, there are positive correlations between neighboring on times and similarly between successive off times. This means that a short on (off) time is likely followed by another short on (off) time. The absence of these correlations in earlier models has now motivated researchers to develop new theories that, while maintaining power-law distributed probability densities, naturally predict these correlations.
One such model is a multiple recombination center (MRC) model, which postulates that every QD has a few (approximately 10) nonradiative recombination centers.6 Physically, these could be deep traps or localized charge distributions on the QD surface. Then, if an electron-hole pair is excited by an external light source, it can recombine through these recombination centers without emitting a photon. The rate of this nonradiative recombination process (referred to as a "trapping rate"), associated with each center, can randomly change between a high and a low value—corresponding to an active and a passive state, respectively. Therefore, the recombination centers are effectively stochastic two-level systems where the total nonradiative rate is simply the sum of individual trapping rates. In turn, fluctuations of the total nonradiative rate lead to fluctuations in the fluorescence emission intensity, resulting in emission blinking. These individual nonradiative rates are distributed over timescales spanning several orders of magnitude and can interact, causing the appearance of correlations between subsequent on or off events.
A predictive model
The MRC model thus explains key features of the blinking process, such as 1) the (truncated) inverse power-law distribution of on- and off-times, 2) the strong threshold dependence of on- and off-time statistics, 3) the 1/f power spectrum of the blinking trajectories, and 4) correlations between subsequent switching events. At the moment, the MRC model is the only model that predicts such temporal correlations and therefore represents an important stepping stone in unraveling and ultimately controlling blinking.
Toward this end, by better identifying the physical nature of trap states invoked in the MRC model, we can eventually develop rational synthetic approaches that exclude their participation in relevant photophysical processes of the quantum dot. In practice, this could entail identifying more appropriate surface ligands for nanocrystal surface passivation or other chemical passivation schemes that negate blinking or, at the very least, exclude long timescale intermittency events that are a nuisance to imaging applications. Finally, the MRC model suggests a broader theoretical framework for understanding fluorescence intermittency in other systems such as fluorescent proteins and dye molecules.
REFERENCES
1. X. Wu et al., Nat. Biotech., 21, 41 (2003).
2. X. Wang et al., Nature, 459, 686 (2009).
3. R.G. Neuhauser et al., Phys. Rev. Lett., 85, 3301 (2000).
4. P.A. Frantsuzov et al., Nat. Phys., 4, 519 (2008).
5. F.D. Stefani et al., New J. Phys., 7, 197 (2005).
6. P.A. Frantsuzov et al., Phys. Rev. Lett., 103, 207402 (2009).
Felix Vietmeyer and Sandor Volkan-Kacso are research assistants, Pavel Frantsuzov is visiting research assistant professor, Masaru Kuno is associate professor, and Boldizsar Janko is professor and director of the Institute for Theoretical Sciences at the University of Notre Dame, Notre Dame, IN 46556; e-mail: [email protected]; www.nd.edu.