## Abstract

In single particle spectroscopy, the degree of observed fluorescence anti-bunching in a second-order cross correlation experiment is indicative of its bi-exciton quantum yield and whether or not a particle is well isolated. Advances in quantum dot synthesis have produced single particles with bi-exciton quantum yields approaching unity. Consequently, this creates uncertainty as to whether a particle has a high bi-exciton quantum yield or if it exists as a cluster. We report on a time-gated anti-bunching technique capable of determining the relative contributions of both multi-exciton emission and clustering effects. In this way, we can now unambiguously determine if a particle is single. Additionally, this time-gated anti-bunching approach provides an accurate way for the determination of bi-exciton lifetime with minimal contribution from higher order multi-exciton states.

© 2013 OSA

## 1. Introduction

The second-order photon cross correlation function, g^{(2)}(τ), plays a critical role in characterizing the photon emission statistics of nanoscale light emitters including single molecules [1–3], quantum dots [4,5], carbon nanotubes [6], and diamond nanocrystals [7–9]. It is widely used to test the correlation of intensities on two different photon detectors in a Hanburry Brown and Twiss configuration [10]. In a pulsed laser experiment, it is well known that measuring a single quantum emitter will result in photon anti-bunching, which is signified by the disappearance of a peak at zero time delay. This arises from the fact that a single quantum emitter can only emit a single photon per excitation pulse, a property that is utilized in single molecule spectroscopy to prove the isolation of a single quantum emitter [1–13]. In fact, the area ratio between center peak and lateral or side peaks, *R*, has been used to determine the number of *m* identical emitters under illumination: *R = (m-1)/m* [10–12].

In some emitters, such as quantum dots, optical generation of multi-excitons can lead to the emission of multiple photons in a single excitation cycle via a quantum cascade process [14–16]. When the emission of these multi-excitons are not spectrally excluded in a g^{(2)} measurement, they contribute to the zero time delay peak. Thus, it has been shown that for a single, well isolated, quantum dot in the low pump intensity limit, *R* is directly proportional to the ratio between the bi-exciton quantum yield, Q_{BX}, and single exciton quantum yield, Q_{X} i.e. *R =* Q_{BX}/Q_{X} [13,17].

Recently, much effort has been devoted to the study of nanocrystal quantum dots (NQDs) with large bi-exciton quantum yields [17–19]. NQDs with high bi-exciton quantum yield are important for light amplification [20], multi-exciton generation [21,22], and photon pair source applications [23,24]. As Q_{BX} plays a critical role in understanding the relationship between NQD PL blinking [25,26] and the suppression of Auger recombination, determining Q_{BX} is an important part of this puzzle [17,18,27–29].

Importantly, g^{(2)} based Q_{BX} measurements hinge upon the complete isolation of individual NQDs. However, when the combined effects of both clustering and multi-exciton emission are considered together, the ratio *R* can be expressed as:

Here we assume that all *m* NQDs are identical: having equal excitation probabilities and identical Q_{X} and Q_{BX} values. Therefore, when measuring large Q_{BX} values, or rather, *R* > 0.5, an independent measure of NQD isolation becomes necessary [30]. Independent verification of NQD isolation might include high-resolution techniques such as AFM, SEM, or TEM. However, such techniques have limited utility in that they require additional specialized instrumentation, they are time consuming, and it is difficult to perform these high resolution microscopies and single dot spectroscopy on the same set of NQDs. Most importantly, these techniques are not appropriate for many types of samples such as NQDs embedded in polymer films or biological tissues. In this regard, an all-optical approach capable of disentangling the contributions of clustering and multi-exciton emission is essential to interpret g^{(2)} measurements without ambiguity.

We now demonstrate that the application of an appropriate time gate to a g^{(2)} measurement allows for the separation of various factors contributing to *R*. Specifically, this technique allows for the determination of the extent of clustering (*m*), as well as the Q_{BX}/Q_{X} ratio. Furthermore, by conducting a detailed analysis on the decay functionality of *R* with respect to the gate delay time (GDT), we determine general conditions required for effective utilization of this approach. Analyzing this decay functionality also yields a method for determining lifetimes of bi-exciton states (τ_{BX}). Existing approaches for measuring τ_{BX} involve fitting PL decays measured as a function of pump power with multi-exponential functions [31,32]. In contrast, our approach requires fewer fitting parameters, and more importantly, allows for the determination of τ_{BX} at low pump powers, where contributions from higher order multi-excitons are minimal. In this regard, this approach could serve as a valuable tool in investigating technologically important multi-exciton processes in NQDs, such as optical amplification and multiple exciton generation.

## 2. Time-gating principle

Our approach exploits the fact that the bi-exciton emission always precedes single exciton emission. We selectively detect the photon emissions of the single excitons through the use of a time gate (only photons arriving within the gate time are analyzed) where the beginning of the gate (the gate delay time – GDT) is set to be much longer than the lifetime of bi-excitons. The g^{(2)} function constructed out of these photons will be essentially void of all bi-exciton contributions, yielding the only the degree of clustering, with the familiar form of *R*:

We use the subscript TG to denote that time-gating has been applied. With the ability to remove all bi-exciton contributions, we are able to determine if a NQD is in fact well isolated without ambiguity, even if it has a large bi-exciton quantum yield. Furthermore, we can apply the degree of clustering, *m*, found by Eq. (2) into Eq. (1) to obtain the average Q_{BX} of the NQDs in the cluster as: ${Q}_{BX}=\left(R-{R}_{TG}\right)/\left(1-{R}_{TG}\right).$

Time-gating is a feature available on many Single Photon Avalanche Photodiodes (SPADs) where a digital input signal enables photon detection for a given time window (Fig. 1
). Thus, a hardware based time-gating approach is well-suited for a measurement where the goal is to identify single emitters. While this hardware implementation is simple, it requires a separate photon correlation experiment to acquire g^{(2)} at different GDT values. Since such experiments are time consuming it becomes impossible to study R_{TG} as a function GDT, a study required to validate this approach. For this reason, here we utilize a software-based time correlated single photon counting (TCSPC) approach. Specifically, we record the photon arrival times originating from multiple detectors with respect to both the start of the experiment (macro-time) and to a common excitation sync pulse (micro-time) as depicted in Fig. 1(c). This allows for the simultaneous collection of both the PL decay dynamics and g^{(2)} function. Importantly, this approach allows for selective analysis of only photons that arrive after a certain time delay following the sync pulse, which enables the construction of a time gated g^{(2)} function for any arbitrary GDT value.

## 3. Results and discussion

We applied this software-based time-gating approach to CdSe/CdS core thick shell NQDs (a.k.a giant NQDs), which exhibit nonblinking PL with minimal residual intensity flickering [17,27,30,33]. In Fig. 2
, we display the PL decay dynamics, g^{(2)} before the time gating, decay of R_{TG} as the function of GDT, and g^{(2)} after application of GDTs for two different giant-NQDs (g-NQDs). While the observation that R is > 0.5 in both g-NQDs suggests that these g-NQDs have very high Q_{BX}, this measurement alone cannot rule out the possibility of NQD clusters. However, the decay of R_{TG} towards zero with the increase of GDT (Figs. 2(c) and 2(g)) and observation of near complete photon anti-bunching after time gating at 75 ns (Figs. 2(d) and 2(h)) provide clear evidence that these two NQDs are well isolated and *R* values in each case indeed provide a direct measure of the Q_{BX}.

However, R_{TG} values do not decay completely to zero as they should for single quantum emitters (cf. Figures 2(d) and 2(h)). Our analysis revealed that this residual value originates mainly from the sharp spike at zero time delay, which is composed of two peaks separated to one another by a time delay of ~30 ns. This doublet is the characteristic feature of cross-talk between the two SPADs of our HBT spectrometer. In general, several other factors such as the presence of dark-counts, background noise, and a pulse period that does not allow for complete NQD relaxation can also contribute to this residual value. The total contribution of these experimental artifacts together [10–13] can add error in the range of 2%-10% in determination of center peak area in our measurement.

Figure 2 further shows that R_{TG} for both g-NQDs decays exponentially with GDT (black dashed lines). While the R_{TG} of the first g-NQD (Fig. 2(c)), decays with a time constant of 17.5 ns, that of the second g-NQD, which possesses a higher Q_{BX} value (0.73 compared to 0.5) requires a longer decay time of 32.4 ns. This slow decay of R_{TG} demands longer GDT to reject all bi-exciton emission contributions in this measurement. The rate of decay of R_{TG} as a function of GDT can be related to the lifetime or quantum yield of the bi-exciton state (i.e., τ_{BX} or Q_{BX}). We calculate that the decay should follow an exponential dependence:

_{X}is the lifetime of the single exciton. When it is assumed that the radiative decay rate of the bi-exciton scales up from that of the single exciton with an increase in the number of available recombination channels (i.e. Γ

_{Rad,BX}= 4 Γ

_{Rad,X}), we can express τ

_{BX}in terms of the bi-exciton quantum yield and single-exciton lifetime as${\tau}_{BX}=\left({\tau}_{X}{Q}_{BX}\right)/4$. Now, R

_{TG}can be expressed in terms of the bi-exciton quantum yield and single-exciton lifetime as:where$\beta ={Q}_{BX}/\left(4-2{Q}_{BX}\right)$. From this expression, we can further derive that a gate delay time of$GDT=\beta {\tau}_{X}\mathrm{ln}\left(100\right)$is required to remove 99% of all bi-exciton contributions$\left({R}_{TG}/R=0.01\right),$ meaning that g-NQDs with high Q

_{BX}require exceedingly long GDT values to reject all bi-exciton contributions.

Additionally, our analysis of the decay of R_{TG} provides another powerful way to extract the recombination time of bi-excitons. By substituting the value of τ_{X} and α attained from Figs. 2(a) and 2(c) in Eq. (5), we determine τ_{BX} = 13.4 ns. While this τ_{BX} value is extracted without requiring any assumptions, it is close to a 14.0 ns lifetime that can be calculated from τ_{X} = 112 ns and Q_{BX} = 0.5, assuming the statistical scaling of Γ_{Rad,BX}. The τ_{BX} value extracted for the second g-NQDs (Figs. 2(e)-(h) also exhibits a similar agreement: τ_{BX} = 22.9 ns as calculated using Eq. (5), vs. τ_{BX} = 28.5 ns as calculated from a statistical scaling model (τ_{X} = 156 ns and Q_{BX} = 0.73). Our time-gating experiment is performed at low pump power (i.e., the average exciton population per pulse (μ) is below 0.1). The nearly single exponential decay shown in Fig. 2(a) indicates that bi-exciton emission is minimal, yet we can extract τ_{BX} via R_{TG} versus GDT data. In this way, our time-gated g^{(2)} approach presents a better and more accurate measurement of τ_{BX} in single-particle studies, as this extracted τ_{BX} value is free of contributions from higher order multi-excitons [34].

Thus far, we have considered isolated NQDs and demonstrated the power of our approach in rejecting multi-excitonic contributions to g^{(2)} measurements. However, the significance of our approach becomes even more evident when it is applied to a cluster of NQDs. In this case, our approach not only allows the extraction of the average Q_{BX} and τ_{BX} of NQDs in the cluster, it also provides some insight to the nature of clustering. Figure 3
provides a clear demonstration of this point.

The conventional g^{(2)} measurement shown in Fig. 3(b) yields an *R* value of 0.73. While this value decays exponentially with GDT, the decay asymptotically approaches a value of 0.28. This provides clear evidence of clustering. A simple substitution of R_{TG} and *R* to Eq. (1) and Eq. (2) gives a value of m = 1.39 NQDs with an average Q_{BX} value of 0.63. Furthermore, by including the effects of clustering in the analysis for the decay of the R_{TG}, we extract the average τ_{BX} to be 6.2 ns.

Since a cluster of two identical dots will yield an R_{TG} value approaching 0.5, the value found here of R_{TG} = 0.28 (and thus m = 1.39) provides a clear indication that two (or more) non-identical dots are being examined. The cluster in this study could be composed of NQDs with unequal Q_{X} and/or unequal excitation probability. Specifically, for a cluster of 2 NQDs with equal excitation probability it can be shown that a difference in their Q_{X} value with a ratio of 1:0.2 can give rise to the R_{TG} of 0.28. On the other hand, for 2 NQDs with identical Q_{X}, a difference in excitation probability with the same ratio is necessary. The excitation probability difference can arise from a scenario in which 2 identical NQDs are separated by some distance and are excited by laser spot with Gaussian intensity distribution.

## 4. Summary

In summary, we have demonstrated a new method to successfully discriminate between clusters of similar particles and well isolated single emitters with a high bi-exciton quantum yield. Notably, this method has added utility by providing a means for estimating the size of a cluster, even if it contains particles having high Q_{BX} values, without the need for spectral rejection of multi-excitonic spectral lines. It can also be used as a powerful tool to determine the bi-exciton lifetime of a single particle with much less influence from higher order multi-excitons. Recently a post photon selection approach, in which first and second photon detection events in a g^{(2)} experiment are separated to extract the decay of BX is reported by Canneson et al. [33]. Although this approach can be considered as a more direct approach to extract BX decay, it is incapable of disentangling the contribution of multi-exciton emission and clustering in a g^{(2)} experiment. Furthermore, in the case of a NQD exhibiting PL intensity fluctuations, our approach can be extended to extract Q_{BX}, τ_{BX} and the degree of clustering for different emissive levels of g-NQDs that are often associated with different charge exciton states [27]. Additionally, while we have focused on colloidal quantum dot samples in this letter, this technique should be widely applicable to virtually any type of sample where single-particle analysis is relevant.

## 5. Methods

#### 5.1 Data collection

Quantum dots with 5.5 nm CdSe cores having nominally 15 monolayers of CdS grown using a SILAR method [35] were drop cast onto glass coverslips. Pulsed laser excitation (405 nm, 70 ps pulse-width) was used to elicit PL (centered at 650 nm) from the NQDs, which was detected by a pair of SPADs (Perkin Elmer, SPCM-AQR14) in a custom setup with a HBT detection scheme as described above. Photon arrivals were recorded with a HydraHarp 400 TCSPC module (PicoQuant). A 532 nm long pass filter was used to reject laser scatter and a 775 nm short pass filter was placed in front of each detector to minimize cross-talk. The laser power was reduced to ~200 pW and focused to a near diffraction limited spot (NA 1.3, 100x). Values for μ are estimated to be μ ~0.1.

#### 5.2 Data analysis

Time-gating was applied by removing all photons from the data stream having micro-time (*t)* values less than GDT as depicted in Fig. 1. Subsequently, g^{(2)} plots were created in the same fashion as is done without time gating: a histogram of time differences is created for photons arriving on different channels as depicted in Fig. 1(c).

## Acknowledgments

This work was conducted at the Center for Integrated Nanotechnologies (CINT), a U.S. Department of Energy (DOE), Office of Science (OS), Office of Basic Energy Sciences (OBES) user facility and nanoscale science research center. B.D.M and H.H. acknowledge a Single-Investigator Small-Group Research Award (2009LANL1096), OBES, OS, U.S. DOE. Y.G. is supported by Los Alamos National Laboratory Directed Research and Development Funds. J.A.H. acknowledges NIH-NIGMS Grant 1R01GM084702-01.

## References and links

**1. **F. Treussart, A. Clouqueur, C. Grossman, and J.-F. Roch, “Photon antibunching in the fluorescence of a single dye molecule embedded in a thin polymer film,” Opt. Lett. **26**(19), 1504–1506 (2001). [CrossRef] [PubMed]

**2. **T. Basché, W. E. Moerner, M. Orrit, and H. Talon, “Photon antibunching in the fluorescence of a single dye molecule trapped in a solid,” Phys. Rev. Lett. **69**(10), 1516–1519 (1992). [CrossRef] [PubMed]

**3. **B. Lounis and W. E. Moerner, “Single photons on demand from a single molecule at room temperature,” Nature **407**(6803), 491–493 (2000). [CrossRef] [PubMed]

**4. **P. Michler, A. Kiraz, C. Becher, W. V. Schoenfeld, P. M. Petroff, L. Zhang, E. Hu, and A. Imamoglu, “A quantum dot single-photon turnstile device,” Science **290**(5500), 2282–2285 (2000). [CrossRef] [PubMed]

**5. **P. Michler, A. Imamoglu, M. D. Mason, P. J. Carson, G. F. Strouse, and S. K. Buratto, “Quantum correlation among photons from a single quantum dot at room temperature,” Nature **406**(6799), 968–970 (2000). [CrossRef] [PubMed]

**6. **A. Högele, C. Galland, M. Winger, and A. Imamoğlu, “Photon antibunching in the photoluminescence spectra of a single carbon nanotube,” Phys. Rev. Lett. **100**(21), 217401 (2008). [CrossRef] [PubMed]

**7. **C. Kurtsiefer, S. Mayer, P. Zarda, and H. Weinfurter, “Stable solid-state source of single photons,” Phys. Rev. Lett. **85**(2), 290–293 (2000). [CrossRef] [PubMed]

**8. **A. Beveratos, R. Brouri, T. Gacoin, J.-P. Poizat, and P. Grangier, “Nonclassical radiation from diamond nanocrystals,” Phys. Rev. A **64**(6), 061802 (2001). [CrossRef]

**9. **R. Brouri, A. Beveratos, J.-P. Poizat, and P. Grangier, “Photon antibunching in the fluorescence of individual color centers in diamond,” Opt. Lett. **25**(17), 1294–1296 (2000). [CrossRef] [PubMed]

**10. **W. Becker, *Advanced Time-Correlated Single Photon Counting Techniques* Chemical Physics (Springer, 2005).

**11. **P. Tinnefeld, C. Müller, and M. Sauer, “Time-varying photon probability distribution of individual molecules at room temperature,” Chem. Phys. Lett. **345**(3-4), 252–258 (2001). [CrossRef]

**12. **K. D. Weston, M. Dyck, P. Tinnefeld, C. Müller, D. P. Herten, and M. Sauer, “Measuring the number of independent emitters in single-molecule fluorescence images and trajectories using coincident photons,” Anal. Chem. **74**(20), 5342–5349 (2002). [CrossRef] [PubMed]

**13. **G. Nair, J. Zhao, and M. G. Bawendi, “Biexciton quantum yield of single semiconductor nanocrystals from photon statistics,” Nano Lett. **11**(3), 1136–1140 (2011). [CrossRef] [PubMed]

**14. **E. Moreau, I. Robert, L. Manin, V. Thierry-Mieg, J. M. Gérard, and I. Abram, “Quantum cascade of photons in semiconductor quantum dots,” Phys. Rev. Lett. **87**(18), 183601 (2001). [CrossRef]

**15. **E. Dekel, D. V. Regelman, D. Gershoni, E. Ehrenfreund, W. V. Schoenfeld, and P. M. Petroff, “Cascade evolution and radiative recombination of quantum dot multiexcitons studied by time-resolved spectroscopy,” Phys. Rev. B **62**(16), 11038–11045 (2000). [CrossRef]

**16. **B. Fisher, J. M. Caruge, D. Zehnder, and M. Bawendi, “Room-temperature ordered photon emission from multiexciton states in single CdSe core-shell nanocrystals,” Phys. Rev. Lett. **94**(8), 087403 (2005). [CrossRef] [PubMed]

**17. **Y. S. Park, A. V. Malko, J. Vela, Y. Chen, Y. Ghosh, F. García-Santamaría, J. A. Hollingsworth, V. I. Klimov, and H. Htoon, “Near-unity quantum yields of biexciton emission from CdSe/CdS nanocrystals measured using single-particle spectroscopy,” Phys. Rev. Lett. **106**(18), 187401 (2011). [CrossRef] [PubMed]

**18. **Y. Louyer, L. Biadala, J. B. Trebbia, M. J. Fernée, P. Tamarat, and B. Lounis, “Efficient biexciton emission in elongated CdSe/ZnS nanocrystals,” Nano Lett. **11**(10), 4370–4375 (2011). [CrossRef] [PubMed]

**19. **R. Osovsky, D. Cheskis, V. Kloper, A. Sashchiuk, M. Kroner, and E. Lifshitz, “Continuous-wave pumping of multiexciton bands in the photoluminescence spectrum of a single CdTe-CdSe core-shell colloidal quantum dot,” Phys. Rev. Lett. **102**(19), 197401 (2009). [CrossRef] [PubMed]

**20. **V. I. Klimov, A. A. Mikhailovsky, S. Xu, A. Malko, J. A. Hollingsworth, C. A. Leatherdale, H.-J. Eisler, and M. G. Bawendi, “Optical gain and stimulated emission in nanocrystal quantum dots,” Science **290**(5490), 314–317 (2000). [CrossRef] [PubMed]

**21. **A. Shabaev, A. L. Efros, and A. J. Nozik, “Multiexciton generation by a single photon in nanocrystals,” Nano Lett. **6**(12), 2856–2863 (2006). [CrossRef] [PubMed]

**22. **R. D. Schaller and V. I. Klimov, “High efficiency carrier multiplication in PbSe nanocrystals: Implications for solar energy conversion,” Phys. Rev. Lett. **92**(18), 186601 (2004). [CrossRef] [PubMed]

**23. **A. Muller, W. Fang, J. Lawall, and G. S. Solomon, “Creating polarization-entangled photon pairs from a semiconductor quantum dot using the optical Stark effect,” Phys. Rev. Lett. **103**(21), 217402 (2009). [CrossRef] [PubMed]

**24. **R. M. Stevenson, R. J. Young, P. Atkinson, K. Cooper, D. A. Ritchie, and A. J. Shields, “A semiconductor source of triggered entangled photon pairs,” Nature **439**(7073), 179–182 (2006). [CrossRef] [PubMed]

**25. **M. Nirmal, B. O. Dabbousi, M. G. Bawendi, J. J. Macklin, J. K. Trautman, T. D. Harris, and L. E. Brus, “Fluorescence intermittency in single cadmium selenide nanocrystals,” Nature **383**(6603), 802–804 (1996). [CrossRef]

**26. **M. Kuno, D. P. Fromm, H. F. Hamann, A. Gallagher, and D. J. Nesbitt, “Nonexponential “blinking” kinetics of single CdSe quantum dots: A universal power law behavior,” J. Chem. Phys. **112**(7), 3117–3120 (2000). [CrossRef]

**27. **C. Galland, Y. Ghosh, A. Steinbrück, M. Sykora, J. A. Hollingsworth, V. I. Klimov, and H. Htoon, “Two types of luminescence blinking revealed by spectroelectrochemistry of single quantum dots,” Nature **479**(7372), 203–207 (2011). [CrossRef] [PubMed]

**28. **S. Jander, A. Kornowski, and H. Weller, “Energy transfer from CdSe/CdS nanorods to amorphous carbon,” Nano Lett. **11**(12), 5179–5183 (2011). [CrossRef] [PubMed]

**29. **J. Zhao, G. Nair, B. R. Fisher, and M. G. Bawendi, “Challenge to the charging model of semiconductor-nanocrystal fluorescence intermittency from off-state quantum yields and multiexciton blinking,” Phys. Rev. Lett. **104**(15), 157403 (2010). [CrossRef] [PubMed]

**30. **Y. Chen, J. Vela, H. Htoon, J. L. Casson, D. J. Werder, D. A. Bussian, V. I. Klimov, and J. A. Hollingsworth, ““Giant” multishell CdSe nanocrystal quantum dots with suppressed blinking,” J. Am. Chem. Soc. **130**(15), 5026–5027 (2008). [CrossRef] [PubMed]

**31. **V. I. Klimov, A. A. Mikhailovsky, D. W. McBranch, C. A. Leatherdale, and M. G. Bawendi, “Quantization of multiparticle Auger rates in semiconductor quantum dots,” Science **287**(5455), 1011–1013 (2000). [CrossRef] [PubMed]

**32. **F. García-Santamaría, Y. Chen, J. Vela, R. D. Schaller, J. A. Hollingsworth, and V. I. Klimov, “Suppressed Auger recombination in “giant” nanocrystals boosts optical gain performance,” Nano Lett. **9**(10), 3482–3488 (2009). [CrossRef] [PubMed]

**33. **D. Canneson, I. Mallek-Zouari, S. Buil, X. Quelin, C. Javaux, B. Dubertret, and J.-P. Hermier, “Enhancing the fluorescence of individual thick shell CdSe/CdS Nanocrystals by coupling to gold structures,” New J. Phys. **14**(6), 063035 (2012). [CrossRef]

**34. **Due to very high Q_{BX}, then second g-NQD shows a bi-exponential decay even at very low pump power. The fast time constant of the PL decay 23.78 ns is in good agreement with 22.9ns τ_{BX} extracted from the decay of R_{TG}.”

**35. **Y. Ghosh, B. D. Mangum, J. L. Casson, D. J. Williams, H. Htoon, and J. A. Hollingsworth, “New insights into the complexities of shell growth and the strong influence of particle volume in nonblinking “giant” core/shell nanocrystal quantum dots,” J. Am. Chem. Soc. **134**(23), 9634–9643 (2012). [CrossRef] [PubMed]