## Abstract

We present a novel approach for convenient tuning of the local refractive index around nanostructures. We apply this technique to study the influence of the local refractive index on the radiative decay time of CdSe/ZnS quantum dots with three distinct emission wavelengths. The dependence of the luminescence decay time on the environment is well described by an effective medium approach. A critical distance of about 80 nm is found for the determination of the effective local index of refraction. An estimation for the emitting-state quantum efficiency can be extracted.

© 2012 OSA

## 1. Introduction

Tuning the photoluminescence of quantum emitters via modification of their local environment has attracted considerable interest in recent years. The optical properties of nanostructures can vary dramatically, as compared to macroscopic structures, due to the emergence of quantum confinement effects. Hence problems of interpretation can arise when optical phenomena are studied at the nanometer scale. The refractive index is particularly important in this context because it has a strong influence on the radiative transition probability of an emitter, as quantified by Fermi’s golden rule: The spontaneous emission rate *k*_{r} of a quantum system in an isotropic medium can be written as [1–3]

*τ*

_{r}is the radiative decay time,

*ω*

_{12}is the angular emission frequency,

*ρ*(

*ω*

_{12}) is the density (per unit energy) of field oscillators at frequency ω

_{12}, 𝓔

_{loc}is the local zero-point field at the position of the emitting center, and

*μ*

_{12}is the transition dipole moment between its ground and excited states. The refractive index of the surrounding medium affects both

*ρ*(

*ω*

_{12}) and 𝓔

_{loc}. An additional influence of the environment can result from optical coupling of the emitter to a cavity, which changes the density of field oscillators

*ρ*(

*ω*

_{12}). This effect can be used for emission enhancement [4] or for controlling the radiation pattern [5, 6], as well as to determine the quantum yield of a luminescent system at the level of single nano-objects [7]. The influence of the refractive index becomes more pronounced when fluorescence decay times or luminescence quantum yields of nanoparticles are considered. This environmental effect has been demonstrated for many types of emitters embedded in various media, for example molecules [8], semiconductor quantum dots [9, 10] or doped insulators [11–13]. If such experiments deal with nanostructures then it is commonly assumed that the bulk refractive index remains meaningful for the surrounding material at the micrometer scale, which implies that a cut-off distance for a “local” refractive index around the emitters can be established, which allows to define a sphere of interaction. We have previously explored this issue by means of an active nanolayer covered by passive nanolayers of varying thicknesses, and we thus extracted a radius of sensitivity between 100 nm and 150 nm for an emitter at a wavelength of 610 nm (Gd

_{2}O

_{3}:Eu

^{3+}) [14]. However, the applicability of this concept to different types of emitters as well as the wavelength dependence of this radius remain open questions. In this article, we therefore present the application of this approach to CdSe/ZnS quantum dots of three different sizes, which enables us to study the same emitting species at three distinct emission wavelengths. We furthermore explore the possibility to extract the quantum yield of the emitter by means of controlling the surrounding medium.

## 2. Experimental section

#### 2.1. Sample preparation

We used water-soluble CdSe/ZnS core/shell quantum dots (QDs, Invitrogen) in three different sizes, whose luminescence spectra are shown in Fig. 1(a). A dilute mixture of these QDs, with a concentration of *c* = 10^{−8} mol/l for each of the three species in the same aqueous solution, was spin-cast onto a suprasil substrate to obtain a film that contained all three sizes of QDs, so that they could be studied under identical conditions. The total luminescence intensity detected from an excitation focus of 500 nm diameter agrees with a monolayer coverage of the substrate with an average of 35–70 nanoparticles of each size in the focus, in accordance with the dilution series carried out to find the optimum concentration of QDs before spin-casting. The QD film was covered with a BK7 lens of focal length *f* = 12.5 mm, leading to a varying air gap between the substrate and the lens as illustrated in Fig. 1(b). The minimum surface-to-surface distance between lens and substrate, *d*_{min} in Fig. 1(b), was calculated from the diameter of the first green interference ring for white-light transmission and the known radius of curvature of the lens, (13.5 ± 0.3) mm; *d*_{min} was thus found to be (44 ± 7) nm. The diameter of the largest QD species in the sample is *d* = 18 nm (including shell and organic ligands) according to the manufacturer. We attribute the difference between this value and our measured *d*_{min} to the nonzero probability of occasionally finding an agglomerate with a height of two times the particle diameter.

#### 2.2. Confocal microscopy and fluorescence lifetime measurements

The substrate-lens construct was mounted on a home-built confocal microscope based on a Zeiss Axiovert 135 TV in combination with a feedback-controlled sample scanning stage (Physik Instrumente, E-710.3CD) with nanometer positioning accuracy. A pulsed laser diode operating at 473 nm and 10 MHz repetition rate (Picoquant GmbH, LDH-P-C-470) was used to excite the QDs, and an oil-immersion objective (Zeiss plan-neofluar, 100×, NA=1.25, focal spot size below 500 nm) served to both focus the excitation beam and to collect the QD luminescence. The position-dependent luminescence intensity and decay time were recorded by raster-scanning the sample though the excitation focus; selective collection of the luminescence of each individual QD species was assured by appropriate bandpass filters in the detection path. Time-correlated single-photon counting was performed by focusing the QD luminescence onto the active area of a single-photon counting avalanche photo diode (APD) (Perkin Elmer, SPCM 200) and analyzing the timing of the output pulses of the APD with fluorescence lifetime imaging electronics (Picoquant, Picoharp 300 and SymPhoTime software package). All measured luminescence decay curves were fitted with a mono-exponential function, *A* exp(–*t*/*τ*) + *B*, after deconvolution with the instrument response function.

## 3. Results and discussion

Our experimental geometry allowed straightforward selection of emitters that experience different effective refractive indices depending on their distance from the optical axis of the lens. The influence of the local refractive index was thus quantified by measuring the resulting changes of the QD luminescence decay time in a home-built confocal microscope with fluorescence lifetime imaging capabilities. Figure 1(c) shows the raw data of two luminescence decay curves measured for QD565. The curves are well described by a mono-exponential decay function; inhomogeneities and QD interactions obviously play a minor role due to the low substrate coverage and the averaging in the small diffraction-limited focal spot. One finds the expected increase of the luminescence lifetime with growing distance from the optical axis of the lens, which reflects the decreasing effective index of refraction caused by the widening air gap between substrate and lens. The evolution of the decay times for the three QD sizes as a function of the width *d* of the air gap is shown in Fig. 2, which represents the averaged results of twenty different line-scans across the center. As anticipated, one finds the same trend of lengthening luminescence decay time with widening air gap for all three samples, with a saturation of the effect occurring at a lens-substrate distance of about 100 nm, when the lifetime reaches its limiting value.

Before we apply the effective medium approximation, we have to address a possible source of complication for the interpretation of the data: The field mode density, and thus the radiative lifetime, can also be affected by coupling to a cavity, as has been mentioned in the introduction. We therefore have to ensure that such cavity effects are negligible compared to the role of the refractive index in our experimental geometry. Our samples are assembled with a substrate and a lens that are both uncoated, and the quality factor of a cavity formed by these low-reflectivity surfaces is small. The consequences of this fact are illustrated in Fig. 3, which compares the white-light transmission spectrum of such an uncoated glass cavity to the dramatically different response of a metallic microresonator that was used by Chizhik et al. [7] for a deliberate modification of single-molecule emission rates via the cavity effect. The transmission spectrum of the glass cavity does exhibit a single maximum resulting from constructive interference of the light inside the optical resonator, but its quality factor *Q* is almost 50 times lower than the one of a cavity with metallic mirrors. It is therefore reasonable to neglect the modification of the radiative rate that could arise from the electromagnetic mode structure of the low-quality resonator formed by the two glass surfaces.

To find the critical distance for the effective index of refraction, we apply the Bruggeman effective medium approximation, as described by Aspnes [15], to a sphere of interaction with radius *R* that is centered on the QD layer. The width of the air gap above the QD layer at a given position can be calculated from the known radius of curvature of the lens, which yields the volume fractions *f*_{SiO2}of silica and *f*_{BK7} of BK7 in the sphere of interaction. The effective refractive index *n̄* is then obtained as the positive real solution of the equation

*n*

_{SiO2}= 1.46,

*n*

_{BK7}=1.52 and

*n*

_{air}= 1 are the refractive indices of silica, BK7 and air, respectively. This solution for

*n̄*was determined numerically in our fitting procedure. The volume fractions

*f*

_{X}of silica and BK7 were calculated by modeling the space occupied by these materials as spherical caps of heights

*h*

_{X}, which have a volume of [16] where

*R*is the radius of the interaction sphere. Division by the total volume of the interaction sphere, $V=\frac{4}{3}\pi {R}^{3}$, yields the volume fraction

*f*

_{X}=

*V*

_{SC}/

*V*as

*h*

_{X}of the spherical caps are given by

*d*outside the sphere of interaction, i. e., for

*d*>

*R*+

*d*

_{min}/2, the height

*h*

_{BK7}was set to zero. The effective index of refraction

*n̄*for a given distance

*d*thus depends on the radius

*R*and on

*d*

_{min}, the minimum distance between the lens and the substrate, which was kept fixed at

*d*

_{min}= 44 nm (see experimental section) in all our fits. Given an interaction sphere with a radius

*R*around 100 nm, this means that our experimental geometry explored a range of effective refractive indices

*n̄*between 1.15 and 1.32.

Our procedure for calculating the effective refractive index does not take into account the presence of the nanocrystals themselves, which contain CdSe/ZnS core/shell structures with a high refractive index, each covered by a polymer hull. We now want to show that the volume fraction of CdSe/ZnS on the substrate is indeed negligible: According to the manufacturer, the 565 nm QDs have a spherical CdSe/ZnS core/shell center with a diameter of 4.6 nm. The two other species in our sample contain core/shell structures in the shape of prolate ellipsoids, whose diameters (principal axes) are 4 nm and 9.4 nm for QD605, and 6 nm and 12 nm for QD655. All three species are covered by a low-index polymer, resulting in roughly spherical particles with overall diameters of 14, 16, and 18 nm, respectively. Given this geometry of the quantum dots, one can calculate the combined volume of CdSe/ZnS that is present on our substrate, which contains 35–70 particles of each species in an area with a diameter of 500 nm (the focal spot). When we compare this volume of CdSe/ZnS to the total volume of a film with a thickness of 18 nm (the diameter of the largest particles), we find that the high-index volume fraction remains below 1 %; it is therefore an acceptable approximation to use *n*_{air} for the entire space between lens and substrate as we have done for the calculation of *n̄* outlined above.

Several approaches have been proposed to describe the influence of the refractive index on the radiative decay rate: Macroscopic observable properties such as electric field strength and polarization density of the medium can be related to the local field at the position of the emitter by introducing a cavity around the emitter to establish a (conceptual) boundary between its local environment and the bulk of the medium, which is treated as a classical dielectric. Two limiting cases for setting up this cavity are the virtual cavity (VC) model [17], which assumes that the cavity is completely filled with the host dielectric, and the the empty cavity (EC) model [18], in which the cavity only contains the emitter. An alternative approach is the fully microscopic (FM) model [19] for local-field effects, which treats the atoms of the dielectric host as two-level systems and quantifies their interaction with the emitter at the microscopic level. The expressions for the dependence of the radiative depopulation rate *k*_{r} on the effective refractive index *n̄* resulting from these three models are

*k*

_{rv}=

*k*

_{r}(

*n*= 1) is the radiative decay rate in vacuum. The predicted increase of

*k*

_{r}with increasing refractive index is most pronounced for the VC model, less strong for the EC model and even weaker for the FM model. All these expressions assume an isotropic medium for which there is no preferred direction for the local zero-point field 𝓔

_{loc}; this justifies the form of the golden rule in Eq. (1), where the average 〈|𝓔

_{loc}··

*μ*

_{12}|

^{2}〉

_{osc}over all local field oscillators [3] has been replaced by $\frac{1}{3}$|𝓔

_{loc}|

^{2}|

*μ*

_{12}|

^{2}. Our experimental geometry is not isotropic due to the presence of the silica-air and the air-BK7 interface in the sphere of interaction. However, we can anticipate 35–70 randomly-oriented nanoparticles of each size in the excitation focus contributing to our overall decay curves, so that a sufficient amount of orientational averaging takes place to justify the application of the above formulas to our data.

The overall luminescence decay constant *k* is given by the sum of the radiative and non-radiative decay rates, i. e., *k* = *k*_{r} + *k*_{nr}, where *k*_{r} depends on *n̄* according to Eq. (6). The non-radiative rate *k*_{nr} is assumed to be dominated by internal radiationless relaxation processes of the quantum dots, which arise from electron-phonon coupling. The high-quality core-shell structure of the QDs is specifically designed to isolate the exciton in order to suppress energy transfer to the environment and to maximize luminescence emission. We can therefore expect *k*_{nr} to remain unaffected by changes in the local environment, hence we introduce it as an *n̄*-independent parameter. We applied each one of the models to our data to extract *R*, *τ*_{rv} = 1/*k*_{rv}, and *τ*_{nr} = 1/*k*_{nr} as fit parameters for the three QD sizes. The resulting best-fit curves have been included in Fig. 2 and the corresponding sets of parameters are given in Table 1, together with the resulting quantum efficiency Φ = *k*_{r}/(*k*_{r} + *k*_{nr}) of the emitting state.

As can be seen in Fig. 2 and Table 1, both the VC and the EC model can reproduce our data with values for the quantum yield Φ between 0.5 and 0.7. The FM model, on the other hand, predicts a much weaker dependence of *k*_{r} on *n̄* and thus cannot describe our data even for the maximum possible quantum efficiency of Φ = 1, which corresponds to *k*_{nr} = 0. A larger *n̄*-independent non-radiative contribution would further diminish the predicted relative change in luminescence lifetime and thus further reduce the agreement of the FM model with our data. If we remove the *k*_{nr} ≥ 0 restriction from the fit algorithm, we find that quantum efficiencies Φ between 1.2 and 1.4 would be required to bring the curve of the FM model as close to our data as those of the other two models. (Such values of Φ > 1 in consequence of *k*_{nr} < 0 are of course physically meaningless; we only mention these values here as an indication of the extent of disagreement between our data and the FM model.)

Typical quantum efficiencies reported for colloidal semiconductor QDs in the literature are around 50% [10, 20], but it has been suggested that the standard cuvette measurement, which is based on a comparison of fluorescence intensities, may systematically underestimate the quantum yield of QDs due to their particular photophysical properties (long-lived dark states, co-existence of “bright” and “dark” sub-populations) [9,21]. On this basis, we conclude that the quantum efficiencies found by applying the EC or the VC model to our data all fall into a reasonable range. However, our efficiencies are lower than the values of more than 95% reported by Brokmann et al. [9], who used a classical-dipole model to interpret changes in luminescence lifetime due to the presence of a dielectric interface close to the emitter. Both our experiment and that of Brokmann et al. rely on lifetime measurements and should therefore be comparable, because such measurements are insensitive to the effects of blinking which complicate comparison with quantum efficiencies measured in the standard way. The lower quantum efficiencies that we find may be caused by a lower structural quality of our nanocrystals and/or they may reflect fundamental differences between the dipole-interface approach and our effective refractive index method.

The most common application of the EC, VC, and FM models to quantum dots relates to measurements of the fluorescence lifetime in different solvents. In this context, we find agreement with the work of Duan et al. [20], who conclude that the the VC model is compatible with their measurements and quantum efficiencies between 55 % and 90 %. Wuister et al. [10], on the other hand, show a clear preference for the FM model and they conclude that the VC and EC models both overestimate the influence of the solvent refractive index. It is important to note that Berman and Milonni [22] have recently criticized the FM model of Crenshaw and Bowden and have proposed an alternative microscopic theory of spontaneous emission in dielectrics. The main improvement introduced in the Berman-Milonni model is the consideration of the magnetic sublevels of the atoms in the dielectric, which is absolutely necessary to obtain an isotropic index of refraction in an isotropic medium [22]. Unfortunately the new microscopic theory only covers effects that are first-order in the density of the dielectric, at which level the VC and EC expressions are indistinguishable, and it does not directly yield a formula that can be applied to higher-density condensed dielectrics such as silica and BK7. Nevertheless, our results may provide a first experimental confirmation of the problems with the FM model that were pointed out by Berman and Milonni.

Regarding the wavelength dependence of the interaction sphere, our study points to a magnitude for *R* around 80 nm for both the EC and the VC model, and we cannot detect a clear tendency for the dependence of *R* on *λ* in the small wavelength range covered by our experiments.

## 4. Conclusion

We have presented a straightforward and efficient approach to tune the local dielectric environment of nanoparticles and study the resulting changes in their emission dynamics and thereby their quantum yield. We found that our data is well described by adopting an effective medium approach to determine a local effective index of refraction and then applying either the empty cavity or the virtual cavity model for the relation between refractive index and radiative lifetime. The FM model, however, for this relationship was found incapable to describe our experimental results. Our technique permits covering a wider range of effective refractive indices by replacing the BK7 lens with one made of a higher-index material – commercial lenses with *n* ≈ 2 are readily available, which could allow us to probe effective refractive indices of up to 1.6 – and furthermore it can be combined with advanced single-particle/molecule techniques such as full determination of the 3D orientation of the transition dipole and a detailed analysis of the emission pattern. As such, it can be expected to serve as a versatile tool to test and improve various models for the influence of the nanoscale environment on the radiative dynamics of different types of emitters, such as organic chromophores, semiconductor quantum dots, and doped-insulator nanoparticles.

## Acknowledgments

This work was conducted in the framework of the European collaboration Nanolum.

## References and links

**1. **E. Fermi, “Quantum theory of radiation,” Rev. Mod. Phys. **4**, 87–132 (1932). [CrossRef]

**2. **E. M. Purcell, “Spontaneous emission probabilities at radio frequencies,” Phys. Rev. **69**, 681–681 (1946).

**3. **D. Toptygin, “Effects of the solvent refractive index and its dispersion on the radiative decay rate and extinction coefficient of a fluorescent solute,” J. Fluoresc. **13**, 201–219 (2003). [CrossRef]

**4. **A. Chizhik, F. Schleifenbaum, R. Gutbrod, A. Chizhik, D. Khoptyar, A. J. Meixner, and J. Enderlein, “Tuning the fluorescence emission spectra of a single molecule with a variable optical subwavelength metal microcavity,” Phys. Rev. Lett. **102**, 073002 (2009). [CrossRef] [PubMed]

**5. **A. Belarouci, F. Menchini, H. Rigneault, B. Jacquier, R. M. Montereali, F. Somma, and P. Moretti, “Spontaneous emission properties of color centers based optical microcavities,” Opt. Commun. **189**, 281–287 (2001). [CrossRef]

**6. **J. Vučković, D. Fattal, C. Santori, and G. S. Solomon, “Enhanced single-photon emission from a quantum dot in a micropost microcavity,” Appl. Phys. Lett. **82**, 3596–3598 (2003). [CrossRef]

**7. **A. I. Chizhik, A. M. Chizhik, D. Khoptyar, S. Bär, A. J. Meixner, and J. Enderlein, “Probing the radiative transition of single molecules with a tunable microresonator,” Nano Lett. **11**, 1700–1703 (2011). [CrossRef] [PubMed]

**8. **G. Lamouche, P. Lavallard, and T. Gacoin, “Optical properties of dye molecules as a function of the surrounding dielectric medium,” Phys. Rev. A **59**, 4668–4674 (1999). [CrossRef]

**9. **X. Brokmann, L. Coolen, M. Dahan, and J. P. Hermier, “Measurement of the radiative and nonradiative decay rates of single cdse nanocrystals through a controlled modification of their spontaneous emission,” Phys. Rev. Lett. **93**, 107403 (2004). [CrossRef] [PubMed]

**10. **S. F. Wuister, C. D. Donega, and A. Meijerink, “Local-field effects on the spontaneous emission rate of cdte and cdse quantum dots in dielectric media,” J. Chem. Phys. **121**, 4310–4315 (2004). [CrossRef] [PubMed]

**11. **R. S. Meltzer, S. P. Feofilov, B. Tissue, and H. B. Yuan, “Dependence of fluorescence lifetimes of y2o3 : Eu3+ nanoparticles on the surrounding medium,” Phys. Rev. B **60**, 14012–14015 (1999). [CrossRef]

**12. **H. Schniepp and V. Sandoghdar, “Spontaneous emission of europium ions embedded in dielectric nanospheres,” Phys. Rev. Lett. **89**, 257403 (2002). [CrossRef] [PubMed]

**13. **D. W. Cooke, J. K. Lee, B. L. Bennett, J. R. Groves, L. G. Jacobsohn, E. A. McKigney, R. E. Muenchausen, M. Nastasi, K. E. Sickafus, M. Tang, J. A. Valdez, J. Y. Kim, and K. S. Hong, “Luminescent properties and reduced dimensional behavior of hydrothermally prepared y(2)sio(5): ce nanophosphors,” Appl. Phys. Lett. **88**, 103108 (2006). [CrossRef]

**14. **V. LeBihan, A. Pillonnet, D. Amans, G. Ledoux, O. Marty, and C. Dujardin, “Critical dimension where the macroscopic definition of refractive index can be applied at a nanometric scale,” Phys. Rev. B **78**, 113405 (2008). [CrossRef]

**15. **D. E. Aspnes, “Local-field effects and effective-medium theory - a microscopic perspective,” Am. J. Phys. **50**, 704–709 (1982). [CrossRef]

**16. **I. N. Bronshtein, K. A. Semendyayev, G. Müsiol, and H. Mühlig, *Handbook of Mathematics*, 5th ed. (Springer, 2007).

**17. **J. Knoester and S. Mukamel, “Intermolecular forces, spontaneous emission, and superradiance in a dielectric medium - polariton-mediated interactions,” Phys. Rev. A **40**, 7065–7080 (1989). [CrossRef] [PubMed]

**18. **R. J. Glauber and M. Lewenstein, “Quantum optics of dielectric media,” Phys. Rev. A **43**, 467–491 (1991). [CrossRef] [PubMed]

**19. **M. E. Crenshaw and C. M. Bowden, “Effects of local fields on spontaneous emission in dielectric media,” Phys. Rev. Lett. **85**, 1851–1854 (2000). [CrossRef] [PubMed]

**20. **C. K. Duan, M. F. Reid, and Z. Q. Wang, “Local field effects on the radiative lifetime of emitters in surrounding media: Virtual- or real-cavity model?” Phys. Lett. A **343**, 474–480 (2005). [CrossRef]

**21. **J. Yao, D. R. Larson, H. D. Vishwasrao, W. R. Zipfel, and W. W. Webb, “Blinking and nonradiant dark fraction of water-soluble quantum dots in aqueous solution,” Proc. Natl. Acad. Sci. U.S.A. **102**, 14284–14289 (2005). [CrossRef] [PubMed]

**22. **P. R. Berman and P. W. Milonni, “Microscopic theory of modified spontaneous emission in a dielectric,” Phys. Rev. Lett. **92**, 053601 (2004). [CrossRef] [PubMed]