Plasmonic antennas excited at resonance create highly enhanced local fields,1 which are key for surface-enhanced Raman scattering,2, 3 biosensing,4, 5 and nonlinear photoemission down to nanometer-scale volumes.6, 7 Plasmonic antennas also enable the control of the optical emission from single quantum emitters.8–18 However, a major challenge in coupling a single emitter to a plasmonic antenna is that the proximity of the quantum emitter to the metal results in energy transfer to the density fluctuations of the free electron gas that simultaneously enhances radiative emission and generates ohmic losses.8–10 Depending on the experiment configuration and the balance between radiative and nonradiative decay rates, either luminescence enhancement or quenching is reported.16–18 This apparent confusion arises as the luminescence signal combines modifications on excitation intensity, quantum yield, and collection efficiency. Moreover, there are only few reports experimentally quantifying the antenna's influence on excitation and emission.11, 14 New experimental methods are required to fully investigate the antenna response, and quantify separately the enhancement factors for excitation and emission.
Colloidal quantum dots (QDs) offer much wider possibilities than organic fluorophores for the investigation of the antenna-emitter coupling. The broad absorption spectrum of QDs can be used to probe the photoluminescence intensity on and off the antenna resonance to derive the excitation intensity enhancement.19, 20 Another approach takes advantage of sequential resonant photon absorption and multiply excited states occurring in QDs.21–23 We use this property here to quantify the local excitation intensity enhancement. The transient photoluminescence dynamics of the QDs contain at least two contributions, respectively from the singly and doubly (and higher) excited states. An important feature is that the ratio of doubly to singly excited state photoluminescence amplitudes quantifies the local excitation intensity independently on the emission process.24 Another advantage of this approach is that in the same measurement, the single exciton lifetime and intensity can be measured, from which the emission gain can also be deduced, providing the full information about the antenna's influence.
Here, we investigate the luminescence of a quantum dot deterministically coupled to a gold nanoantenna, and quantify the antenna's influence on the excitation intensity and the luminescence quantum yield separately. We report on monomer and dimer disk antennas with resonances close to the excitation and emission wavelengths. These antenna designs yield higher excitation enhancement but also higher quenching losses as compared to previous studies on polystyrene microspheres and subwavelength apertures.24 Thus, the separate investigation of excitation and emission processes is key to fully understand the antenna's influence on the QD luminescence.
Typical fabricated antennas are shown in Figure 1a. Each gold particle has a 90 nm diameter, 40 nm thickness, and the gap size is either 14 or 30 nm for the results reported here. The dimer antenna design requires the precise near-field coupling of the emitter to the gap antenna. To achieve this, we perform a two-step electron beam lithography process combined with chemical functionalization and binding of the Au structures and the QDs: the first lithography step defines the antenna structures on an ITO substrate, the second lithography step defines the area for chemical binding of the QDs (Figure 1b).13 As a last step, custom-made core/shell/shell CdSe/CdS/ZnS QDs are immobilized on the functionalized areas, and the excess QDs are removed by a lift-off step of the remaining PMMA. The QDs have quasi-spherical shapes with 10 nm diameter (Figure 1c) and peak emission at 660 nm with 30 nm FWHM. Details about the antenna fabrication, QD synthesis, functionalization and chemical coupling route to the nanostructures are given in the Experimental Section and in the Supporting Information. Extinction spectra representative of the studied antennas are shown in Figure 1d (prior to the QD binding procedure, which red-shifts the resonances by about 20 nm). This ensures the nanostructures resonances are close to both the excitation wavelength and the spectral range for the luminescence detection. Given the broad spectral response of the antennas as compared to the 30 nm FWHM emission from the QD, the antenna is assumed to induce only marginal modifications to the luminescence spectrum.
Figure 2a displays confocal photoluminescence images corresponding to the case of a dimer antenna with excitation polarization parallel or perpendicular to the dimer long axis. Over 80% of antennas show QD photoluminescence, and about 50% show polarization sensitive emission (red circles on Figure 2a represent the antennas selected for further investigations). For these antennas, the photoluminescence ratio between parallel and perpendicular excitation is higher than 10. The QD luminescence also turns into a clear linear polarization parallel to the long axis of the antenna, with a degree of linear polarization of 0.8 indicating QD coupling to the antenna.13
Figure 2b presents typical photoluminescence time traces. For QDs lying on bare ITO substrate, the blinking dynamics are strongly suppressed as compared to QDs on glass. We also observe a reduced exciton lifetime of 2.0 ns for QDs on ITO whereas the exciton lifetime is 9.5 ns on glass. These features relate to energy transfer from the QDs to the conductive ITO layer, as previously reported.25, 26 Interestingly, when the QD is coupled to a plasmonic antenna the blinking dynamics are partly retrieved while the luminescence lifetime is further reduced.27 This is indicative of an enhancement of the QD emission rate upon coupling to the metal antenna.
Another signature of the near-field coupling between dimer antenna and QD is found in the radiation pattern. The images in Figure 2c record the photoluminescence intensity distribution on the back focal plane (Fourier plane or momentum space) of the high numerical aperture (NA) objective, and contain the directions of emission toward the substrate. Two distinct circles are seen in the polar angle. The outer circle is the maximum collection angle of the 1.2 NA water-immersion objective (64°). The inner circle is the critical angle for the glass-air interface (NA = 1 or 41.1°), where a dipole close to a glass interface is expected to emit with a sharp maximum. The radiation patterns of a QD on ITO (left column) and of a QD coupled to a single gold particle (center) are isotropic in the azimuthal angle, as a direct consequence of the QD degenerate transition dipole moment.28, 29 When the QD is coupled to a dimer gap antenna (Figure 2c, right column), the radiation pattern changes dramatically and transforms to that of a linear dipole horizontally aligned respective to the interface. Hence, the antenna mode fully determines the QD radiation pattern, which is further evidence of coupling between the QD and the antenna.13, 30, 31
As a consequence of the strong quantum confinement of the free charge carriers, QDs can undergo sequential resonant photon absorption, and sustain multiply excited states.21 Below photoluminescence saturation, the singly excited (X) state mostly results from the absorption of a single photon while the doubly excited state (BX) state is created after the sequential absorption of two photons (Figure 3a). Immediately after pulsed photoexcitation, the X and BX relative populations thus scale linearly and quadratically (respectively) with the excitation intensity. The X and BX populations can be distinguished by monitoring the QD's transient emission dynamics,22 since the X state radiative lifetime is on the order of a few nanoseconds while the BX lifetime is dominated by Auger recombination and ranges between ten and a few hundreds of picoseconds (Figure 3a). The time correlated single photon counting trace s(t) is decomposed as a sum of two exponentials:
where aX and aBX are the amplitudes of characteristic decay times τX and τBX, respectively.
The amplitude aX relates to the singly excited X state, while aBX relates to the doubly excited BX state, as demonstrated later by the excitation power dependence. Contrary to earlier work,24 the X state decay of the QDs used in this study is well modeled by a single exponential (see the Supporting Information). The aX and aBX amplitudes can be expressed as function of the radiative decay rate ki rad for the i-th mulitexciton state and the average number Nabs of absorbed excitation photons per QD per pulse:22, 24
Equations (2) and (3) contain two important points. First, both aX and aBX are linearly proportional to the radiative rate kX rad. Hence both aX and aBX sense the same emission rate enhancement on plasmonic antennas. Second, aX bears a linear dependence with Nabs and the local excitation intensity while aBX has a quadratic dependence. While computing the ratio aBX/aX, the emission rate contribution cancels out, and only a term proportional to the local excitation intensity remains. As r is a constant, aBX/aX is a direct probe of the local excitation intensity. An increase in this ratio on a nanoantenna as compared to a flat interface is a direct demonstration of an increased local excitation intensity, independently on the number of emitters involved and the emission enhancement.
Figure 3b–d presents experimental decay traces and the corresponding biexponential fit according to Equation (1). These decay traces are normalized to better reveal the modifications of the decay dynamics and the increased relative weight of the fast BX transient component as the excitation power is raised. For the QD on ITO, the exciton lifetime amounts to τX = 2.0 ns, while for the QD coupled to a monomer antenna and for the QD coupled to a dimer antenna we find τX = 0.18 ns after deconvolution from the instrument response function.14 This modified exciton lifetime corresponds to a lifetime reduction or excitonic decay rate enhancement of 11.1 when the QD is coupled to the antenna. For all samples tested, the biexciton lifetime is found to τBX = 0.12 ns which is presently limited by the instrument response function. This short lifetime confirms that the Auger recombination route strongly affects the BX to X decay. The excitonic and biexcitonic amplitudes aX and aBX are displayed in the last column of Figure 3b–d. The excitonic amplitude aX is found to grow linearly with the excitation intensity, while the biexcitonic amplitude aBX grows quadratically, as expected from Equations (2) and (3). The evolution of aBX with the excitation power confirms that the fast transient component in the decay trace corresponds to the biexcitonic process BX.
From the data set in Figure 3b–d, we compute the ratio aBX/aX, which follows a linear trend as the excitation power is increased (Figure 4a). For all excitation powers, the aBX/aX ratio on plasmonic antennas exceeds the one found for the ITO reference. As demonstrated by Equations (2) and (3), this effect is directly related to the excitation intensity amplification induced by the antenna. To quantify the excitation intensity enhancement, we take the ratio of the slopes in Figure 4a, which provides a better estimate with a typical relative uncertainty of 10% for the excitation enhancement factors on the different selected antennas.
The orange bars in Figure 4b summarize our main results. The excitation enhancement factor starts at 5.1 for a single disk antenna. For the dimer antennas, as the gap size is reduced, the electromagnetic coupling between the gold particles is increased.11, 31 Our observations confirm this trend by an increase in the excitation enhancement from 13.1 to 15.9 when the gap is reduced from 30 to 14 nm. Let us stress that these enhancement factors for the excitation intensity sensed by the QDs are independent on the number of emitters involved in the luminescence signal and on the emission properties. The experimental results also stand in good agreement with numerical simulations based on the finite-difference time-domain FDTD method (see the Experimental Section for details). The only exception concerns the case with highest enhancement (dimer with 14 nm gap, parallel excitation), for which nanofabrication deficiencies, nonideal QD positioning and more complex photodynamics have a non-negligible influence.
The QDs used in this work are custom synthetized to bear a high BX population. If this is an advantage to determine the aBX/aX ratio, it also has the negative consequence to blur any photon antibunching experiment because of the emission of two photons within the BX state decay. Hence standard Hanbury–Brown–Twiss experiments cannot be successfully implemented to guarantee that there is a single QD under investigation. Instead of this, clues for the single emitter on the nanoantenna are brought by (i) the blinking dynamics down to the background level (see Figure 2b for typical time traces), and (ii) the luminescence intensity level on the selected antennas as compared to other antennas where most likely more than one QD is present (compare the intensity levels in Figure 2a). For the reference on the ITO substrate, we focus on an area with very low QD coverage density, where each bright spot is diffraction-limited and the intensity of bright spots follow a Poisson-like distribution. Selecting the spots with minimum intensities promotes the cases where most likely a single QD is present. Hence to quantify the luminescence enhancement, we compute the ratio of the average levels of luminescence from the time traces, and assume the detected signal stems from a single QD.
Figure 4c summarizes the enhancement factors found for the photoluminescence intensity. While a value of 2.7 is found for the single disk antenna, the luminescence is only enhanced by a factor of 1.3 for a dimer gap antenna with 14 nm gap and parallel excitation, although this configuration leads to a 15.9 excitation intensity enhancement. The situation is even worse for the dimer with perpendicular polarization orientation, where the luminescence factors drops to 0.25 of ITO reference. We remind that for the different cases the exciton decay rate is increased by a factor 11.1 as compared to the ITO reference. The fact that the luminescence enhancement is smaller than the decay rate enhancement and the excitation intensity enhancement is an indication of quenching, i.e., the non-radiative energy decays take the lead over the radiative routes.
In the excitation regime below photoluminescence saturation, the luminescence enhancement is proportional to the gains in collection efficiency, quantum yield and excitation intensity. Thus, from the measurements of the excitation and luminescence enhancement factors (Figure 4b,c), we compute the enhancement factor for the QD quantum yield multiplied by the collection efficiency. The results are displayed in Figure 4d. For all configurations, values below 1 are obtained, which further evidence luminescence quenching. Our observations support the fact that the quenching losses also increase as the gap size is reduced:18 antennas with gaps of 14 or 30 nm are found to provide almost the same quantum yield enhancement, although the excitation enhancement was found to be significantly higher in the case of the 14 nm gap. A remarkable feature of our study is that excitation intensity enhancement can be recorded despite this quenching phenomenon, and that near-field intensity information can be extracted from emission even in the presence of strong nonradiative losses. To conclude, this work provides new routes to experimentally investigate the physics of optical antennas, and optimize the excitation and emission processes independently for the future development of bright single-photon sources and biochemical sensors. Positioning a quantum dot respective to an optical antenna provides also an interesting approach to investigate the exciton-photon interaction beyond the classical dipole approximation.33, 34