1. Introduction

Optical antennas [1–3], similar to their radio frequency counterparts, capture free-space electromagnetic radiation and focus it to a small region beyond the diffraction limit. Optical antenna systems consisting of single or coupled nanoparticles have been studied extensively over the last years[4–19]. It is well established that two closely spaced nanoparticles show greater field enhancement than single particles, due to the coupling of their localised surface plasmon resonances (LSPRs) [14–16]. The strong field enhancements provided by LSPRs can be used to enhance optical spectroscopies such as SERS [20, 21] and multiphoton absorption [22] or modify the spectral characteristics of nanoscale emitters (dyes, quantum dots...) in the vicinity of the nanoparticles [23]. In the case of nanoemitters, fluorescence enhancement can be obtained via three phenomena. On the one hand, an increase of the emitter’s radiative decay rate can be induced by the interaction between the emitter and nanoparticles resonant with the emission spectral profile [24]. On the other hand the nanoantenna can improve the transduction of the excitation light field from near field to far field [6, 7, 9, 10, 11, 25, 24]. Finally, nanoantennas can exert directional control on the fluorescence emission, allowing for a higher portion of the light to be emitted within the finite aperture of the collection optics [26, 27].

To maximize the field enhancement, a careful optimization of the optical antennas’ characteristics needs to be explored [28–30]. It has been demonstrated that the field enhancement can be increased by changing the antenna dimensions, shrinking the size of the antenna feed gap, or using arrays of nanoantennas to improve the near-field coupling [2, 31, 32]. Recently studies have pointed out a very effective alternative to increase the field enhancement and the scattering cross section of metal nanoantennas via placing them on a continuous thin dielectric film on a metal underlayer. Specifically it was demonstrated that Metal-Insulator-Metal(MIM) structures working as resonators can be used to maximize the absorption in the visible regime when they are combined with nanoantennas arrays [33, 34]. The addition of a reflecting underlayer provides alternative ways of tuning optical antennas by changing the dielectric layer thickness as well as increasing the near-field enhancement, achieving high quality factors [35, 36]. This phenomenon can be explained by a near-field coupling between the antenna and its mirror image on the metal layer [37] and has been used to improve SERS [38] or the directivity on optical antenna substrates [39], allowing for single molecular detection [40]. Also, Gong et al. [41] show an enhancement in the emission intensity of Erbium-doped SiN placed in Metal-Insulator-Metal plasmonic cavities. However, none of these articles have investigated the possibility to use this concept to improve the radiation properties of single nanoantennas in terms of enhancement of the fluorescence of organic nanoemitters.

In this article, we investigate both theoretically and experimentally a plasmonic resonant system consisting of gold dimer nanoantennas separated by a thin dielectric layer of SiO₂ from a gold metal underlayer. The nanoemitter chosen for our study is meso-Tetraphenylporphyrin (TPP), an organic dye which presents two well-separated emission peaks at 650 and 750 nm, as shown in Fig. 1(c) [42]. We demonstrate that the addition of the metal layer in the nanoantenna fabrication leads to an increased scattering cross section and field enhancement. This is accompanied by an increase of photoluminescence enhancement. The asymmetry of the dimer leads to two LSPRs in different wavelength ranges along the two perpendicular polarisations. Those two spectrally separated LSPRs enhance selectively the emission bands of TPP depending on the polarisation.

 

Fig. 1 (a) Experimental setup for TPP fluorescence measurement. The excitation laser at a wavelength of 405 nm is coupled into a 40× reflective objective that focuses the laser beam on the sample and collects the emitted light. The fluorescence light is filtered using a dichroic mirror (DM) and a high pass filter (HP) at 420 nm in order to cut off the incident light scattered from the sample. The beam can be either sent to a spectrometer (SPEC) combined with a CCD camera, or to a pair of detectors. In the latter case, a 50 μm-pinhole PH selects the fluorescent light coming from the focal spot only. Lens L2 collimates the light transmitted through PH, which is then directed to a polarising beam splitter (PBS) that splits the two polarization components. Finally the light is focused on two APDs (avalanche photodiode) detectors that collect simultaneously both polarizations along x and y axes. (b) Schematic sample cross section. A film of TPP embedded in a PMMA matrix is deposited on top of the antennas at a thickness approximately equal to that of the antennas (40 nm). (c) Measured absorption and photoluminescence (PL) emission spectra from a film of TPP into a PMMA host matrix.

Download Full Size | PDF

2. Methods

Our study uses glass substrates covered by a 100 nm thick gold layer, further capped by a continuous 25 nm SiO₂ film (n_d = 1.45). The nanoantennas were nanofabricated on top of the SiO₂ film, and are 40 nm-thick gold dimers composed of two opposing long arms separated by a 30 nm gap, as illustrated in Fig. 1(b). The structures were fabricated by electron beam lithography (Elonix 100KV EBL system) with the process described in [43]. The antennas were fabricated in arrays at a separation of 4μm allowing to measure the optical response of a single antenna. The emitting layer is produced by spin-coating a film of TPP-doped PMMA film on the sample, at a thickness of approximately 40 nm. The weight concentration ratio PMMA/TPP is chosen to be 75 in order to maximize the TPP film fluorescence. Since TPP possesses a low quantum yield (<20%) [42], it is an ideal candidate for photoluminescence enhancement using plasmonic nanostructures [44].

Figure 1(a) presents a schematic diagram of the experimental setup. The excitation laser is set at a wavelength of 405 nm to match the absorption band of TPP [Fig. 1(c)]. A 40×, NA=0.6 microscope objective is used to focus the laser beam on the sample and collect the emitted light from the sample. The fluorescent light is filtered in order to cut off the incident light scattered from the sample. The fluorescent light can be sent to a spectrometer or to a polarizing beam splitter that splits the two polarization components, which are collected independently on two APDs (avalanche photodiode) detectors. The scattering cross sections of the nanoantennas were measured on individual nanoantennas using dark-field spectroscopy with a polarization-adjustable white-light illumination at an incident angle θ = 60° with respect to the normal of the sample surface, as described elsewhere [45]. Numerical simulations were performed to evaluate the scattering cross sections of the antennas as well as their near-field enhancement and the radiated power of an emitter in the presence of the antennas. The simulations were performed using the commercial 3D full-wave electromagnetic wave solver Lumerical FDTD Solutions, at similar conditions to the measurements using the experimental sample geometries. The optical constants were obtained by fitting the values found in [46] to a multi-coefficient model.

3. Results and discussion

We firstly describe and characterize the antenna system used in our study. The nanoantenna system is not yet covered with the dye-doped layer, and we choose a nanoantenna composed of two opposing 75 nm long arms (elongated disks) with a 30 nm gap. Figure 2(a) shows a simulation of the near-field enhancement as a function of the excitation wavelength by spatially integrating the field intensity at every point in the surrounding antenna volume with and without the reflecting layer. Since the confocal setup probes a zone limited by diffraction, we choose to integrate the near-field intensity in a volume comparable to the probing area of excitation (300 nm × 200 nm × 100 nm). This allows us to estimate the average near-field enhancement on the TPP molecules, which are randomly spatially distributed and oriented.

 

Fig. 2 Effect of a gold underlayer on the antenna properties. The nanoantennas are composed of two 75 nm long arms with a 30 nm gap on a glass substrate and surrounded with air (n=1). (a) Comparison between the near field (NF) enhacement, integrated over a volume of 300 nm × 200 nm × 100 nm surrounding the antenna, of dimer antennas placed on a glass substrate with (black line) and without (red line) gold underlayer. For this comparison, the polarization is selected parallel to the antenna axis, X-pol. Dotted red line: antenna with ≈ 95 nm long arms, without gold layer, to match the resonance at 655 nm. Inset: simulated power radiated by a dipole placed at the antenna gap for different values of SiO₂ thickness. The dipole is placed at the center of the gap cavity with polarization parallel to the antenna main axis in order to maximize the coupling. (b) Comparison between the simulated scattering cross section (continuous lines) and dark field scattering (dotted lines), for an incident polarisation along the main axis of the antenna. Experimental and simulated data in black refers to dimer antennas placed on a glass substrate and red data to dimmer antennas without gold underlayer. Inset: scanning electron microscope (SEM) image of one nanoantenna being considered.

Download Full Size | PDF

The simulation was carried out considering a broadband plane wave as illumination source and a SiO₂ separation of 25 nm. This comparison shows a field enhancement factor being amplified up to 5 times when the gold underlayer is used in these antennas. In addition, the resonance of the nanoantenna shifts to lower energies when the SiO₂ separation decreases, because of the near-field interaction between the antennas and the gold underlayer. This is explained by the hybridisation of the antenna dipole with its dipole image in the Au film leading to a red shift when the interaction is stronger [47]. Moreover in Fig. 2(a) we present the average enhancement corresponding to nanoantenna without gold reflecting layer (dotted line), with the dimensions such that the resonance position matches the position of the antenna with the reflective layer (long arm of 95 nm instead of 75 nm). This allows us to estimate the maximum amplification enhancement that the mirror layer introduces at a certain wavelength. In this case (resonance at 655 nm) the factor is approximately 3. To link this to the enhancement of the emission of a dye placed next to a nanoantenna, the inset of Fig. 2(a) presents the enhancement of the radiated power of a dipole ideally coupled to the gold nanoantenna, i.e. positioned at the center of the gap and oriented along the main axis of the dimer, influenced by the addition of the Au underlayer on the antenna substrate. Here the nanoantennas are placed on top of a SiO₂ layer at different thickness. The enhancement of the power radiated by the dipole increases gradually as the SiO₂ separation layer gets smaller, achieving amplification factors of more than 100 times compared to the case without the gold underlayer. This comparison illustrates the benefits of the underlayer on the radiation properties of dyes, albeit in the ideal situation where the emitter is perfectly placed and oriented along the gap. This picture is challenging to reproduce experimentally in our samples, TPP molecules are distributed randomly around the nanoantenna and with random orientation. In addition, the red-shift of the wavelength, previously discussed, allows us to select 25 nm as the SiO₂ layer thickness for our samples to optimize the overlap between the near field antenna resonance and the first emission peak of TPP at 640–680 nm. It is important to mention that the predicted near field antenna resonance under plane wave illumination matches the predicted antenna radiation resonance under the dipole excitation.

Figure 2(b) shows a comparison of the simulated scattering cross section (solid lines) and the dark field scattering of the antennas (dotted lines) placed on a glass substrate with (black) and without (red) gold underlayer. For these and all subsequent simulations, the PMMA layer is not taken into account because the measurement of the scattering cross sections of the antennas after deposition of the PMMA/TPP layer only exhibited a minor shift of the resonances (≈10 nm). Spectra are normalised relatively to the intensity of the resonance parallel to the dimmer axis for the case of using a gold underlayer. Far field projections (not shown) of the light emitted by a dipole in the situation described in Figure (2)a show that the underlayer does not change the radiation profile: the photoluminescence enhancement seen is not due to a beaming effect. For this reason, it is not necessary to take into account the finite NA of the microscope objectives used in the simulations. Hence we always estimate the scattering cross sections of the nanoantennas by integrating over the full 4π space. We observe a good agreement between experiment and simulation, both demonstrating that using a gold underlayer the scattering of the antenna increases by a factor up to 4 in the current configuration.

We then evaluate the effect of the gold underlayer on the photoluminescence (PL) enhancement of the TPP emitters placed on the antenna vicinity. Figure 3(a) shows the spectral overlap between the emission band of TPP and the simulated near-field resonances for the polarization along the main axis of the dimer with and without the gold underlayer. In both cases, the near-field resonances predominantly match the first emission peak at 640–680 nm. We thus filter the PL in this wavelength range to maximize the intensity contrast Fig. 3(b). The fluorescence comparison shows that antennas without the Au under layer induces a 10% PL enhancement while the antennas with the gold underlayer reach up to 50%, as highlighted on the profiles on Fig. 3(b–d). This result is in good agreement with the values predicted by the simulations as discussed previously. The ratio of the enhancements is slightly higher than what is expected from the near field integrations discussed in Fig. 2(a), probably because the overlap between the resonance of the nanoantenna and the emission peak of TPP is stronger in the case with the reflective gold layer. Note while, the total maximum enhancement seen is modest, this is due to the fact that the radiation collected is integrated over the whole confocal volume, which contains many dyes, randomly positioned and oriented - only the ones in the centre of the gap experience for optimum enhancement.

 

Fig. 3 (a) Spectral overlap between the emission band of TPP and simulated NF resonances with and without gold underlayer, for the polarization parallel to the dimer main axis. (b) (resp. (c)) Photoluminescence intensity around single resonant antennas for polarization of the detected light parallel to the antenna axis without (resp. with) the gold underlayer (scale bars 2 μm). Only light in the range 640–680 nm is collected, which corresponds to the first emission peak of TPP. The PL intensity I is normalized to the intensity measured away from the nanoantennas, I₀. Bottom: horizontal profiles as indicated on the images above. The antennas without the Au underlayer induced a 10% PL enhancement, while antennas with the gold underlayer reach up to 50% PL enhancement.

Download Full Size | PDF

It has been reported that the fluorescence enhancement can be written as the product of two factors related to different properties of LSPRs [23]. On the one hand, the near-field enhancement $\left|\frac{E}{E_o}\right|$ (where E_o is the free-space electric field that illuminates the nanoantenna) is related to the capability of plasmons to concentrate electromagnetic energy into sub-wavelength volumes and it is evaluated at the excitation frequency, effectively increasing the absorption cross section of the nanoemitter. The other factor is the quantum yield enhancement $\frac{\eta }{{\eta }_o}$,which is related to the modification of the photonic density of states at the emitter position induced by the localized resonance and it is determined by the emission frequency. These two factors can be simultaneously much larger than unity, which leads to strong enhancements of the fluorescence intensity [48]. Considering that the emission and absorption of TPP are spectrally well separated [Fig. 1(c)], and the excitation wavelength at 405 nm does not excite plasmon resonance in our antennas, the observed PL increase should be mainly due to the direct emission enhancement

Furthermore, we study the effect of polarization for the case of a dimmer antenna with the reflecting underlayer. Figure 4 shows the two resonances sustained by a gold dimer for both parallel and perpendicular polarizations with respect to the dimmer axis. For this study, the dimer antennas used were chosen with a larger length in order to have the main resonance along the X axis matching the low energy emission peak of TPP at 700–740 nm. The resonance along the X axis is much more intense than the Y-axis resonance due to the strong coupling between the particles as reflected also in the near-field profile [Fig. 4(c)]. Dark field measurement and scattering simulation in Fig. 4(a) clearly show the difference in intensities and the spectral separation between both resonances.

 

Fig. 4 Influence of the polarization on the scattering of a gold dimer consisting of two elongated disks with the dimensions 105×60×40 nm³ (long axis × short axis × thickness) separated by a 30 nm gap. (a) Dark field spectra (solid lines) and simulation of the scattering cross section (dotted lines) of the nanoantenna for X and Y polarizations In all cases the values are normalised to the value at the peak for the X polarisation. Inset: SEM image of the gold antenna considered (scale bar: 100 nm). (b) Near field intensity distribution of the antenna for illumination of polarization perpendicular (Y direction) and (c) parallel (X direction) to the antenna axis.

Download Full Size | PDF

Figure 5(a) compares the emission band of TPP and the simulated near-field resonances for an excitation along the main axis of the dimer (X-polarised) and perpendicular to it (Y-polarised). There is a good overlap between the low energy peak at 700–740 nm and the X-resonance and a partial overlap of the Y-resonance with the high energy peak at 640–680 nm. In addition, a small contribution from the X-resonance tail overlaps with the high energy peak. The spectrum of PL of the TPP is acquired without specific filtering, on the antennas and away from them. The results, as shown in Fig. 5(b), show a PL enhancement in all the emission ranges, which is more pronounced at the low energy emission peak, with a 25% increase in the peak intensity. This is in agreement with the overlap presented in Fig. 5(a). The enhancement is 5% for the high energy peak, attributed to the smaller contribution from both resonances in this spectral region at 640–680 nm. Figure 5(c) illustrates the PL enhancement difference for both peaks by imaging three identical nanoantennas, either selecting only the first peak (top) or only the second (bottom).

 

Fig. 5 (a) Spectral overlap between the emission bands of TPP (blue curve) and the simulated near field resonances for the X and Y polarizations, black curve and red curve respectively. The PL collected is filtered either in the area highlighted in blue (640–680 nm) or in green (700–740 nm). (b) TPP emission spectra collected from one antenna (black curve) and on the plain TPP film (red) without a filter. (c) Filtered photoluminescence intensity images around single resonant antennas: filtering for the first emission peak at 640–680 nm (top) and for the second emission peak at 700–740 nm (bottom). Note that here no particular polarisation is selected for the PL. Scale bars: 2μm.

Download Full Size | PDF

We finish our discussion by analysing the effect of both resonances independently and experimentally test the overlap discussed in Fig. 5(a). To do so, we collect independently the two polarization components of the PL emitted, as described in Fig. 1. Figure 6(a) shows the PL enhancement in the X polarization. In this case, the low energy peak is enhanced up to 80% whereas the high energy peak is enhanced up to 20%. The ratio between the two enhancements is in a reasonable agreement with the predicted overlap between the near field resonance in the X-axis resonance and the two TPP emission peaks. Similarly, Fig. 6(b) shows that for the polarization perpendicular to the antenna axis, the TPP emission peak around 640–680 nm is more enhanced than the low energy peak, which is again consistent with the magnitude of the field enhancement provided by the nanoantenna for this polarisation. These observations support that doubly resonant systems, like gold dimers, can be used to selectively influence the polarization of emitted light in a specific wavelength range.

 

Fig. 6 Study of the influence of the nanoantenna on the polarisation properties of the PL emitted. (a) Top: PL intensity around single nanoantennas for polarization of the detected light parallel to the nanoantenna axis, filtered at around 640–680 nm and 700–740 nm. Scale bars: 1μm. Bottom: horizontal profiles comparing the PL of the two emission peaks for X-polarised detected light. The TPP emission peak in the 700–740 nm range is preferentially enhanced over the peak at 640–680 nm. The inset is an SEM image of the antenna considered, scale bar 100 nm. (b) PL intensity for polarization of the detected light perpendicular to the antenna axis. The peak around 640–680 nm is preferentially enhanced in this case.

Download Full Size | PDF

4. Conclusions

We have demonstrated that the addition of a gold underlayer to the fabrication of nanoantennas can improves their properties, by offering new possibilities of resonance tuning by means of controlling the spacer thickness and increasing their scattered radiation power and near-field enhancement. These in turn increase the photoluminescence enhancement induced by the antenna on a low efficiency light emitter such as TPP. In addition, this improvement enhances the intensity of the two resonances associated with the perpendicular light polarizations existing in a gold dimer, allowing to measure their influence in a specific wavelength of the dye emission. Such a system presents a way to selectively influence the polarization of emitted light in organic dyes with multiple emission peaks.