1. Introduction

Single molecule fluorescence studies using conventional microscopes are limited by diffraction to detection volumes in the femtoliter range. This volume dimension limits the maximum molecular concentration required to have a single molecule in the detection volume (typically in the nanomolar range), and the maximum signal-to-background ratio that can be achieved [1]. To go beyond these limits, a first approach uses circular apertures of 50 to 200 nm diameter milled in a metal film (so-called zero-mode waveguides [2]) to reduce the detection volume down to the attoliter range. However, the fluorescence signal per molecule drops dramatically as the aperture diameter is reduced below 100 nm, as a consequence of increased quenching losses [3]. To further confine the light and increase the molecular brightness, plasmonic nanoantennas stand out as powerful tools to control the light emission from nanoscale volumes [4–7]. These exciting possibilities have been applied to the detection of fluorescent molecules in solution using gold nanoparticles [8–11], silver island films [12], or more complex antenna designs [13–15].

Individual metal nanoparticles have a strong appeal for practical applications to detect fluorescent molecules in solution owing to their large availability, low intrinsic cost, and tunable spectral response. However, two technical issues are challenging the applications: (i) the large contribution in the fluorescence signal from molecules tens of nanometer away from the nanoparticles [9, 10, 12, 13] and (ii) the molecular binding to the metal [8, 11, 14]. Quantifying the near-field volume where the electric field intensity is enhanced together with the average fluorescence enhancement over this volume is of high interest [16], yet quantitative estimates for fluorescent molecules in solution are still lacking.

Here we use individual spherical gold nanoparticles to perform enhanced single molecule fluorescence analysis in solutions at high (micromolar) concentrations. Figure 1(a–c) present a schematic of the experimental configuration. Nanoparticles diameters ranging from 60 to 150 nm are used to tune the local surface plasmon resonance (LSPR) across the fluorescence excitation and emission spectra (Fig. 1(d)). At the targeted concentration of fluorescent molecules of 10 μM, the 0.5 fL confocal detection volume (diffraction-limited) contains about 3000 molecules, while only a few molecules are expected to be in the sub-attoliter near-field volume around the nanoparticle. Hence, the collected fluorescence signal is a sum of two contributions: the enhanced fluorescence from a few molecules in the nanoparticle near-field plus a fluorescence background from several thousands of freely diffusing molecules within the confocal volume. To discriminate between these contributions, we use emitters with low quantum yield and take advantage of the higher fluorescence enhancement factors obtained with them [5, 15]. We add 200 mM of methyl viologen to the solution containing the Alexa Fluor 647 dyes to quench the dye quantum yield to 8%. Lastly, to quantify the near-field detection volume and the fluorescence enhancement, we use fluorescence correlation spectroscopy (FCS) to analyze the fluorescence trace. FCS records the temporal fluctuations of the fluorescence signal and computes its temporal correlation. For a homogeneous sample with a single fluorescent species, the amplitude of the correlation function scales with the inverse of the average number of molecules in the detection volume N. The knowledge of N together with the (calibrated) concentration of fluorescent molecules quantifies the detection volume as well as the average fluorescence brightness computed per molecule.

 

Fig. 1 (a) Gold nanoparticle on a glass substrate for enhanced single molecule analysis at high concentrations. (b) Finite-difference time-domain computation of excitation intensity enhancement near a 80 nm gold nanoparticle. The incoming light is horizontally polarized at a wavelength of 633 nm. (c) Bright-field optical image of 80 nm gold nanoparticles immobilized on a glass susbtrate. (d) Normalized experimental scattering cross-section of the nanoparticles (color-shaded curves, the label on top on the graphs indicates the nanoparticle diameter in nm), taken in pure water in the absence of the fluorescent species. The dashed and solid dark blue lines indicate the normalized absorption and emission spectra of Alexa Fluor 647 dye. The vertical lines indicate the 633 nm laser line used for excitation and the 650–690 nm region used for fluorescence detection.

Download Full Size | PDF

2. Materials and methods

Throughout this work, we use commercial gold nanoparticles (BBI Solutions) of calibrated diameters ranging from 60 to 150 nm with a typical dispersion below 10% in diameter [13]. The nanoparticles are diluted in pure water, dispersed and dried on a glass coverslip to isolate single nanoparticles, which is confirmed by confocal laser scanning and checking the invariance on the fluorescence results upon turning the linear polarization orientation of the excitation laser beam. The nanoparticle sample is covered by the solution containing Alexa Fluor 647 fluorescent dyes at micromolar concentrations together with 200 mM of methyl viologen (1,1′-Dimethyl-4,4′-bipyridinium dichloride). The experimental set-up is based on a confocal inverted microscope with a 40× 1.2NA water-immersion objective. The excitation source is a He-Ne laser at 633 nm with 10 μW excitation power. Positioning the nanoparticle at the laser focus spot is obtained with a three-axis piezoelectric stage. A dichroic mirror and a long pass filter separate the fluorescence light from the scattered laser light. After a 30 μm confocal pinhole conjugated to the sample plane, the detection is performed by two avalanche photodiodes with 670 ± 20 nm fluorescence bandpass filters. The fluorescence intensity temporal fluctuations are analyzed with a hardware correlator (ALV6000, ALV GmbH). The analysis of the FCS data considers two species with different fluorescence brightness: N^* molecules in the nanoparticle near-field with brightness Q^*, and N₀ background molecules with brightness Q₀ diffusing away in the confocal volume. The temporal correlation G of the fluorescence intensity F can be written [1]:

3. Experimental results

Confocal experiments on a solution at 4.5 μM fluorophore concentration yield a large fluorescence intensity with reduced temporal fluctuations (Fig. 2(a)), corresponding to a weak correlation amplitude of 8 × 10⁻⁴ (Fig. 2(b)). Fitting the confocal FCS data indicates an average number of N_sol = 1330 molecules with a diffusion time of 65 μs, as expected for the 0.5 fL detection volume calibrated previously and the 4.5 μM concentration. The brightness per molecule at 10 μW excitation power with 200 mM methyl viologen is Q_sol = 0.14 kHz. Experimental results are summarized in Tab. 1.

 

Fig. 2 (a) Fluorescence time trace and (b) FCS correlation functions for the reference confocal (blue) and the 80 nm nanoparticle (red). Alexa Fluor 647 concentration is 4.5 μM, with 200 mM of methyl viologen as chemical quencher. Thick lines are numerical fit of the data using Eq. (1). The insert displays normalized FCS traces.

Download Full Size | PDF

Table 1. Fitting parameter results for the FCS curves in Fig. 2(b).

View Table

Performing FCS experiments with a 80 nm spherical gold nanoparticle provides correlation amplitudes about 12 times higher with significantly improved signal-to-noise ratio. FCS analysis quantifies an average number of N^* = 0.74 molecules in the 80 nm nanoparticle near-field with brightness Q^* = 8.9kHz. This molecular brightness corresponds to a fluorescence enhancement of Q^*/Q_sol = 64 times. Given the calibrated Alexa concentration of 4.5 μM, we relate the N^* = 0.74 molecules in the nanoparticle near-field to a detection volume of 270 zL (1zL = 10⁻²¹L), or equivalently a detection volume reduction of N_sol/N^* = 1800 times as compared to the diffraction-limited confocal volume. Moreover, a clear reduction of diffusion time down to 11 μs is observed with the nanoparticle, which confirms that the translational diffusion events take place from a sub-diffraction volume.

FCS measurements were taken for several individual gold nanoparticles of diameter ranging from 60 to 150 nm to quantify the fluorescence enhancement factor and the near-field detection volume. The results are summarized in Fig. 3(a) and 3(b), together with estimates of the near-field detection volume derived from finite-difference time-domain simulations using the approach described in [17]. At least 10 different nanoparticles are tested for each diameter. The error bars displayed in Fig. 3 indicate the standard deviations of our measurements.

 

Fig. 3 Fluorescence enhancement (a) and near-field detection volume (b).

Download Full Size | PDF

Figure 3(a) clearly indicates an optimum diameter around 80 nm for the highest fluorescence enhancement factor. Remarkably, the 150 nm nanoparticle diameter for which the LSPR wavelength coincides with the peak fluorescence emission wavelength (Fig. 1(d)) does not provide the highest fluorescence enhancement. We find that the optimum fluorescence enhancement appears when the emission wavelength is red-shifted from the LSPR wavelength. This follows from the trade-off that must be found between radiative and non-radiative rates enhancement to achieve high fluorescence enhancement [18]. As predicted numerically [19, 20], the peak wavelength for enhanced non-radiative decay rate coincides with the LSPR wavelength but then drops rapidly towards the red side of the LSPR, while the radiative rate has a longer tail towards the red. Consequently, the quantum yield bears a more pronounced enhancement for wavelengths red-shifted as compared to the LSPR.

Lastly, series of FCS curves were recorded for increasing concentrations of fluorescent probe (Fig. 4(a)). We obtain that the number of molecules in the near-field region N^* follows a linear relationship with the fluorophore concentration (Fig. 4(b)), demonstrating that our approach can accurately quantify the number of detected molecules in the concentration regime exceeding 20 μM, and provide single molecule sensitivity at relevant physiological concentrations.

 

Fig. 4 (a) Fluorescence correlation functions for increasing concentrations of fluorescent probes using a 100 nm diameter gold nanoparticle. (b) Number of detected molecules in the nanoparticle near-field versus the molecular concentration.

Download Full Size | PDF

4. Conclusion

Taking advantage of the 60-fold fluorescence enhancement found for low quantum yield fluorescent species near metal nanoparticles, we perform fluorescence correlation spectroscopy analysis in detection volumes down to 270 zL (three orders of magnitude beyond the diffraction barrier) within a solution of concentration exceeding ten micromolar. We provide quantitative measurements of the near-field detection volume and fluorescence enhancement factors for various sizes of gold nanoparticles. Colloidal nanoparticles represent an inexpensive and versatile platform to perform a wide variety of biochemical assays in solution with single molecule resolution at the biologically relevant micromolar concentration regime.