1. Introduction

Optical fiber probes offer exciting opportunities to broaden the application range of fluorescence spectroscopy to conditions where conventional optical microscopy cannot be used [1–4]. Up to date, the large luminescence background generated in the silica cores of conventional fibers has been a severe limit to the detection sensitivity [5]. To circumvent this issue, one approach restricts the fluorescent probes to bright sources such as fluorescent nanospheres of at least 24 nm in diameter [6–8]. Another approach works only at fluorophore concentrations larger than 500 nM to guarantee that several hundreds of fluorescent molecules contribute to the detected signal [9, 10]. Detecting a single fluorescent molecule at concentrations down to a few nanomolars with an optical fiber probe remains a challenge [11].

Here, we demonstrate that it is possible to detect and analyze single fluorescent molecules in solution with concentrations down to the nanomolar range using a hollow-core photonic crystal fiber (HC-PCF) probe. HC-PCF guide the light in an air core, with only a small fraction of the light penetrating into the glass [12, 13]. Hence, the luminescence background generated in the HC-PCF can be reduced by about two orders of magnitude compared to conventional silica fibers [14–16]. The large mode size and low numerical aperture of the HC-PCF command the use of an objective coupler lens to efficiently excite and collect the fluorescence from the sample. Here, we use a 20 μm polystyrene microsphere set at the HC-PCF end-face. Recently, it was demonstrated that a micron-scale dielectric sphere with a refractive index of approximately 1.6 can focus light in a beam of subwavelength transverse dimensions and low divergence termed “photonic nanojet” [17–19]. Dielectric microspheres offer attractive solutions for designing focusing devices with excellent optical properties and compactness at minimal costs [9, 20–23].

To analyze the fluorescence signal, we implement fluorescence correlation spectroscopy (FCS), which is a versatile and widely used method for the characterization of fluorescent probes [24, 25]. FCS analyzes the temporal correlation of the fluorescence intensity fluctuations, and provides information about any molecular process inducing a change in the fluorescence intensity. Applications go much beyond fluorescence detection based on photon counts levels, and include determining translational and rotational diffusion, molecular concentrations, chemical kinetics, and binding reactions.

2. Materials and methods

2.1. Hollow-core photonic crystal fiber

The large-pitch Kagome-lattice fiber is fabricated using the conventional stack-and-draw technique [13]. Figure 1(a) shows a scanning electron micrograph of its end-face. Its transmission spectrum, measured by the cutback technique, is shown on Fig. 1(b) together with the 633 nm laser line and the 650–690 nm spectral region used for fluorescence collection. The background luminescence spectrum, acquired by a spectrometer with a Peltier-cooled CCD camera (Horiba Jobin Yvon iHR320), is shown on Fig. 1(c). The background luminescence intensity measured by the avalanche photodiode in the normal conditions for FCS experiments (Fig. 1(d)) typically amounts to 1.2 kHz for 750 μW excitation power at the fiber input.

 

Fig. 1 (a) Scanning electron micrograph of the Kagome-lattice HC-PCF. (b) Transmission losses. The red line indicates the 633 nm wavelength used for excitation and the orange shaded region is the spectral range used for fluorescence collection.(c) Luminescence background spectrum (red) with maximum 11 mW input power and 10 s integration time. The black line is the background noise of the spectrometer. (d) Experimental setup. (e) Computed electric field intensity with a 20 μm polystyrene sphere illuminated by the fundamental Gaussian-shaped fiber mode at λ = 633 nm. (f) Horizontal cut at the best focus plane. (g) Vertical cut along the microsphere center.

Download Full Size | PDF

2.2. Polystyrene microsphere as focusing lens

Numerical modelling is based on the finite-difference time-domain FDTD method using the commercial software Rsoft Fullwave version 6.0. The model considers mesh sizes of 20 nm and 50 nm along the horizontal and vertical directions, respectively, and perfect matched layers boundary conditions on all faces. The 20 μm sphere has a refractive index of 1.59, while the surrounding medium is water (refractive index 1.33). The sphere is illuminated by a Gaussian-shaped beam at λ = 633 nm with 14 μm transverse waist, corresponding to the HC-PCF fundamental mode. Figure 1(e) shows the computed electric field intensity. The focusing ability of the sphere is quantified from Fig. 1(f) and (g) to transverse and axial FWHMs of 560 nm and 6.5 μm respectively. This corresponds to what can be achieved with a water-immersion objective of 0.6 NA.

2.3. Experimental setup

Alexa Fluor 647 (Invitrogen, Carlsbad, CA), a common dye for FCS with absorption / emission maxima at 650 and 670 nm, is used as fluorescent probe. The excitation beam is provided by a linearly polarized CW He-Ne laser operating at 633 nm, and is coupled into the 1.5 m long HC-PCF by a 150 mm lens, matching the 0.02 NA of the HC-PCF. The total transmission of the HC-PCF, including coupling and transmission losses, is about 60%. To avoid filling the HC-PCF with the fluorescent liquid, a standard microscope coverslip (thickness 140 μm) is set in contact with the fiber end-face (Fig. 1(d)). The same glass coverslip holds the polystyrene microsphere, which is centered respective to the HC-PCF core with the help of a micrometer driven three-axis translation stage under visual control with an optical microscope equipped with 10x objective. The microsphere used for laser focusing also collects the fluorescence light and couples it into the HC-PCF, where both laser and fluorescence beams are counter-propagating. We use a dichroic mirror (Omega Filters 650DRLP), a long-pass filter (Omega Filters 640AELP) and a bandpass filter (Omega Filters 670DF40) to separate the fluorescence light from the reflected and elastically scattered laser light. A 20 μm pinhole is conjugated to the fiber output with a 1:1 magnification to spatially filter the HC-PCF mode carrying the fluorescence light. The fluorescence signal is finally detected by an avalanche photodiode (Perkin-Elmer SPCM-AQR-13). To perform FCS, the fluorescence intensity temporal fluctuations are analyzed by computing the temporal correlation of the APD signal with a Flex-02C hardware correlator (Correlator.com). Each FCS measurement is the average of 10 runs of 10 s duration.

2.4. FCS analysis

As for earlier work [9, 22], analysis of the FCS data is based on the analytical model established for Brownian three-dimensional diffusion in the case of a Gaussian molecular detection efficiency [24, 25]:

3. Experimental results with a 20 μm sphere

Figure 2(a) and (b) show the fluorescence time trace and correlation function obtained while dipping the HC-PCF-microsphere probe into the Alexa Fluor 647 solution at 3 nM concentration. We stress that no temporal correlation was found in the absence of the microsphere; this highlights the crucial role of the microsphere as focusing element to confine the excitation beam and collect the fluorescence light. Thanks to the fiber hollow core, the background noise in Fig. 2(a) amounts to only 1.2 kHz, which is more than 40x lower than the fluorescence intensity at 3 nM concentration and appears almost negligible. Earlier work with a microsphere attached to a silica fiber probe [9] reported for background noise of 270 kHz, which was at the same level as the detected fluorescence signal at a much higher 500 nM concentration.

 

Fig. 2 Fluorescence intensity time trace (a) and correlation function (b) obtained with the HC-PCF and the 20 μm sphere (excitation power 750 μW at the fiber input). The curves in (c) and (d) are for a conventional confocal microscope with a 63x 1.2NA water-immersion objective (excitation power 40 μW). Black lines in (b), (d) are numerical fits using Eq. (1). For both cases, the Alexa Fluor 647 concentration is 3 nM.

Download Full Size | PDF

To compare the performance of the HC-PCF-microsphere probe to conventional microscope objectives, Fig. 2(c) and (d) show the fluorescence time trace and correlation function obtained with a 1.2 NA planapochromat objective. The results of the FCS fits according to Eq. (1) are indicated in Fig. 2(b) and (d) respectively. About 10 molecules are detected on average with the HC-PCF-microsphere probe at 3 nM dye concentration, this is 11x larger than the number of molecules detected with the 1.2 NA objective and relate to the larger focal volume with the HC-PCF-microsphere probe. Accordingly, the diffusion time is larger with the HC-PCF-microsphere (225 μs) than with the 1.2 NA objective (64 μs). Given the calibrated translational diffusion coefficient D = 3.0×10⁻⁶ cm²/s for Alexa Fluor 647 at 21°C [26, 27] and the relation τ_d = w²/4D, a transverse waist of w = 520 nm is inferred for the HC-PCF-microsphere probe; this stands in good agreement with the numerical simulations shown in Fig. 1(e) and (f) and further confirms the focusing properties of the 20 μm polystyrene sphere.

Figure 3(a) shows the linear evolution of the total number of molecules detected with the HC-PCF-microsphere versus the concentration of fluorescent dye. Remarkably, experiments could be conducted down to 0.4 nM, which is 1000x lower than with the previous system [9]. From the slope in Fig. 3(a), we quantify the analysis volume to 6.0 fL. Figure 3(b) presents the fluorescence count rates computed back to per molecule for different excitation powers. A transition to fluorescence saturation occurs for excitation powers higher than 1 mW at the fiber input (focal spot intensity 70 kW/cm²). The background noise from the fiber remains remarkably below 2 kHz, even at high excitation powers. At 750 μW excitation power, the average count rate CRM for a single molecule is about 4x higher than the total background noise. This allows for routine detection of single molecule in solution with the HC-PCF-microsphere probe. Please note the background noise in Fig. 3(b) also includes the APD dark counts of 480 counts/s. Further reduction of this noise could be obtained by selecting a photodiode with lower dark counts and/or time gating the photodetection events.

 

Fig. 3 (a) Number of molecules in the FCS detection volume versus the molecular concentration. The inset shows the corresponding fluorescence correlation functions. (b) Fluorescence count rate computed back to per molecule (red) and background noise (gray) versus the excitation power.

Download Full Size | PDF

One of the main interest of FCS is to distinguish between different fluorescent molecules based on their diffusion properties, which are directly related to their molecular weight and hydrodynamic radius. As a proof-of-principle, we investigate the detection of the cellular protein Annexin A5b labelled with the Cyanine Cy5 fluorescent dye. Annexin A5b is a 36 kDa phospholipid-binding protein and has a high affinity to phospholipid phosphatidylserine in the presence of physiological concentrations of calcium [28]. Figure 4(a) and (b) show the fluorescence correlation functions obtained with the HC-PCF-microsphere and the 1.2NA objective, respectively. As compared to the Alexa Fluor 647 free dye (dashed lines), the diffusion time τ_d is shifted from 225 μs to 910 μs for the A5B-Cy5 labelled protein measured by the HC-PCF-microsphere. A similar ∼ 4× increase in the diffusion time is observed with the 1.2NA objective, confirming the results with the HC-PCF probe. Translating these results into hydrodynamic radii, we find an hydrodynamic radius of 2.8 nm for A5B-Cy5, while the hydrodynamic radius of Alexa Fluor 647 is 0.7 nm. These results demonstrate that the HC-PCF-microsphere probe can be readily used to distinguish between molecules of similar fluorescence spectra based on the FCS analysis of their translational diffusion properties.

 

Fig. 4 Fluorescence correlation functions for the cellular protein Annexin A5b labelled with Cyanine-5 obtained with the HC-PCF-microsphere (a) and the 1.2NA objective (b). the dashed lines are the reference correlation functions for the Alexa Fluor 647 free dye. For all cases, the dye concentration is 13.6 nM.

Download Full Size | PDF

4. Matching the HC-PCF numerical aperture

To further improve the system sensitivity, the fluorescence signal per molecule should be increased by tighter focusing the excitation light and better collecting the fluorescence radiation. This could be conveniently done by reducing the microsphere diameter [20]. However, experiments using 2 μm microspheres instead of 20 μm were a failure, as no correlation functions could be obtained. This can be explained as the fluorescence beam exiting the 2 μm microsphere has necessarily a diameter of d = 2 μm, which sets an angular divergence limited by diffraction of θ ≈ 2λ/πd ≈ 0.2. For efficient collection of the fluorescence beam by the HC-PCF, the angular divergence θ has to match the fiber numerical aperture. However, because of the large core and single-defect design, the Kagome-lattice HC-PCF used here has a numerical aperture of only 0.02, which is about 10x lower than the angular divergence θ emerging from a 2 μm microsphere and results in improper collection of the fluorescence signal by the HC-PCF.

Some relay optics need to be inserted in order to match the HC-PCF numerical aperture (0.02) to the angular divergence (0.2) of the fluorescence beam exiting the 2 μm microsphere. This can be elegantly done by a microscope objective of 10x magnification (the angular magnification being the inverse of the lateral magnification). The experimental setup is pictured in Fig. 5(a). With this configuration, the results found with the HC-PCF and the 2 μm microsphere (Fig. 5(b) and (c)) indicate a 0.8 fL detection volume, short diffusion times and high count rates. The performance of this configuration is comparable to that of a conventional confocal microscope (Fig. 5(d) and (e)), while preserving the property of foldable light path. However, since the relay optics introduce some complexity in the optical alignment of the probe, we believe the simple configuration with the 20 μm sphere is more relevant for practical applications. The experiment in Fig. 5 is rather an exploration of the influence of the HC-PCF numerical aperture and of the best performance that could be reached without considering optical complexity.

 

Fig. 5 (a) Experimental setup with the HC-PCF combined with a 10x 0.3NA objective and a 2 μm diameter polystyrene microsphere. The inset shows the electric field intensity computed for a 2 μm polystyrene sphere set at the focus of the 0.3NA objective. (b) Fluorescence intensity time trace and correlation function (c) obtained with the HC-PCF (excitation power 500 μW at the fiber input). The curves in (d) and (e) are for a conventional confocal microscope with a 63x 1.2NA water-immersion objective (excitation power 40 μW). For both cases, the Alexa Fluor 647 concentration is 30 nM.

Download Full Size | PDF

5. Conclusion

Hollow-core photonic crystal fiber probes offer reduced luminescence background as compared to conventional glass fibers. The combination of the HC-PCF with a polystyrene microsphere realizes a new probe for remote fluorescence sensing with single molecule sensitivity down to nanomolar concentrations. As compared to earlier work [9], the HC-PCF probe enables a 200x improvement of the signal-to-noise ratio for a single molecule detection event, and a 1000x reduction of the minimum detectable concentration with FCS. Further improvement of the detection sensitivity can be obtained by selecting photodiodes with lower dark counts and time gating the photodetection events [11].

Realizing portable single molecule fluorescence sensing devices is of great interest due to the widespread popularity of molecular-fluorescence-based measurements in fields such as chemistry, molecular biology, materials science, and medicine. Portable systems for fluorescence analysis with similar sensitivity as conventional microscope can be foreseen, with applications for instance in early diagnosis of Alzheimer’s disease [29], oxygen sensing [30], temperature sensing [31] or molecular sorting [32]. This system could also be extended to laser contact surgery [23] or optical transfection [33].