1. Introduction

Evanescent wave spectroscopy is a powerful method for the study of the optical properties of liquid and solid samples at interfaces [1]. The technique relies upon the absorption or dispersive shift of an evanescent wave created through total internal reflection at the interface between two dielectric media. In fiber-based methods, one of the media is an unclad optical fiber and the second the analyte [2, 3]. The main advantages of this approach are stability, fast response and the possibility to perform in situ measurements which make it specially suited for sensing applications [4]. Information about the structure and configuration of the particles at the interface can be obtained via absorption [2, 5] as well as fluorescence [6] spectroscopy.

In this context, optical nanofibers with diameters smaller than the wavelength of the guided light exhibit a strong radial confinement of the light and a pronounced evanescent field surrounding the fiber [7]. This makes them well suited for the controlled interaction of the guided light with particles near or on the fiber surface and opens the route towards applications in fields like optical sensing [8], nonlinear optics [9] or cold atoms physics [10, 11]. It has been demonstrated that surface absorption spectroscopy using nanofibers is several orders of magnitude more sensitive than conventional methods based on free beam absorption [12]. The resulting short integration time allows us to study and characterize dynamical processes of the molecules on the surface. Furthermore, the fluorescence from a very small number of resonantly irradiated atoms around a nanofiber has been coupled into the guided fiber mode and spectrally analyzed [13]. For thin molecular films the measurement of both, absorption and fluorescence, could provide complementary information about the molecules on the surface.

Here, we show that the fluorescence light emitted by 3,4,9,10-perylene-tetracarboxylic dianhydride molecules (PTCDA) adsorbed on the surface of a nanofiber is efficiently coupled into the guided mode of the fiber yielding a high degree of sensitivity for spectroscopic studies. Moreover, our method is very stable and reproducible. To our knowledge this is the first time that a nanofiber is used to collect the fluorescence of surface-adsorbed particles. We measure fluorescence spectra for surface coverages as low as 1 ‰ of a compact monolayer (ML). Interlaced measurements of absorption and fluorescence spectra are performed in order to determine the respective surface coverage. Furthermore, we observe self-absorption in our system, i.e., a partial reabsorption of the emitted fluorescence by circumjacent molecules along the nanofiber. While the high sensitivity of our method allows us to perform measurements in a regime of low surface coverages where self-absorption is negligible, it has to be taken into account for higher surface coverages. Therefore, we present detailed theoretical considerations in order to quantify this effect.

2. Experimental setup

For our experiments, we use the nanofiber waist of a tapered optical fiber (TOF). We fabricate the TOFs by stretching a standard single mode optical fiber while heating it with a travelling hydrogen/oxygen flame [14]. Our computer controlled fiber pulling rig allows us to produce TOFs with a homogeneous waist diameter down to 100 nm and a typical extension of 1-10mm [15]. We have carried out a series of test measurements using electron microscopy showing that the fabrication of tapered fibers is highly reproducible and has an a priori accuracy of ±5% of reaching the target diameter. In the taper sections, the mode of the unstretched fiber is adiabatically transformed into the strongly guided mode of the ultrathin section and back, resulting in a highly efficient coupling of light into and out of the nanofiber. Beyond that, the surface of the nanofiber waist exhibits a very high degree of smoothness and cleanliness. This property which usually demands special sample preparation techniques [16] is crucial for spectroscopy of thin molecular films. For the presented measurements we used a 320 nm diameter nanofiber with a length of 1mm fabricated from a Nufern 460-HP fiber. This diameter ensures that only the fundamental mode is guided in the fiber waist for wavelengths longer than 450 nm, thereby matching the single mode cutoff wavelength of the unprocessed fiber. Moreover, this diameter yields the maximum sensitivity for absorption spectroscopy in the visible domain, as discussed in [12]. We obtain a typical transmission of up to 70% in the wavelength range between 450 and 650 nm. The measurements were performed on PTCDA molecules, taking advantage of their stability under evaporation at ambient conditions, their high quantum yield and the experimental and theoretical knowledge concerning their spectral characteristics [17]. Moreover, PTCDA molecules significantly change their spectral properties depending on their arrangement on the surface [18]. Therefore, these organic molecules are well suited as a model system for sensitivity studies.

Figure 1 shows the schematic experimental setup for fiber-based absorption and fluorescence spectroscopy. The molecules are deposited on the fiber waist by placing a crucible with PTCDA crystals below the fiber and by heating it to 325°C. By convection, sublimated molecules are carried to the fiber waist where they are adsorbed. The absorption of the deposited molecules is measured using a conventional absorption spectrometer configuration with a tungsten white light source (Avantes AvaLight-HAL) and a thermoelectrically cooled spectrograph (Andor SR 303i with DU920N-BR-DD detector). Fluorescence spectra are obtained upon excitation by a freely running diode laser at 406 nm (Sharp GH04020B2A). The laser light is coupled into the fiber upon reflection from a glass plate. The counterpropagating fluorescence is collected in transmission through the glass plate by the same spectrograph as for the transmitted white light. This configuration, in combination with a razor edge filter (Semrock BLP01-405R-25) placed directly in front of the spectrograph’s entrance slit, reduces the background signal due to the excitation light. To subsequently record absorption and fluorescence spectra, two shutters alternately open the path of the white light and the laser light. The shutter signal also triggers the spectrograph, thus synchronising the data aquisition. For timing reasons, the spectra are taken with an integration time of 420 ms.

 

Fig. 1. Scheme of the experimental setup. Molecules are deposited on the 320 nm diameter nanofiber waist of a tapered fiber from a heated crucible. Controlled by alternating shutters, either absorption is probed via transmission of white light from a tungsten lamp or fluorescence is excited by laser radiation. Both signals are detected by a CCD spectrograph.

Download Full Size | PDF

3. Theoretical considerations

In the following, we will derive an expression for the fluorescence signal at the fiber output. For this purpose, we have to take into account multiple processes of absorption from and emission into the guided fiber mode.

We consider a nanofiber with radius R and length L covered with n molecules, yielding a surface coverage, i.e., number of molecules per surface area, of θ=n/2πRL. The absorbance of light at a wavelength λ _exc propagating in the fundamental mode of this nanofiber is given by η(λ _exc)=-lg(P _sig(λ _exc)/P _ref(λ _exc)), where P _sig(λ _exc) and P _ref(λ _exc) are the transmitted powers at the excitation wavelength λ _exc in the presence and absence of molecules, respectively. According to [12], the absorbance η(λ _exc) can be approximated to

if the molecular absorption cross section σ(λ _exc) is much smaller than the effective area of the guided fiber mode A_eff(λ _exc) which is defined by

I _surf(λ _exc) denotes the fraction of the evanescent field intensity at the fiber surface exciting the molecules. Since the molecules lie flat on the surface [19], their transition dipole moment is oriented perpendicular to the radial electric field component of the guided fiber mode. Therefore, only the axial and azimuthal components of the electric field, calculated according to [7], contribute to I _surf(λ _exc).

A molecule at position z along the fiber waist absorbs the fraction σ(λ _exc)/A_eff(λ _exc) of the power P _sig(λ _exc,z):

For the fiber diameter and the wavelengths considered here, about 20%of the fluorescence of a single dipole emitter on the fiber surface is expected to be coupled back into the fundamental guided mode of the fiber, 10 % in each direction of the fiber [20]. As a result, the fluorescence emitted into the fiber mode by a molecule at a position z along the fiber waist is proportional to the absorbed power:

where the proportionality factor C(λ) ∝ q(λ)/A _eff(λ) includes the wavelength dependent fluorescence quantum yield of the PTCDA molecule q(λ) and the average fractional emission of the molecule into the guided fiber mode. The latter is an average over all possible orientations of the molecule on the fiber surface. Furthermore, it is determined by the intensity of the evanescent field at the fiber surface which is proportional to 1/A _eff(λ), see Eq. (2).

The spectral overlap between the absorption and emission spectra in our system results in a partial reabsorption of the emitted fluorescence by circumjacent molecules before it reaches the output of the fiber. In order to quantify this effect and to determine the fluorescence signal exiting the fiber at one end, we need to multiply Eq. (4) by a factor that takes into account all molecules covering the fiber waist between the origin z of the emitted signal and the end of the fiber (z=0). The total fluorescence detected at the fiber output can be calculated by integrating over all n molecules adsorbed on the fiber waist:

For small surface coverages, i. e. η(λ)+η(λ _exc)≪1, this equation can be approximated by

Hence, if the reabsorption is small enough to be neglected, we can infer the wavelength dependence of the fluorescence quantum yield of the PTCDA molecule directly from the measured signal. We only need to correct for the known wavelength dependent coupling to the fiber mode characterized by 1/A _eff(λ). For higher surface coverages however, reabsorption will become important and the detected signal has to be corrected according to Eq. (5).

For even higher surface coverages, multiple processes of emission and reabsorption may occur and can, in principle, also be corrected for. However, for the surface coverages below 1.5 % considered here, it will in general be enough to account for one more emission process, i. e., the reemission of the reabsorbed fluorescence, leading to an additional signal P ¹ _out(λ) at the end of the fiber. Therefore, the measured fluorescence signal at the fiber output P ^tot _out(λ) can be approximated by:

To calculate the reemission P ¹ _out(λ), we first have to consider the power P _reabs(z,z̄) that a molecule at position z̄ will reabsorb from the fluorescence emitted by a molecule at position z. For each wavelength λ̄, it will reabsorb the fraction σ(λ̄)/A _eff(λ̄) of the power of the fluorescence emitted by a molecule at position z. This power is given by Eq. (4), but taking into

account that part of the light has already been reabsorbed by other molecules on the fiber waist between the positions z and z̄. Integration over all wavelengths λ̄ yields:

The reemission of this molecule can be calculated in analogy to Eq. (4):

 

Fig. 2. Absorption (right) and corresponding fluorescence (left) spectra of surface-adsorbed PTCDA molecules during deposition. Five representative spectra within a range of surface coverages between 0.10 and 1.28% ML are shown.

Download Full Size | PDF

Taking into account reabsorption of the reemitted fluorescence light on the way to the fiber output and integrating over all molecules emitting and reemitting, we can obtain an expression for the total reemission measured at the fiber output (z=0):

4. Results and discussion

Figure 2 displays a series of fluorescence and absorption spectra, recorded in situ during the deposition of molecules on the fiber waist. We show representative spectra for five different surface coverages, ranging from 0.10% to 1.28% (see below) of a compact monolayer (ML) of flat lying PTCDA molecules, arranged in the herringbone structure of the (102) plane of the PTCDA bulk crystal [21].

The absorption spectra with 6 nm effective spectral resolution are taken with an optical power of about 3 nW of white light within the spectral interval between 400 and 600 nm, corresponding to an intensity of less than 1W/cm². Since the typical saturation intensities for single dye molecules are on the order of a few kW/cm² [22], we are far from saturating the molecules. All spectra show the typical effective vibronic progression for excitation between the ground and first excited electronic states of PTCDA. In order to determine the absorbance of molecules on the fiber surface, a reference spectrum is recorded before deposition. The absorption spectra have been corrected for the wavelength dependent coupling to the fiber mode characterized by 1/A _eff(λ _exc), so that the obtained spectra are directly proportional to the molecular absorption cross section σ(λ _exc). The shown spectra are the mean of two consecutive absorption spectra, one recorded 420 ms before and the other 420 ms after the corresponding fluorescence spectrum. Within this time interval, the deposition rate remains roughly constant, thus enabling us to determine the surface coverages underlying the measured fluorescence spectra by this averaging process. Using σ=4.1×10^-16 cm² [23] in Eq. (1), we infer from the absorbance of 0.04–0.51 at the absorption maximum at 2.47 eV that 0.8–10.6×10⁶ molecules cover the fiber waist. This corresponds to a surface coverage of 0.8–10.6×10¹¹ cm^-2 or 0.10–1.28% ML.

The fluorescence spectra are recorded upon excitation with a power of about 8 µW of laser light transmitted through the nanofiber at 406 nm, corresponding to the high energy edge of the absorption spectrum. They present maxima at about 2.44 and 2.26 eV, thus showing a vibronic progression as reported for PTCDA in solution [24]. The absolute peak positions are shifted with respect to the solution spectra, and the fluorescence exhibits a smaller Stokes shift. We attribute this fact to the interaction of the adsorbed molecules with the fiber surface. With respect to the luminescence spectra of perylene derivatives at room temperature reported in [25], the smallest coverages for which spectra can be measured are about two orders of magnitude smaller in our case. The shown fluorescence signal has been corrected for the wavelength dependent coupling to the fiber mode according to Eq. (6). The spectral response of our setup has been measured with a calibrated white light source (Avantes AvaLight-HALCAL).

The influence of self-absorption on the recorded spectra is shown in Fig. 3 for about 0.75% and 0.20% ML. It displays the same spectra as in Fig. 2, the spectra corrected for reabsorption (using Eq. (5)) and the ones corrected for reabsorption and reemission (according to Eq. (7) and Eq. (10)). For higher surface coverages like 0.75% ML, the reabsorption of fluorescence light significantly changes the spectrum. Only after correction for this effect, we retrieve the mirror symmetry between the fluorescence and absorption spectra which is characteristic for numerous molecular systems. For lower surface coverages like 0.20% ML, the effect of reabsorption and reemission is negligible. Hence, we conclude that for surface coverages below 0.2% ML, the spectral quantum yield of the molecule can directly be deduced from the measured fluorescence signal according to Eq. (6). The sensitivity of our method is therefore high enough to perform self-absorption-free fluorescence spectroscopy on isolated surface-adsorbed molecules. Finally, Fig. 3 shows that, due to the reduced efficiency of multiple coupling of the fluorescence back into the fiber, reemission has only a minor effect for both surface coverages considered.

5. Conclusions

Summarizing, we have shown that optical nanofibers are a highly efficient tool for fluorescence spectroscopy at interfaces. Excitation and detection of fluorescence via a single fiber mode yields an entirely fiber-based method that can be used for carrying out spectroscopy at a remote location. Using PTCDA molecules deposited on the fiber surface, we have shown that such fibers allow to collect and spectroscopically analyze the fluorescence of extremely low surface coverages, exceeding the sensitivity of previous studies by two orders of magnitude. For higher surface coverages, we have to account for reabsorption of the emitted fluorescence light by circumjacent molecules along the fiber waist. Based on theoretical considerations, we are able to quantify this effect and thus to extend our technique to a large range of surface coverages.

 

Fig. 3. Influence of self-absorption and reemission on the measured fluorescence signal for 0.75% and 0.20% ML. For 0.75%ML, the signal corrected for reabsorption (magenta) differs significantly from the measured spectrum (blue). For 0.10%ML (red and black, respectively), the difference is only marginal. The correction for reemission does not have a significant influence on the spectral shape, as can be seen in the cyan/green curves. The absorbance for both surface coverages is shown for comparison of the spectral shape.

Download Full Size | PDF

Our approach could, e. g., prove useful for bio-sensing, extending the possible applications from label-free detection using absorption measurements [26] to label-based detection using fluorescence measurements [27]. The interaction of the molecules with the nanofiber mode could possibly be enhanced further by using nanofiber-loop resonators [28], which might also allow us to implement a nanofiber-based dye laser [29]. Moreover, our setup enables us to measure both, absorption and fluorescence spectra for a given surface coverage of molecules. This property, in conjunction with the stable and reproducible fluorescence collection efficiency, makes our method ideally suited for analyzing and quantifying dynamic processes at surfaces. As an example, it should allow us to characterize the dynamics of the agglomeration of a constant number of molecules in situ at ambient conditions.