1. Introduction

Refractive index (RI) variations are currently monitored on Surface Plasmon Resonance (SPR) imaging devices in order to detect biological target/probe interactions in liquid media [1–9] or chemical interactions in gaseous media [10–12]. Indeed, the use of surface plasmons is an efficient way to probe these molecular interactions through the local RI variations they generate at the surface of the sensors. The standard approach to excite the surface plasmons of a thin metallic layer relies on the Kretschmann configuration [13]. In this configuration, an incident electromagnetic wave injected trough a high RI prism generates free electrons oscillations at the surface of a noble metal film in contact with the analyzed dielectric medium [14]. The adsorption of organic matter on the metal surface locally changes the RI leading to a plasmonic resonance shift. This detection method presents two essential advantages of real time and label free measurements.

A wide range of biosensors has been reported on waveguides such as optical fibers [14–23]. The principle of surface plasmon excitation is similar to the one described for prism-based configuration. The fiber core plays the role of the high RI dielectric prism. The optical cladding is removed on several millimeters in order to unravel the core which is coated with a thin metallic layer to form the active area of the biosensor. Plasmon resonance is obtained when the propagation constant of the surface plasmon modes matches the propagation constant of the guided fiber modes (see [24–27] for a detailed review). The coupling between surface plasmons and guided fiber modes is associated with a loss in the transmitted or retro-reflected light. Most of the SPR interrogation methods are based on the detection of this loss. Different interrogation methods have been considered: spectral interrogation of transmitted [28–30] or retro-reflected [31,32] light and intensity measurement of transmitted [33–35] or retro-reflected [36–38] light. In the following, we will focus on the SPR sensors based on intensity measurement. Depending on the configuration, the detection of the RI variations of a medium in contact with the active area is performed through the measurement of transmitted or retro-reflected global light intensity (i.e. spatially integrated value of the intensity). These kinds of measurements are quite easy to perform since they require the simple use of photodetectors.

Nevertheless, a change in RI near the active area does not only affect the global light intensity value but it can also affect its spatial distribution. The change in spatial distribution of the light intensity following a variation in RI around the sensor’s active area is mainly due to the different interactions between the optical modes, which propagate into the fiber and the generation of plasmons. Each optical mode can interact differently: some will see their intensity increase or decrease and others may not be influenced by the surface plasmons. Generally, when several modes can propagate into a fiber, all these different behaviors result, in average, in only a weak effect on the light intensity distribution. The difficulty to observe this effect explains why, even if optical modes are well-known and exploited in a variety of ways [39–43], the differential perturbation of these modes by plasmons was so far never used as a transduction method to probe molecular interactions through RI variations.

In this paper, we first present the preparation of an optical fiber fulfilling the required conditions to observe the effect of the plasmons/optical modes interaction on the intensity distribution. This phenomenon is then illustrated and exploited to improve the sensitivity of the optical fiber sensor. For this, the sensitivity to RI changes is determined by measuring both global intensity variations and by analyzing its spatial distribution. We demonstrate that by using intensity distribution it is possible to enhance the sensor performances: a 25-fold increase of the sensitivity is observed.

2. Experimental and theoretical methods

2.1 Microstructuration of the optical fiber

The optical fiber used in this study belongs to a bundle composed of 6000 individually cladded 3-4 µm diameter silica imaging fibers (Sumitomo Electric Industries, IGN-035/06 provided by Fiber and Technology) with a global diameter of 270 µm. The main characteristics of the optical fiber bundle have been reported previously [44]. Thanks to the cladding, the cross-talk between neighboring optical fibers should be limited. Consequently, each of the individual fiber can be used as an independent sensor. The behavior of a representative number of fibers of the bundle is illustrated in Figs. S1 and S2 in Supplement 1 and Visualization 1. In the following, we will concentrate on the response and sensibility of an individual fiber. The individual fiber core diameter is 3.5 ± 0.5 µm. The numerical aperture is 0.41 for a total fiber length of 25 cm. The RI of the core and the cladding are 1.497 and 1.438, respectively. One end of the fiber was microstructured in a truncated cone shape, called micropillar, in order to form the active area [21]. The structuration of this active area was obtained by a wet-etching procedure [45] of the optical fiber bundle that have been initially cleaved. One facet was left for 48 min in a buffered HF solution consisting of saturated NH₄F and 40% HF in proportion 6/1 and was then rinsed with deionized water (Caution: HF etching solutions are extremely corrosive and safety procedures must be followed accordingly).

2.2 Metallization process

A metallic layer was deposited with an electron gun sputter-coater. 10 nm of titanium were first deposited as an adhesive layer, followed by 570 nm of gold on the top of the micropillar, resulting in a 50 nm thick gold layer on the lateral surfaces. The layer thickness on the sides was optimized to couple the light effectively with the plasmons and the top layer thickness was sufficient for a mirror behaviour [21]. A specific support was designed to adapt the process to the cylindrical geometry of optical fibers. This support allows both rotation and tilt of the sample for homogenization of the deposited films during metallization. The thickness of the gold layer, which is a crucial parameter in the remote detection mode, was measured using the quartz microbalance incorporated into the sputter-coater.

2.3 Instrumentation and image treatment

The instrumentation dedicated to image acquisition and the whole process of signal treatment from image acquisition to sensitivity determination have been precisely described previously [21]. The sensitivity and resolution determination principles are explained hereafter. An optical setup with high-resolution camera (ORCA 4.0 LTE, Hamamatsu, Japan) able to image and quantify the retro-reflected light coming from the fiber bundle was conceived and allowed the characterization of the sensitivity and resolution of each individual fiber. The retro-reflected intensity $I(n )$ was imaged and measured at different RI n. $i(n )$, the normalized retro-reflected intensity, is defined as follow: $i(n )= \; \frac{{I(n )}}{{{I_{ref}}}}$ where $ {I_{ref}} = {I_{{H_2}O}}$ represents the retro-reflected intensity in pure water. The sensitivity S, defined as the slope of the normalized retro-reflected intensity per Refractive Index Unit (RIU) $\frac{{\delta i}}{{\delta n}}$ was measured. To characterize the performance of the system, the resolution ${n_{res}}$, i.e. the smallest RI variation detectable by the sensor, was determined from S and $ \sigma $ the maximum noise of normalized intensity among all given steps, then:

2.4 Characterization of the resolution and sensitivity to global RI changes

A range of glycerol solutions of known RI were prepared. The fiber gold-coated end-face was placed successively into the different solutions and twenty images were registered per RI value. The sensitivity S and resolution were calculated as described in the previous paragraph. All the experiments were performed at constant room temperature.

2.5 Optical modes simulation

The following parameters were used for the effective RI calculation and for the intensity profiles simulation: a core radius equal to 1.5 µm, a parabolic core RI profile from 1.497 (${n_{core}})$ to 1.438 (${n_{clad}}$), an incident light wavelength equal to 590 nm. The guided modes and their effective indexes were obtained from a numerical resolution of the Wentzel-Kramers-Brillouin (WKB) phase-tuning condition [46] using Scilab. The intensity profiles of the guided modes were simulated using Comsol 5.5.

3. Results and discussion

3.1 Optical fiber preparation

To observe a significant modification in the spatial distribution of light intensity following a change in RI near the active area, several conditions have to be fulfilled. First of all, the optical fiber must tolerate only a few modes (2 to 10) with intensity profiles as much different as possible. Secondly, the numerical aperture has to be large (NA > 0.2 for example) in order to have modes with sufficiently different effective indexes so that they behave differently. Finally, the geometry of the active area have to promote different mode/plasmon interactions. These conditions can be created in optical fibers with small cores (typically, less than 5 µm in diameter). But these small dimensions’ cores require imaging devices providing a sufficient resolution. An optical fiber fulfilling these conditions was prepared as follow.

A truncated cone also called micropillar was obtained onto the distal face of the optical fiber by wet chemical etching [16,17,22,45,47–51] (Fig. 1). Such a shape was fabricated by controlling the wet-chemical etching parameters: composition of the etching solution and etching time. Figure 1 displays the structure formed using the optical fiber as a starting material. Selective etching was achieved due to the different etching rates between the high-RI GeO₂ gradient doped core and the low-RI fluorine doped cladding. With the conditions described above, the fluorine cladding is primary dissolved and spatial structuration of the GeO₂ core as micropillar was obtained. The shape of the micropillars (i.e. height and angle) can be varied by changing the etching parameters. In the present study, micropillar with the following characteristics have been fabricated: base diameter of 3.5 ± 0.5 µm, half apex angle of 5° and top plateau diameter 1.2 µm in diameter. The lateral surface of the truncated cone was coated with 50 nm of gold which correspond to 570 nm of gold on the top mirror.

 

Fig. 1. A) Schematic representation of SPR phenomenon occurring in the active area of an optical fiber. B) Scanning Electron Microscopy (SEM) image of the active area geometry fabricated by etching the optical fiber bundles.

Download Full Size | PDF

3.2 Effect of changes in the RI on the spatial distribution of intensity

The response of the fiber to RI changes was investigated by immersing the gold-coated etched face in glycerol solutions with known RI ranging from 1.332 to 1.455. Excitation light was injected onto the cleaved facet (i.e. flat facet without microstructuration) of the optical fiber and then guided by total internal reflection in the core through the fiber. The light reaching the microstructured facet was confined to the etched core and senses the local RI. A fraction of the light was retro-reflected and collected by the same core, transmitted, and eventually detected at the cleaved facet. An image stack of the cleaved facet used to both inject the excitation light and collect the retro-reflected light was recorded for each RI value. The imaged retro-reflected intensity collected from the optical fiber core and its spatial distributions are presented in Fig. 2 for seventeen RI values. The corresponding movie representing the evolution of the 3D intensity profile is also available (Visualization 2).

 

Fig. 2. Effect of the RI on the retro-reflected intensity spatial distribution. Images 1 to 17 plot the spatial distribution of the retro-reflected intensity as recorded by the camera. The intensity profiles are presented under each image with the 3D Surface Plot plugin from ImageJ. RI corresponding to the media in which the optical fiber active area is immerged are presented on the scale in the bottom of the figure.

Download Full Size | PDF

The effect of the RI on the spatial distribution of the retro-reflected intensity is clearly visible: the intensity profile slowly evolves, from a “mountain” shape (one maximum) to a “valley” shape (two separate maxima). The relevancy of the « intensity spatial distribution » criteria compared to the « global intensity » criteria is highlighted in Fig. 3. For RI of 1.336 and 1.423, images 4 and 15 are obtained, respectively. The normalized intensities measured from these images are 98.9% and 98.8% and differ from less than 0.2%. However, the corresponding intensity profiles are clearly different (mountain vs valley shapes).

 

Fig. 3. Illustration of the relevancy of the « intensity spatial distribution » criteria compared to the « global intensity » criteria. For two RI giving very close global intensity values, corresponding intensity profiles are clearly differentiable.

Download Full Size | PDF

3.3 Characterization of the sensitivity to RI changes

The normalized retro-reflected intensity was measured for three different regions of the images. The three regions are presented in Fig. 4(A) (insert): region 1 includes the entire core; the measurement on this region corresponds to the measurement of the global intensity; region 2 represents the center of the core and region 3 a bottom part of the core. Regions 2 and 3 were defined on areas of highly variable intensity. The normalized retro-reflected intensity is plotted on each region in Fig. 4(B) as a function of the RI. The error bars presented in Fig. 4 were evaluated from $\sigma $ the gray level relative standard deviation observed on images corresponding to twenty individual measurements at the same RI.

 

Fig. 4. A) Representation of the different regions for which intensities and sensitivities were measured. B) Normalized retro-reflected intensity measured on regions 1, 2 and 3. C) Focus on the RI range with higher sensitivities for regions 2 and 3 versus region 1.

Download Full Size | PDF

The intensity variations are much higher on regions 2 and 3 compared to the variation observed on the whole core (region 1). On region 2 the intensity decreases with the RI whereas it increases on region 3. It corresponds in region 2 to the apparition of the valley while the mountain from the initial region 2 split and shifted towards region 3. The weak variation in global intensity results from an average effect and leads to a small sensitivity. The mean sensitivities for each region, for RI between 1.34 and 1.40 were calculated from the curves presented in Fig. 4(C). The uncertainties on the sensitivities (slopes) were estimated from the dispersion of the points. A sensitivity of 23 ± 12%/RIU is obtained on region 1, it reaches −341 ± 28%/RIU on region 2 and 584 ± 46%/RIU on region 3. The highest sensitivity corresponds in our experiment to a resolution ${n_{res}}$ = 8.10⁻⁴, $\sigma $ being equal to 0.16%. As the sensitivities present opposite signs on regions 2 and 3, they compensate on the whole core (region 1). By measuring the intensity on different regions, the sensitivity is increased by a factor of 15 in absolute value on region 2 and by a factor of 25 on region 3 compared to the global sensitivity. It is thus possible to drastically enhance the sensitivity of this sensor by analyzing the spatial distribution of optical modes intensity instead of analyzing the global intensity coming out from the core.

The transmission devices used by Chiu’s and Liu’s teams present sensitivities of 4000%/RIU and 1136%/RIU respectively [33,34]. Even if they were calculated from global intensity measurements, these sensitivities are higher than the one obtained in the presented study. However, the transmission configuration presents an important drawback: contrary to our device, it is not adapted to end fiber analysis and so incompatible with in situ interactions detection. The device of Kurihara [36], in reflection configuration, is based on the use of a single-mode fiber and a lock-in amplifier to improve the collected retro-reflected intensity. Nevertheless, it has a sensitivity of 320%/RIU which is lower than the one obtained here. Schuster’s and Suzuki’s teams obtained higher sensitivities of respectively 1000 and 2100%/RIU [37,38]. The high sensitivity of Schuster’s device can be explained by first, the use of a long-period single-mode fiber grating and secondly a active area of 1.5 mm which is far longer than our 15 µm-long active area. Concerning the device proposed by Suzuki’s team, they are not satisfied with a simple intensity measurement but use alternatively two light-emitting diodes whose well-chosen wavelengths correspond to a positive and a negative sensitivity. By taking the difference between the two opposite-signed signals, the final sensitivity is two-fold increased. Moreover, their active area is 20 mm long. However, as they use a multimode optical fiber, their sensitivity may be enhanced by imaging the optical modes’ intensities.

The reported improvement of sensitivity was obtained easily in our case as the spatial distribution of the retro-reflected intensity was imaged by a high-resolution camera. This was possible because of the weak optical modes number supported by the fiber. However, this analysis method could be adapted for highly multimode fibers by different ways. A priori, in a multimode fiber, only a few modes may interact with plasmons and their intensities’ variations may also compensate. Drowned among many insensitive modes (because too confined in the center of the core for example), the intensity variation of all sensitive modes may not be sufficient to significantly modify the spatial distribution of the total intensity coming out of the fiber. Two solutions may be used in this case: it may be possible by acting at the inlet of the fiber to select and inject only the sensitive modes using for example the modes selection process described by Kresing et al. [40], or to sort out sensitive and insensitive modes at the outlet of the fiber using spectrophotometry for example as reported by Nicholson et al. [39].

3.4 Modal interpretation

Although a quantitative study of the modal decomposition of the intensity is far beyond the scope of the present paper, we will now give a qualitative explanation of the observed intensity profiles variations. Using Scilab, and entering the parameters of the optical fiber (core diameter, core and cladding indexes, core index profile, wavelength), we were able to demonstrate that in theory 4 modes could be guided in the fiber. These are the linearly polarized LP₀₁, LP₁₁, LP₂₁, and LP₀₂ modes whose intensity profiles are shown in Fig. 5. We can assume that the 4 guided modes are orthogonal [52] and that the total intensity after propagation in the fiber is a linear combination of the intensities of each mode [53]. The effective indexes of the 4 modes propagating into the fiber were calculated by the Wentzel-Kramers-Brillouin (WKB) method [46] taking into account the fiber parameters and the light’s wavelength. Since the numerical aperture of the fiber is large (0.41), the effective indexes of the few guided modes (respectively equal to 1.4795; 1.4686; 1.4517; 1.4439 for LP₀₁, LP₁₁, LP₂₁, and LP₀₂) are very spaced over the interval $[{{n_{core}},\; {n_{clad}}} ]$ = [1.497, 1.438]. Consequently, one can expect very different individual responses from each mode.

 

Fig. 5. Intensity profiles of LP₀₁, LP₁₁, LP₀₂ and LP₂₁ modes. The 3D colored profiles and the 2D gray level views were obtained with Comsol 5.5.

Download Full Size | PDF

When looking at the intensity distributions of the modes propagating in the fiber, it seems that the starting intensity profile (first image in Fig. 2, ${n_{water}}$) mainly corresponds to a combination of LP₀₁ and LP₁₁ modes’ intensity profiles. As the RI increases (image 2 to 17 in Fig. 2), it seems that LP₀₁ mode intensity decreases, probably because the coupling of this mode with plasmons increases with the RI which results in a decrease of its reflected intensity. The intensity distribution shown in the last image (image number 17 in Fig. 2) appears to correspond mainly to that of LP₁₁ mode. In this experiment, we are able to identify only LP₀₁ and LP₁₁ modes. The fact that LP₂₁ and LP₀₂ are not visible may come from several reasons. First, we may not have a sufficient optical resolution and/or sensibility to distinguish them. Then, the number of guided modes is very sensitive to the radius of the fiber. Simulation was carried out for a radius of 1.5 µm, but if the radius decreases by even a few tenths of a micron, the number of guided modes also decreases. Given the low effective indexes of LP₂₁ and LP₀₂ (they are the modes with the lowest effective indexes), they may simply not exist in the fiber under study. These qualitative observations indicate that, with a sufficient resolution, it may be possible, from an image analysis, to reconstruct the total intensity profile from a linear combination of the intensity profiles of the four guided modes. This operation mathematically corresponds to a decomposition of the global optical signal in the base formed by the four guided modes.

4. Conclusion

In this study, we have demonstrated that a few multimodal optical fibers with an appropriate gold-coated micropillar-etched tip could reveal the effect of the optical modes/plasmons interactions on the spatial distribution of intensity. This fiber was used to detect remotely RI variations. The resulting intensity distribution variations were qualitatively interpreted from the intensity profiles of the four modes guided by the fiber. The sensitivity to RI changes was evaluated in remote detection mode performed by imaging through the etched optical fiber itself. The sensitivity was measured by taking into account the overall retro-reflected intensity coming from the entire core and by taking into account only selected regions of the retro-reflected intensity distribution. The exploitation of the intensity spatial distribution heterogeneity, due to different optical modes/plasmons interactions, allowed us to increase the sensitivity of our sensor by a factor of 25. This approach is particularly interesting for reflection-based biosensors because contrary to transmission-based biosensors, it is difficult to increase the size of the sensitive area in this configuration. Thus, this approach opens exciting perspectives to improve significantly the in situ detection of biological or chemical interactions in liquid or gaseous media.