1. Introduction

Optical methods have found widespread use in monitoring of flow, reaction kinetics or particulate matter in microfluidic systems [1–3]. Measurements of absorbance, fluorescence, scattering and chemiluminescence have allowed monitoring of solute concentrations, reactor kinetics and nanoparticle aggregation, to name a few.

Measurements of refractive index in liquid systems are of importance in a number of fields, ranging from molecular biology (e.g. refractive index of proteins) to oceanography (e.g. refractive index of sea water). The refractive index of a liquid can be determined by refractometry, diffractometry, interferometry or spectroscopy [4–12]. The development of optics able to interrogate microfluidic systems has been particularly important in recent years, and several types of refractive index probes have been developed. Such systems utilize gratings [7, 8], capillaries [13–15], light propagation in and from micro and nanofluidic channels [16–19], optical fibers [20–22] and photonic crystals [23], and some of them have a relative resolution as low as 10⁻⁷ refractive index units (RIU). Combining two or more optical systems into a single setup opens up new possibilities, as in the cases for microfluidic refractometry based on gratings [24] or microinterferometry for spectral studies [25]. To this end, microfluidic refractometry is a potent technique allowing accurate monitoring of fluids [8, 26], and recent research has demonstrated self-calibration and automatic calculation of refractive index [27].

Optical sensing of dyes in solution is of interest from both basic research and industrial points of view, and systems have been built for accurate monitoring of refractive index and concentration [9, 10]. The polarizability of a solvent is related to its refractive index through the Clausius-Mossotti relationship, and it is known that the solvent polarizability may significantly influence the absorption and fluorescence spectra of a dye [28–31]. The excited and ground states of the dye are influenced by the polarity of the liquid, often through dipolar coupling, thereby leading to characteristic shifts in the spectral peaks of the dye. Bathochromic and hypsochromic shifts (red and blue shifts, respectively) occur due to different polarities of the ground and excited states, thus decreasing or increasing the difference in energy between these levels. In order to investigate the solvatochromism of dyes one needs to determine their spectroscopic properties as well as the polarity of the solvent. Typically, the refractive index of the solvent and the absorption spectrum of the solute are measured using separate experimental setups [29]. Here an optical system is presented that is able to do absorption spectroscopy and refractive index measurements simultaneously.

In Refs [16, 17]. interferometric systems using laser fringe shifts to determine the relative changes in refractive index of microfluidic volumes were investigated. The method for measurement of refractive index presented here is similar, but applies an interference pattern generated due to the scattered light from the top and bottom of the microfluidic channel with an obliquely incident laser beam, as opposed to the direct backscattering scheme of Refs [16,17]. However, as stated above measurement of refractive index alone is not always sufficient for a complete characterization. In addition to be able to measure the refractive index, the system presented here also applies a fiber-based spectrometer for the determination of the absorption spectrum of liquids in microfluidic channels. The two different measurement modes can be operated independently, and the novelty of the presented system lies in its ability to provide simultaneous measurements of both the absorption spectrum and refractive index. In order to test the applicability of the system, the solvatochromism of the laser dye Rhodamine 6G in glycerol solution is studied.

2. Experiment

A schematic drawing of the experimental setup is shown in Fig. 1 .

 

Fig. 1 Schematic drawing of the experimental setup.

Download Full Size | PDF

The homemade optical microscope utilizes a white-light LED as light source. The nearly collimated light beam is focused onto the microfluidic channel using a 10x Olympus objective (NA = 0.25). The light reflected by the microfluidic channel is collected by the objective and propagated to a fiber spectrometer (USB2000 + from Ocean Optics) trough a 2 m long multimode fiber of core diameter 100 μm (QP100-2-UV-VIS, Ocean Optics). The spectrometer was configured with a 10 μm slit, a grating of groove density 600 lines/mm, and a Sony ILX511b detector with 2048 pixels. The bandwidth of the spectrometer was approximately from 250 to 850 nm, and during the experiments we set 55 ms integration time and averaged over 50 subsequent recordings.

A 1.1. mm diameter circular beam from a 635 nm diode laser (VHK circular beam, Edmund Optics) is scattered off the microfluidic channel. The beam, which has a divergence of 0.70 mrad and is polarized parallel to the microfluidic channel (s-polarized), is reflected off the lower and upper part of the microfluidic channel to generate an interference patterns detected by a CMOS camera (DCC1645C from Thorlabs). The camera has 3.6 μm square pixels, 1280 x 1024 pixels and can record 25 frames per second. It is located 14 cm from the channel and the laser beam makes and angle of about 30° with the plane of the microfluidic channel. The videos recorded by the camera allowed one to record the dynamics of the interference pattern as it moved with changing refractive index of the medium within the channel. The microfluidic channel was made of cyclo-olefin copolymer (thinXXS microtechnology), was 1.6 mm thick containing channels of depth and width D = 300 μm. A piezo-driven pump was used to inject the fluid into the microfluidic channel through standard tygon tubing at a volume rate between 0 and 3 mL per minute. The microfluidic chip was placed on a metal holder to keep it at room temperature. The room temperature was kept at T = (293 ± 1) K for all the experiments, but was not measured to vary by more than about 0.1 K during a single experiment. The variations in temperature during an experiment did not result in any noticeable fringe shifts, so the fringe shifts must have been significantly smaller than one pixel.

The insets of Fig. 1 show the interference pattern recorded by the CMOS when the microfluidic channel was filled with water (a), and when it was filled with a water-solution containing 34 mM glycerol (b). The entire interference patterns shifted to the left when the refractive index increased, and the corresponding fringe shift was measured.

3. Results and discussion

As in Ref [17], we used glycerol to test the calibration of the refractive-index measurement system. Figure 2 shows the fringe shift versus glycerol concentration. The range 0 to 1.4 M corresponds to a refractive index change of approximately 1.4 × 10⁻². The solid line of Fig. 2 is a linear fit of zero bias (y = ac_g, where y is the fringe shift and c_g is the glycerol concentration) with slope a = (1350 ± 30) pixels/M. It should be noted that at the uncertainty was rather large for large glycerol concentrations ( ± 95 pixels). This is related to the fact that temporal refractive index fluctuations appeared frequently at high concentrations, thus making it harder to accurately identify the displacement of the fringes. At low concentrations this problem was not observed and the 3σ (i.e. 3 standard deviations) detection limit was found to be 10⁻⁵ refractive index units, which is more than sufficient to study the samples of interest. The refractive index versus glycerol concentration was also measured using a HI 96802 refractometer (Hanna Instruments), in order to check the linearity and calibration of the laser system. Good agreement between the two methods was found.

 

Fig. 2 Fringe shift versus glycerol concentration.

Download Full Size | PDF

A simple model for the observed interference pattern and its change with refractive index can be found by only considering the two plane waves propagating towards the camera as seen in Fig. 3 .

 

Fig. 3 Simplified ray model of the light propagation towards the camera.

Download Full Size | PDF

The two rays seen in the figure meet at the camera due to the finite curvature of the top and bottom of the microfluidic channel. The curvature of the microfluidic channel was due to the manufacturing procedures of the chip, and it was estimated that it caused a beam convergence of Δθ ≈10⁻³ rad. In the experiments, it was found that the laser beam had to be moved closer to one side of the channel in order to generate this convergence (see Fig. 3) and corresponding interference pattern (see Fig. 1). Let us now represent the two plane waves corresponding to ray 1 and 2 by $\exp (i{k}_{0}x\sin \Delta \theta /2)$and $\exp (i\pi +i2kD\cos {\theta }_t-i{k}_{0}x\sin \Delta \theta /2)$. Here

θ_t is the angle of refraction inside the channel, n is the refractive index, k = nk₀ = 2πn/λ₀ and λ₀ is the wavelength in vacuum. The two rays meet and form an interference pattern at the position of the camera. Assuming that the angle Δθ is small, such that sinΔθ/2 ≈Δθ/2, it can be shown that the change Δx in fringe position with a change Δn in refractive index is given by Δx ≈2D(cosθ_t/Δθ)Δn. If we assume that Δn ≈10⁻⁵, we find that Δx ≈4.5 μm, i.e. a movement of approximately one pixel. Thus, the very simple theory presented here predicts a smallest observable fringe shift in reasonable agreement with experimental data.

The spectrometer setup was tested by comparing the absorbance spectrum of rhodamine R6G in water with data found in the literature. However, in order to extract the absorbance one needs to take into account the background caused by light reflected or scattered from the surfaces of the microfluidic chip. Let us assume that some of the light is reflected or scattered from outside the channel, and is registered by the spectrometer as a power per area I_out. Similarly, some of the light propagates through the liquid inside the channel and will result in a power per area I_m on the detector. If these two contributions sum up incoherently, the total intensity reaching the spectrometer when the channel is filled with water can be approximated as I₁ = I_out + I_in. Upon adding dye of concentration c, and assuming that the light passing the liquid can be described by Beer-Lambert’s law, the intensity becomes I₂ = I_out + I_in.exp(−2εcL), where ε is the molar absorption coefficient channel. The transmittance is then given by T = I₂/I₁, or

The ratio γ = I_in/ I_out is the only unknown constant in Eqs. (1) and (2), and in this work it was determined by filling the channel with increasing concentration of materials with high molar absorbance coefficient until the transmittance saturates. When this happens, Eq. (1) becomes T = 1/(1 + γ), and allows us to extract the constant γ. We followed this procedure for rhodamine R6G, as seen in Fig. 4 . The transmittance at the wavelength of minimum transmittance is shown as a function of concentration in Fig. 4 (boxes), and the corresponding fit (solid line) using Eq. (1) with γ = 0.9 ± 0.1 and ε = (8.9 ± 0.5) × 10⁶ m⁻¹M⁻¹. The same procedure was followed by using iron oxide (Fe₃O₄) solution made according to Ref [32]. Iron oxide is known to have a completely flat spectrum in the spectral range of interest (i.e. absorbs equally well for all wavelengths between 420 nm and 620 nm). Upon increasing the iron oxide concentration until T reached saturation, it was found that γ = 1.1 ± 0.1. It was found that the ratio γ found for rhodamine R6G solution appears to be slightly lower than that found for iron oxide, which may be attributed to light scattered out of the channel by the rhodamine dye. The ratio γ is probably most reliably obtained from the measurements using iron oxide exhibiting nearly flat spectral characteristics. It should also be pointed out that the molar absorption coefficient of rhodamine 6G depends on the concentration in the range considered here, and using a constant ε is only an approximation.

 

Fig. 4 The transmittance versus concentration of rhodamine R6G.

Download Full Size | PDF

The transmittance of rhodamine 6G versus concentration was found by injecting 2.5 mM rhodamine R6G into the microfluidic channel, and then gradually adding water until the concentration was 0.07 mM. We then measured the corresponding transmittance, and the spectra at some of the concentrations are shown in Fig. 5(a) .

 

Fig. 5 The transmittance as a function of wavelength for four different rhodamine 6G concentrations (a). The transmittance is converted to molar absorption coefficient using Eq. (2) with γ = 1.1 (b)

Download Full Size | PDF

The local transmittance minimum at 500 nm is due to the formation of dimers in concentrated rhodamine 6G solutions. The dimeric structure disappeared when the rhodamine 6G concentration decreased, and the formation and spectral characteristics of such dimeric structures have been reported in detail elsewhere [33, 34]. Of most interest here is the local transmittance minimum at 525 nm, which can be attributed to the rhodamine 6G monomer. The molar absorption coefficient of rhodamine 6G is shown as a function of wavelength for different concentrations in Fig. 5(b). Here we used γ = 1.1 and L = 300 μm. The monomer spectral shapes and magnitudes are in good agreement with those found in Refs [33, 34], although the dimer peak at higher concentrations appears less pronounced in our work. It is known that the local environment influences the formation of the dimer peak, and it should be noticed that the available surface area in a microfluidic channel is larger than in a standard spectroscopic cuvette. More importantly, the liquid solution is just stopped for a small while (less than 30 seconds) during measurements, and flow near boundaries may have an influence on the dimer formation. Further studies of the dimer peak are outside the scope of the current work.

The strength of the described system lies in its ability to simultaneously do spectroscopy and probe bulk refractive index of a solution in a microfluidic channel. That is, the system allows one to measure the fringe shift simultaneously with the spectral measurements. The observed fringe shift upon adding rhodamine 6G to water was within the measurement uncertainty, which is not surprising since rhodamine 6G gives rise to a very small change in refractive index at 635 nm [35].

The total volume in the channel plus beaker was 3.4 mL. Glycerol was gradually added to the beaker in order to increase the concentration. At the same time, the rhodamine R6G concentration decreased. The resulting fringe shift upon increasing the glycerol concentration is shown in Fig. 6 . The solid line is a linear fit with zero bias of slope a = (1168 ± 80) pixels/M. Note that the value of the slope coincides with that of Fig. 2 to within the uncertainty stated, which is as expected since the change in refractive index due rhodamine 6G itself cannot be detected with current system. The transmittance versus wavelength for the increasing glycerol concentration (and decreasing rhodamine 6G concentration) is shown in Fig. 7 .

 

Fig. 6 Fringe shift as a function of glycerol concentration in presence of rhodamine 6G.

Download Full Size | PDF

 

Fig. 7 The transmittance of rhodamine 6G as a function of wavelength for seven different glycerol concentrations. The glycerol concentrations are indicated in the figure, while the rhodamine 6G concentrations are given in the text.

Download Full Size | PDF

Here the red line corresponds to a 0.29 mM rhodamine 6G spectrum in absence of glycerol (0 M glycerol), the green line to a 0.23 mM rhodamine 6G spectrum in 1.25 M glycerol, the blue line to a 0.19 mM rhodamine 6G spectrum in 2.03 M glycerol, the magenta line to a 0.19 mM rhodamine 6G spectrum in 2.54 M glycerol, the yellow line to a 0. 14 mM rhodamine 6G spectrum in 2.96 M glycerol, the cyan line to a 0.11 mM rhodamine 6G spectrum in 4.37 M glycerol, and finally the black line to a 0.10 mM rhodamine 6G spectrum in 4.61 M glycerol. It is seen that the addition of a glycerol-water mixture generally decreases the transmittance and removes the dimer peak as was also observed in the case of dilution by pure water (as was seen in Fig. 5). More interestingly, we also observe that the rhodamine 6G monomer peak exhibits a redshift of about 5 nm as the glycerol concentration increases from 0 M to 4.6 M. Such a redshift was not observed when adding pure water, and can be attributed to solvatochromism, wherein the polarity (i.e. refractive index) of glycerol causes a spectral shift [29]. Figure 8 shows the redshift as a function of glycerol concentration, where the boxes are measurements and the solid line is a linear fit (y = ax) with a = (1.0 ± 0.1) nm/M. The relatively large error bars ( ± 0.4 nm) are due to the fact that a simple fiber-spectrometer with low spectral resolution was used. Higher resolution could be obtained by using a spectrometer with smaller exit slit, at the cost of reduced light flux.

 

Fig. 8 The wavelength shift of rhodamine 6G as a function of glycerol concentration. The corresponding rhodamine 6G concentrations are given in the text.

Download Full Size | PDF

Although the hybrid optical system presented here has only been tested on rhodamine 6G, one could in principle study a range of dyes in various solvents. In general, it would allow one to probe refractive index (at a fixed wavelength) and the transmittance spectrum of any liquid as long as the parameters are not below the measurement sensitivities of the system. It should be noted that the setup only measured fringe shifts (corresponding to changes in refractive index), and the absolute refractive index may in principle have changed slightly from experiment to experiment if the temperature varied. However, in the current experiments we were mainly interested in the fringe shifts due to changes in glycerol concentration (and refractive index). Moreover, we did not observe any changes in the curve of Fig. 6 (within the error bars given) during different experiments, which suggest that temperature variations did not play a significant role.

4. Conclusion

We have demonstrated a system which allows one to simultaneously monitor the change in refractive index and spectral properties of a microscopic volume of liquid. We applied the system to study rhodamine 6G in glycerol, and found that while the interferometric laser system allowed precise monitoring of the change in refractive index, the spectrometer system gave data about the transmittance of the sample. This is the first time these two parameters have been obtained simultaneously in microfluidic channels.

The hybrid sensor system presented here may have several other applications, and is therefore expected to become a useful tool for studying refractive index (at one wavelength) and spectroscopic data simultaneously. One possible application under consideration for future studies would be to monitor the refractive index and localized surface plasmon resonance simultaneously. It is known that localized surface plasmons can be probed by visible light spectroscopy, and are very sensitive towards local changes in refractive index [36]. Thus, simultaneous measurements of bulk and local refractive index may be a very potent combination.