1. Introduction

The design of new types of optical miniature sensors, which can be used for the determination of physical, chemical, and biological properties of surrounding medium, is of great importance both for practical applications and scientific research [1–18]. One of the ways to sense an object optically is to measure the value and variation of its refractive index. The latter can be determined from the transmission (reflection) spectrum of the light propagating through the object. Usually, the sensitivity of detection grows with the length of the path that light propagates in the tested medium. Therefore, accurate sensing is more problematic for small objects. There are two known solutions for this problem. The first solution is based on the resonant propagation of light when the probe light passes the tested object many times. In this case, very small changes in refractive index can noticeably shift the resonant peaks of the transmission spectrum. The resonant microspheres [1,2], microrings [3–7], microdisks [8,9], microcylinders [10,11], and microcapillaries [12–15], as well as the Bragg grating sensors [16,17] are related to this group of resonant sensors. The second solution is based on the strong nonresonant attenuation of light caused by scattering and radiation loss. The simplest example of this kind is a microfiber evanescent refractive index sensor [18], which becomes very responsive when the refractive index of the liquid surrounding the microfiber approaches that of the microfiber. Some of optical sensors, such as the sensors based on Bragg gratings, can exhibit both the resonant and attenuation features [16, 17].

This paper presents the miniature optical refractive index sensor of liquids that can combine the features of the resonant and attenuation sensors. The design of this sensor is based on the recently introduced liquid-core optical ring-resonator sensor (LCORRS) [12–15]. The LCORRS sensor is a silica capillary coupled to the waist of an optical fiber taper. The light propagating along the taper waist excites the whispering gallery modes (WGMs) inside the capillary wall. The capillary wall has thickness of a few microns. For this reason, the evanescent part of the WGMs penetrates into the interior part of the capillary and probes the liquid flowing inside the capillary. The LCORRS has been demonstrated as a very efficient tool for chemical and biological sensing [14, 15]. However, significant problems still exist in simultaneous increasing of the sensitivity of LCORRS and its robustness. In fact, in order to increase the sensitivity, the wall of the capillary has to be decreased, which leads to fragility of this device. In addition, the waist of the optical fiber (which can also be replaced by a lithographically fabricated waveguide [13]) is prone to contamination and corrosion, which reduces the transmission power of the detected light. This paper introduces a sensor that solves this dilemma. Briefly, we immersed the LCORRS into a curable low index polymer, which was further solidified by UV curing. Therefore, the taper and the capillary surfaces of our device were fully protected from corrosion and contamination. Crucially, because the capillary was imbedded into the solid polymer matrix, we were able to etch the capillary wall down to submicron values and eventually fully eliminate it. In our sensor, a significant part of resonant WGMs is located inside the tested liquid (rather than propagates primarily along the capillary wall, as in the LCORRS [12–15]), which considerably increases its sensitivity. In order to emphasize this fact, we call this device the liquid ring resonator optical sensor (LRROS). Section 2 describes the fabrication principles of LRROS. Section 3 demonstrates the performance of an LRROS with a very thin silica capillary wall and an LRROS without silica wall. After removal of the wall, the sensitivity of our device was as high as 800 nm/RIU. Finally, Section 4 includes discussion and summary of the results.

2. Fabrication of the LRROS

The proposed LRROS consists of a segment of a capillary fiber (CF) coupled to a microfiber (MF) and immersed into the cured low-index polymer as illustrated in Fig. 1. The CF was fabricated by drawing from a hollow silica preform with a wall having less than 4% thickness nonuniformity. The CF was fabricated by drawing from a hollow silica preform with a wall having less than 4% thickness nonuniformity. The CF, which was fabricated from this preform, had the outer radius 106 μm and wall thickness 13 μm. We suspect that the wall thickness nonuniformity of the CF was proportional to that of the preform, i.e. it was less than 0.04∙13=0.52 μm. The biconical taper with the MF waist of diameter 1.7 μm was fabricated from a standard telecommunication fiber using a CO₂ laser microfiber drawing technique [19]. The ends of the taper were connected to the broadband light source and to the optical spectrum analyzer (OSA) as illustrated in Fig. 1. A segment of the CF was striped from the coating and immersed into a curable low index OFS specialty polymer having refractive index 1.384 at 1550 nm wavelength of light. The MF was also immersed into the polymer and positioned normally to and in direct contact with the CF. After that, the polymer was solidified by UV curing. The refractive index of the polymer (n_p = 1.384 at radiation wavelength 1540 nm) was noticeably smaller than the refractive index of the silica CF and MF (n_s = 1.444 at the same radiation wavelength). For this reason, the light propagating along the MF excited the WGMs in the CF, which were observed in the transmission spectrum measured with the OSA. The transmission spectrum of our device at this step of fabrication process is illustrated in the interval from 1540 to 1545 nm by the black curve in Fig. 2(a). Next, the interior of the CF was etched down to the thickness of about 1 micron or less using the HF solution. The thickness of the CF was estimated by monitoring the transmission spectrum. For a relatively thick wall, the etching process did not affect the transmission spectrum of the CF and it remained the same as shown in Fig. 2(a), black curve. However, the spectrum began to show noticeable change when the thickness became less than a few microns. In the previous design of LCORRS [12–15], the etching was stopped when the light still propagated primarily along the CF wall. The evanescent component of light propagating in the tested liquid was relatively small (this corresponded to the wall thickness of 3-4 microns). Achieving a thinner wall was problematic because the fragility of the capillary increased dramatically with a decrease in the wall thickness. In our device we do not have this problem because the CF wall is supported by the solid polymer matrix. Therefore, we were able to fabricate the LRROS with extremely thin walls and the LRROS where the CF was eliminated completely.

 

Fig. 1. Illustration of an LRROS

Download Full Size | PDF

3. Measurement of the LRROS sensitivity

If the refractive index of the tested liquid, n_c, is greater than that of the polymer n_p, then, with thinning of the CF wall, the WGMs move from the CF wall into the tested liquid. As a result, the sensitivity of the LRROS grows dramatically. However, for n_c < n_p and in the absence of the CF, the WGMs no longer exist and oscillations in transmission spectrum disappear. For this reason, in our first experiment the CF was not removed completely. This allowed us to create a highly sensitive LRROS, which worked both for n_c > n_p and for n_c < n_p. In pratice, we periodically interrupted the etching process in order to measure the transmission spectrum of the LRROS for the low refractive index liquid with n_c = 1.296 and to ensure that the transmission spectrum still had noticeable oscillations. The etching was stopped when the oscillations at 1.296n_c= were as depicted by the black curve in Fig. 2(b). The red curve in Fig. 2(a) shows that at this stage of etching the transmission oscillations for the empty core (i.e. for n_c = 1) disappeared. We measured the sensitivity of our device using the Cargille Labs optical refractive index matching liquids. At the wavelength of 1.54 μm, the refractive indices of these liquids were calculated using the Cauchy equations provided by the Cargille Labs. The liquids were delivered into the CF with a vacuum pump. Figures 2(b) and 2(c) show the representative transmission spectra of liquids with refractive indices 1.296, 1.390, 1.444, 1.496, and 1.606. Most of the curves in Fig. 2 have a characteristic double-dip periodic structure. We suspect that two dips correspond to two polarizations of light similar to the ring resonator (see e.g. [6]). Figure 2(d) shows an example of our measurement of the sensitivity. The spectrum of the LRROS was recorded for close refractive indices n ₁ = 1.444 and n ₂ = 1.446. The sensitivity is expressed through the shift of the dips in these spectra, Δλ , as S = Δλ/Δn. For the spectra depicted in Fig. 2(d), Δλ ≈ 1 nm and Δn = n ₂ ? n ₁ = 0.002 resulting in S ≈ 500 nm/RIU. It is seen that the transmission spectrum of the matching liquid with refractive index 1.444 (blue curve in Fig. 2(b)) well matches the transmission spectrum of unetched silica CF (black curve in Fig. 2(a)) with the same refractive index. A small shift between these curves of 0.35 nm can be explained by inaccuracy of the refractive index of glass and the matching liquid, which are given here with 0.001 RIU accuracy. Actually, with the sensitivity of 500 nm/RIU, the 0.35 nm inaccuracy in wavelength corresponds to the inaccuracy in refractive index of 0.35/500=0.0007 RIU. The plot of sensitivity of the LRROS is shown in Fig. 3, black curve. For the test liquids with larger refractive indices, n_c > 1.5, the sensitivity was as high as ~ 700 nm/RIU. For smaller indices, the sensitivity decreased and was ~ 25 nm/RIU at n_c = 1.296. The sensitivity achieved at larger indices was comparable with the maximum possible sensitivity at refractive index n_c and wavelength λ, S_max = λ/n_c [15]. The dotted curve in Fig. 3 shows this dependence for λ = 1.54 μm.

 

Fig. 2. Transmission spectrum of the LRROS for different indices of the tested liquids. (a) - Unetched CF (black) and etched CF filled with air (red); (b),(c) - Examples of transmission spectra for different refractive indices of liquids; the values of indices are shown on the figures. (d) - Illustration of measurement of the LRROS sensitivity.

Download Full Size | PDF

 

Fig. 3. The LRROS sensitivity measured when the CF was not removed completely (black curve) and for the removed CF (red curve). The dotted violet curve is the maximum theoretical limit of sensitivity. The dashed blue curve is the sensitivity suspected for the Teflon AF matrix.

Download Full Size | PDF

 

Fig. 4. Transmission spectra of the LRROS with the removed CF at refractive indices shown in the figure.

Download Full Size | PDF

In order to further increase the sensitivity of the LRROS, we continued etching the CF down to what we believed corresponded to complete removal of the silica capillary. We stopped etching when the transmission spectrum of the device at refractive indices below the index of the polymer matrix, n_p = 1.384, did not have the WGM oscillations. The corresponding spectra are depicted in Fig. 4. Interestingly, at lower indices, the transmission exhibited significant attenuation with the change of the refractive index. Thus, at indices below n_c ~ 1.4, our device performed as an attenuation (nonresonant) sensor similar to the microfiber sensor of Ref. [18]. For the empty channel (n_c = 1), the transmission spectrum (black curve in Fig. 4) showed more than 4 dB increase in transmission power compared to the similar spectrum of the under-etched device (red curve in Fig. 2(a)). This may be due to the reduction of the power loss caused by radiation from the MF into the silica wall of the CF (the mechanism of radiation is similar to that described in Ref. [20]). At higher indices, the WGMs oscillations re-appear as illustrated by the transmission spectra at n_c = 1.41 and 1.43 in Fig. 4. The corresponding sensitivity of the LRROS is given by the red curve in Fig. 3 obtained by the measurement of the shifts in WGM transmission spectra as described above. This curve demonstrates the increase of the sensitivity at larger indices up to ~ 800 nm/RIU and also a significant growth of sensitivity at lower indices. It should be noted that the resonant transmission spectrum of light propagating in a polymer is by an order of magnitude more sensitive to temperature variations than the light propagating in glass [14]. For this reason, for highly accurate measurement of the refractive index, the temperature variations should be controlled and suppressed.

4. Discussion and summary

The LRROS, which is introduced in this paper, shows very high sensitivity enabling very accurate measurement of refractive index variations. For example, for the sensitivity S = 800 nm/RIU and with 1 pm wavelength measurement resolution, the accuracy of (800 nm/RIU)/pm ~ 10^-6 can be achieved. The shapes of the transmission spectra measured for liquids with different refractive indices were qualitatively different as illustrated in Fig. 2. For this reason, the computer pattern recognition would allow us to achieve very high accuracy in measuring variations of the refractive index as well as its absolute values.

Using the developed fabrication technique, it is straightforward to create a lab-on-the-chip device consisting of many LRROS elements. In application to the LCORRS, this concept has been discussed previously [15]. The advantage of our design is that it has the significantly increased sensitivity and robustness, and that it ensures protection from contamination and corrosion.

It is possible that the sensitivity of LRROS can be increased even more by further modifications. In particular, the polymer matrix with refractive index 1.384 n_p = used in our experiments can be replaced by a very low index Teflon AF with n_p = 1.291. We anticipate that, then, the plot of sensitivity, the red solid curve in Fig. 3, will be shifted to the left towards lower refractive indices by approximately the difference of these indices, 1.384 – 1.291 ~ 0.1. This would significantly increase the sensitivity at lower refractive indices as illustrated by a dashed blue curve in Fig. 3. In the regime of attenuation sensing (Fig. 4) the sensitivity could be increased by using a thinner MF which causes greater radiation loss [20] and stronger dependence on the refractive index n_c.