Development of an integrated optic oxygen sensor using a novel, generic platform

Abstract

This paper describes the development of a generic platform for enhanced, integrated optic sensors based on fluorescence detection. The platform employs a novel optical configuration in order to achieve enhanced performance and has inherent multianalyte detection capability. The sensor element comprises a multimode ridge waveguide that has been patterned with an analyte-sensitive fluorescent spot, which is excited directly using a LED. The platform was applied to the detection of gaseous oxygen as a proof of principle. The sol–gel-derived sensor spots were doped with an oxygen-sensitive fluorescent dichlororuthenium dye complex and intensity-based calibration data were generated from the oxygen-dependent waveguide output. The sensor achieved a LOD of 0.62% and a resolution of less than 0.96% gaseous oxygen, which compares favourably with a similar, recently reported system. This device highlights the combination of inexpensive rapid prototyping techniques and a dedicated sensor enhancement strategy that together facilitate the production of an effective prototype sensor platform.

---

1. Introduction

There is a growing demand for low-cost, effective sensor platforms for both commercial and biomedical applications. A major trend in the development of such platforms is the combination of miniaturisation, higher information density and simplified fabrication protocols. This has led to the development of multianalyte sensor platforms in a variety of configurations. A key technical challenge is the ability to maintain efficient sensor performance when developing low-cost platforms.

Here, we present the development of a generic, integrated optic platform for sensors based on fluorescence detection. The novel optical design facilitates enhanced sensor performance through improved collection of the fluorescent signal and the sensor configuration provides multianalyte detection capability. The optical design derives from a previously reported theoretical analysis of fluorescence emission at a dielectric interface.¹ The analysis focused upon the emission of fluorescent dipoles embedded in a dielectric medium, deposited on a higher refractive index substrate. The fluorescence was found to be strongly anisotropic in nature and, furthermore, to be emitted preferentially into the substrate at angles exceeding the critical angle defined by the two media.

The sensor chip concept adopted here is illustrated in Fig. 1, which shows a ridge waveguide array on a planar substrate. Spots of fluorescent sensor material are deposited on one end of each waveguide and these are excited directly using a LED source. As predicted by the theoretical analysis,¹ the resultant fluorescence is efficiently coupled into the waveguides and propagates along their length to be detected at their respective endfaces by an appropriate detector. As a result of the anisotropic emission into the waveguide, the output intensity distribution at the waveguide endface exhibits a strong angular peak, which dictates the optimal detector orientation.² Direct excitation is an important design feature of the sensor chip as it is considerably more efficient than evanescent-wave excitation. This sensor configuration is inherently suited to multianalyte detection, where each sensor spot is sensitive to a different chemical or biochemical species.

Fig. 1 Sensor concept.

Fig. 2 SEM image of 2 adjacent SU-8 ridges.

The goal of this work was to demonstrate the effectiveness of this configuration in a proof of principle sensing application. The choice of fabrication protocol was a key aspect of sensor development and was dictated by the need to produce a prototype sensor quickly and efficiently without the need for complex or time-consuming process steps. For this reason, soft lithography and high-resolution pin-printing were employed for platform fabrication. In particular, soft lithography has been employed in a variety of so-called “rapid prototyping” applications.^3–5

As both a proof of principle and because it is a particularly important analyte with biomedical and environmental significance, it was decided to apply the optical configuration to the detection of gaseous oxygen, employing a single multimode ridge waveguide. The oxygen sensing mechanism is based on the quenching of fluorescence from a sol–gel encapsulated dichlororuthenium complex that is deposited on the waveguide. Changes in oxygen concentration result in a modulation of the detected output intensity, thereby providing the basis for the sensor. Multimode ridge waveguides are produced using the procedure known as micromoulding in capillaries⁶ (MIMIC), a soft lithographic technique. Photocurable, sol–gel-derived silica materials are introduced into a poly(dimethylsiloxane) (PDMS) mould and exposed to UV radiation. Sensor spots are then deposited on the resultant ridge waveguides by high-resolution pin-printing. Characterisation of the sensor chips in a calibration chamber demonstrates excellent sensor performance with a low limit of detection.

The work presented here is significant for a number of reasons. First, it demonstrates a generic enhancement strategy to develop simple, integrated optic sensor platforms that exhibit both efficient performance and multianalyte detection capability. In particular, it highlights the application of this strategy to produce a miniaturised oxygen sensor with excellent performance characteristics. Furthermore, it illustrates the importance of rapid prototyping techniques in producing miniaturised sensor platforms for evaluation.

2. Methodology

2.1. Fabrication of SU-8 ridge waveguide template

The use of SU-8 photoresist as a material for microsystems fabrication has been widespread in recent years due to the broad range of film thicknesses that can be deposited and the mechanical stability of the crosslinked material.^7,8 It is, therefore, an attractive choice for the fabrication of high quality microsystem templates for use as masters in soft lithographic applications. In order to produce the desired master for this application, a film of SU-8 was deposited on a silicon substrate by spin-coating. The coating process was designed to produce a film approximately 100 µm thick. After deposition, the film was soft-baked in a two-step contact hot plate process (5 min at 65 °C, followed by 20 min at 95 °C) and then allowed to cool slowly. This step was followed by exposure of the film through an appropriate mask using a Karl Suss MA 56 contact mask aligner (with i-line filter). The sample was then post-baked using a similar process to that described for the soft-bake, after which it was developed under sonication for 10 min. Fig. 3 shows a SEM image of the SU-8 ridges formed, taken using a Hitachi S-3000N scanning electron microscope. The near-square profile of the ridges is apparent with the average ridge measuring 100 µm ± 1 µm in width ×95 µm ± 2 µm in height.

Fig. 3 Principle of micromoulding in capillaries (MIMIC).

2.2. Fabrication of PDMS ridge waveguide mould

PDMS is the material most often used to fabricate the patterning element for soft lithographic applications. It is a cheap, non-toxic, elastomeric material available commercially as Sylgard 184 from Dow Corning. It is optically transparent down to approximately 330 nm, which is an important characteristic in this work as it facilitates the photopolymerisation of UV-curable materials confined within the mould. In order to fabricate the mould for this application, liquid prepolymer was cast against the SU-8 ridge waveguide master and allowed to de-gas in ambient conditions. The PDMS was then cured at 70 °C for 1 hour before being peeled from the master. It should be noted that the SU-8 master was first silanised to facilitate lift-off of the cured PDMS mould without causing damage to the master. The mould was then cut to provide an opening that facilitated capillary action (see Section 2.6).

2.3. Preparation of UV-curable sol

It was decided to employ UV-curable sol–gel⁹ as the ridge waveguide material. The choice of this material was influenced by several factors, including the ability to create dense sol–gel structures of high optical quality without the need for a high temperature annealing stage, the ease with which the material refractive index can be modified and the relatively high film thickness achievable when using such a material.

The substrate most commonly employed for the fabrication of ridge waveguides is silicon due to its ability to be cleaved cleanly, thereby producing waveguide endfaces of optical quality. However, owing to the high refractive index of silicon (n > 3.4) it is necessary to first deposit a low refractive index buffer layer in order to facilitate the propagation of light within the ridge waveguide by total internal reflection (see Fig. 4). For this reason, two sols were prepared, one to provide a buffer layer and the other to act as the higher index guiding layer.

Fig. 4 Schematic of typical sol–gel ridge waveguide configuration,

2.3.1. Preparation of guiding layer sol. The method used for the preparation of the guiding layer sol is well established and has been explained in detail elsewhere.^10,11 In brief, the sol was prepared by mixing tetraethoxysilane (TEOS), methacryloxypropyltrimethoxysilane (MAPTMS) and water to allow for the hydrolysis of the silica precursors. HCl was added to catalyse the reaction. Zirconium propoxide complexed with methacrylic acid (MAA) was added to increase the refractive index of the guiding layer relative to the buffer layer. The function of the complexing agent MAA was to prevent the formation of zirconia clusters in the sol due to the differing rates of hydrolysis and condensation of the zirconium and silica precursors. The reaction rates for silica and zirconia precursors were thus comparable, which resulted in homogeneous materials. The photopatternability of the sol was achieved by adding a photoinitiator (Irgacure 1800, Ciba Speciality Chemicals) to the sol and allowing it to stir until completely dissolved.

2.3.1. Preparation of buffer layer sol. The buffer layer sol was prepared in a similar manner to that described in the previous section without the addition of the complexed zirconium propoxide, as refractive index modification was not required for this sol. The typical refractive indices of the guiding and buffer layers were 1.51 and 1.49, respectively, as measured using a Metricon prism coupler (Metricon Corporation, USA).

2.4. Preparation of sensor sol

The doped sol–gel formulation employed for the production of an oxygen sensitive membrane is identical to one reported previously.¹² The oxygen-sensitive complex, ruthenium–tris(diphenylphenanthroline) (Ru(dpp)₃Cl₂), was dissolved in ethanol and mixed with pH 1 HCl whilst stirring. The sol–gel precursor, methyltriethoxysilane (MTEOS), was then added and the mixture was stirred for 4 hours at room temperature. The film was then deposited by pin-printing on the ridge waveguides and cured at 70 °C for 18 hours.

2.5. Patterning of UV-curable sol–gel ridge waveguides by MIMIC

Micromoulding in capillaries (MIMIC) has been used in several applications including the fabrication of field effect transistors¹³ and ridge waveguides.¹⁴ However, it has not, to our knowledge, been used to pattern UV curable sol–gel materials for optical sensing applications. The principle of the technique is illustrated in Fig. 4, which shows the use of capillary action to draw the desired material into a PDMS mould that is in conformal contact with the substrate to be patterned.

In this work, the PDMS mould was fabricated as described in Section 2.2. This mould was then brought into contact with a silicon wafer onto which a UV-curable sol–gel buffer layer had been deposited by spin-coating (see Section 2.4.1 for a description of the preparation of this buffer sol). A drop of UV-curable sol (guiding layer) was then placed at the entrance to the PDMS mould and proceeded to fill the mould by capillary action. Mould filling was typically achieved in less than 1 min.

As shown in Fig. 3, the SU-8 template consisted of an array of ridge waveguides approximately 100 µm × 100 µm. One can pattern a full waveguide array using this technique and, using multiple PDMS moulds, it is possible to produce many times the original number of waveguides simultaneously having fabricated just one template. However, for the purposes of this proof of principle application, it was necessary to pattern only one waveguide.

Both mould and substrate were subsequently exposed to UV radiation from a broadband source (UV Light Technology Ltd., UK) for 5 hours. This step was followed by a thermal crosslinking process that consisted of a 30 min hard-bake at 80 °C using a contact hot plate. The PDMS mould was then peeled from the substrate, leaving the crosslinked ridge waveguides on the buffer layer. The samples were then cleaved to provide endfaces of optical quality before sensor spot deposition. The typical length of a cleaved ridge waveguide was 8 mm. A major advantage of this technique is the ability to produce sol–gel structures of optical quality without the need for a high temperature annealing process, which would destroy the PDMS mould. This is only possible with photocurable materials.

Another important advantage of using MIMIC to pattern UV-curable sol–gel waveguides is the ability to produce much thicker waveguides than is possible by spin-coating, followed by conventional photolithography. By making the waveguides thicker, the number of modes they can support is increased, thereby leading to an increased signal level and improved sensor performance.

2.6. Deposition of sensor spots

High-resolution pin-printing was achieved using a Cartesian Technologies MicroSys 5100 MicroArrayer (Genomic Solutions, UK). With this device it was possible to deposit uniform sensor spots measuring 60 µm in diameter onto the upper surface of the sol–gel ridge waveguides. Fig. 5(a) shows a typical sensor spot deposited onto a waveguide by pin-printing while Fig. 5(b) shows the fluorescence from the same spot under illumination by a blue LED.

Fig. 5 (a).Image of pin-printed spot on upper surface of ridge waveguide. (b) Pin-printed spot under optical excitation.

Both images were recorded using an Olympus BX51M video microsope, with an appropriate gel filter employed to remove excitation light in Fig. 5(b).

2.7. Oxygen sensor system

The oxygen sensing mechanism is based on the well-established quenching of the fluorescent indicator dye Ru(dpp)₃Cl₂. This oxygen quenching process is described by the Stern–Volmer equation:

I₀/I = 1 + K_SV[O₂] (1)

where K_SV is the Stern Volmer constant, [O₂] is the oxygen concentration and the term I₀/I is the ratio of the the maximum fluorescence intensity, i.e., that obtained for 0% oxygen, relative to the fluorescence intensity at each oxygen concentration.

The experimental setup used in this work is shown in Fig. 6. The silicon substrate bearing the ridge waveguide was placed in a custom-made flowcell through which varying mixtures of oxygen and nitrogen were passed. Gas concentrations were regulated using mass flow controllers (Celerity, Ireland). Excitation of the sensor spots was accomplished using a blue LED (NSPB500S λ_max = 470 nm, Nichia, Japan), which was incorporated into the flowcell itself and the output fluorescence from the waveguides was detected using a linear detector array (LDA—Hamamatsu, Japan). The LDA was positioned at the optimum angle with respect to the waveguide endface predicted by theory in order to maximise capture of the output fluorescence. A Schott BG12 glass bandpass filter (Filter 1, UQG Optics, UK) was used to eliminate the higher wavelength tail from the blue LED. A LEE 135 gel filter (Filter 2, LEE Filters, UK) was used to eliminate excitation light from the detected output signal.

Fig. 6 Experimental setup.

The output from the LDA displayed the light intensity detected as a function of channel number, i.e., the position along the LDA at which the light was detected. This makes it possible to distinguish between the outputs of several ridge waveguides according to their position on the sensor chip. Therefore, using the same detector, it is possible to develop a multianalyte sensor based on a ridge waveguide array with each waveguide carrying a sensor spot for a different analyte.

3. Results and discussion

3.1. Sensor performance

Oxygen calibration data for single ridge waveguides were obtained using the characterisation system described in the last section. A representative Stern–Volmer plot yielded by the data obtained from a pin-printed ridge waveguide sample is shown in Fig. 7. The non-linear nature of the curve is a well established feature and has been attributed to the heterogeneity of the dopant environment in the sol–gel matrix. The data in Fig. 7 were recorded using a 10 s integration time and a 50 s sampling period, which meant that five intensity level readings were recorded every 50 s. The LDA software calculates the average and standard deviation associated with the five readings and plots the average values for each channel in the array (i.e., each photodiode). This results in an intensity trace corresponding to the output detected at the end-face of the waveguide. For the purposes of generating a calibration curve, the peak intensity for each trace was plotted against oxygen concentration.

Fig. 7 Stern–Volmer plot yielded by pin-printed sensor chip.

The 3σ noise level was obtained from the standard deviation of the intensity peak recorded by the LDA at 0% oxygen. From this, it was possible to calculate a limit of detection (LOD) for gaseous oxygen of 0.62%. Similarly, an average resolution of less than 0.96% over the range 0–20% was calculated. Taking into account an equivalent dissolved oxygen (DO) concentration of 9.2 ppm for air saturated water, this converts to a LOD of approximately 300 ppb and a resolution of less than 460 ppb over the range 0–9.2 ppm DO. It is important to note that a longer sampling time would lead to a reduction in the standard deviation and a concomitant improvement in the sensor LOD and resolution.

The results yielded by these sensors compare favourably with those presented in a recent publication¹⁵ reporting a similar platform, where a minimum resolution of 600 ppb was obtained over the range 0.8–24.8 ppm. These results were obtained using evanescent wave excitation of a sensing layer deposited onto the upper surface of a ridge waveguide.

Here, the use of sol–gel technology for the production of oxygen-sensitive spots also makes it possible to tune the sensitivity to specific ranges through correct choice of precursor (hence sol–gel film microstructure), lending a greater degree of flexibility to the sensor development process. This is a major advantage over other immobilisation strategies.¹⁶

Conclusions

We have demonstrated the development of a generic, novel platform for optical chemical sensors based on fluorescence detection. As a proof of principle, we developed an efficient sensor platform for the detection of gaseous oxygen that exploits both soft lithographic fabrication techniques and high accuracy micropatterning technology. The use of these rapid prototyping technologies is central to the development of proof of priciple sensor platforms and allow novel configurations to be evaluated quickly and efficiently.

The use of integrated optical excitation adds to the robustness of this system, which could be further enhanced through use of a dedicated detector, e.g., a Si photodiode or miniaturised photomultiplier tube. However, the most important aspect of this sensor configuration is its potential for multianalyte sensor development combined with efficient fluorescence capture. The latter design feature is frequently absent from fluorescence-based sensor platforms, a large fraction of which employ passive substrates.

The sensor presented here demonstrates a performance comparable to that of a recently reported system over the same concentration range and outperforms it in terms of resolution over lower concentration ranges. This was achieved by employing a simple fabrication protocol and has resulted in a compact, robust and miniaturisable sensor system that is well suited for lab-on-a-chip applications outside a laboratory environment.

Future work will involve the development of referenced fluorescence intensity-based systems for both single and multianalyte detection. More significantly, the platform will be used to develop an optical sensor based on fluorescence decay time monitoring using phase fluorometry.

This work highlights the effectiveness of combining dedicated sensor enhancement strategies with rapid prototyping techniques in order to efficiently develop and evaluate novel sensor platforms.

References

1. L. Polerecky, J. Hamrle and B. D. MacCraith, Appl. Opt., 2000, 39, 3968–3977.
2. J. F. Gouin, A. Doyle and B. D. MacCraith, Electron. Lett., 1998, 34, 1685–1687.
3. J. R. Anderson, D. T. Chiu, R. J. Jackman, O. Cherniavskaya, J. C. McDonald, H. K. Wu, S. H. Whitesides and G. M. Whitesides, Anal. Chem., 2000, 72, 3158–3164.
4. D. C. Duffy, J. C. McDonald, O. J. A. Schueller and G. M. Whitesides, Anal. Chem., 1998, 70, 4974–4984.
5. T. McCreedy, Anal. Chim. Acta, 2001, 427, 39–43.
6. E. Kim Xia and G. M. Whitesides, J. Am. Chem. Soc., 1996, 118, 5722–5731.
7. H. K. Chang and Y. K. Kim, Sensors Actuators A, 2000, 84, 342–350.
8. J. Zhang, K. L. Tan and H. Q. Gong, Polym. Test., 2001, 20, 693–701.
9. S. Aubonnet, H. F. Barry, C. von Bultzingslowen, J. M. Sabattie and B. D. MacCraith, Electron. Lett., 2003, 39, 913–914.
10. M. Oubaha, M. Smaihi, P. Etienne, P. Coudray and Y. Moreau, J. Non-Crystalline Solids, 2003, 318, 305–313.
11. S. I. Najafi, T. Touam, R. Sara, M. P. Andrews and M. A. Fardad, J. Lightwave Technol., 1998, 16, 1640–1646.
12. C. McDonagh, C. Kolle, A. K. McEvoy, D. L. Dowling, A. A. Cafolla, S. J. Cullen and B. D. MacCraith, Sensors Actuators B, 2001, 74, 124–130.
13. J. Hu, R. G. Beck, T. Deng, R. M. Westervelt, K. D. Maranowski, A. C. Gossard and G. M. Whitesides, Appl. Phys. Lett., 1997, 71, 2020–2022.
14. X.-M. Zhao, A. Stoddart, S. P. Smith, E. Kim, Y. Xia, M. Prentiss and G. M. Whitesides, Adv. Mater., 1996, 8, 420–424.
15. D. A. Chang-Yen and B. K. Gale, Lab Chip, 2003, 3, 297–301.
16. B. D. MacCraith and C. McDonagh, J. Fluoresc., 2002, 12, 333–342.

---