Enhanced evanescent wave sensors based on sol-gel-derived porous glass coatings

Sensors and Actuators B, I1 (1993) 29-34 29 Enhanced evanescent wave sensors based on sol-gel-derived porous glass coatings* B. D. MacCraith, Sbhool of Physical Sciences, Dublin City Uniwrsity, Glasnevin, Dublin 9 (Ireland) Optical chemical and biochemical sensors based on evanescent wave interactions on optical fibres are of great interest because of their advantages in a number of applications. Two mechanisms are used: direct spectroscopic sensing and reagent-mediated sensing. The theoretical design criteria underlying both approaches are presented. A novel approach is proposed whereby porous sol-gel glass coatings are used to enhance the performance of both types of evanescent wave sensor. The particular advantages obtained by applying the technique to a pH sensor based on dye fluorescence and a methane gas sensor based on infrared absorption are presented. 1. Iatruductiun When light propagates in an optical fibre or wave- guide, a fraction of the radiation extends a short dis- tance from the guiding region into the medium of lower refractive index which surrounds it. This evanescent field, which decays exponentially with distance from the waveguide interface, defines a short-range sensing vol- ume within which the evanescent energy may interact with molecular species. Optical waveguide sensors for chemical and biological species based on such evanes- cent wave (EW) interactions have attracted consider- able research interest [l]. Two distinct approaches may be adopted in these sensors. First, the evanescent wave can interact directly with the analyte if the interrogating wavelength coincides with an absorption band of the species. Such direct spectroscopic EW sensors are of particular interest in the infrared spectral region, where many species absorb strongly. Alternatively, an inter- mediate reagent, which responds optically (e.g., by fluorescence or absorption change) to the analyte, may be attached to the waveguide. Often reagent-mediated EW sensors provide greater sensitivity than direct spec- troscopic devices. In this paper theoretical consider- ations that indicate the critical design parameters for EW sensors are presented. The use of sol-gel-derived porous glass coatings on both types of sensor is pro- posed. The resulting advantages and enhanced perfor- mance are described. Although the examples given concentrate on optical fibres, the principles are equally applicable to planar waveguide devices. ‘Plenary paper. 0925-4005/93/$6.00 2. Evanescent wave sensors Tbe motivation for adopting the evanescent wave approach in fibre-optic chemical/biochemical sensing derives from a number of advantages offered by the technique in particular applications: (i) Because the interrogating light remains guided, no coupling optics are required in the sensor region and an all-fibre approach is feasible. Furthermore, consider- able miniaturization is possible. (ii) By controlling the launch optics it is possible to conline the evanescent field to a short distance from the guiding interface and thereby discriminate to a large extent between surface and bulk effects (see Section 3). This is particularly important in some applications that involve surface interactions, e.g., fluoroimmunoassay PI. (iii) The technique can provide enhanced sensitivity over conventional bulk-optic approaches. For example, fibre-based EW absorption devices are more sensitive than bulk-optic ATB (attenuated total reflection) crys- tals by virtue of the greater number of reflections per unit length (or, equivalently, due to the greater power in the evanescent field). (iv) It is often diacult or inconvenient to perform accurate absorption measurements on highly absorbing or highly scattering media. Fibre-optic EW spec- troscopy is suitable for such samples because the effec- tive path length is so small and the technique is much less sensitive to scattering. (v) If an optical fibre is configured to be sensitive to EW interactions all along its length, or at discrete zones, then fully or quasi-distributed sensing is possible. This would enable monitoring of the spatial profile of @ 1993 - Elsevier Sequoia. All rights reserved 30 an analyte concentration over substantial distances. Similarly, the line average of a species concentration could be acquired. (vi) In contrast with conventional distal-face op- trodes, the EW approach affords the sensor designer greater control over interaction parameters such as interaction length, sensing volume and response time. EW sensors are not without difliculties. Chief among these is the problem of surface fouling which, if signifi- cant, can reduce sensitivity and necessitate frequent recalibration until the sensor is no longer viable. A number of compensation techniques have been pro- posed, but these have not been implemented experimen- tally [3]. If suitable techniques are not found, then commercial EW devices will be restricted, in some applications, to short-term or disposable use. 3. Theoretical background An awareness of the critical parameters that deter- mine the extent of EW interactions is important for the optimal design of EW sensors. Control over the degree of penetration of the evanescent wave into the low- index medium is important in some applications. This quantity is often characterized by the penetration depth, dr, which is the perpendicular distance from the interface at which the electric field amplitude, E, has fallen to l/e of its value, E,, at the interface, i.e., E = E,, exp( -z/d,) (1) The magnitude of the penetration depth is given by dp = A 2nn,[sin’0 - (n,/n,)2]“2 (2) where 1 is the vacuum wavelength, 0 is the angle of incidence to the normal at the interface, and Q, n2 are the refractive index values of the dense and rare media, respectively. Although dp is typically less than 1, it is clear from eqn. (2) that its value rises sharply as the angle of incidence approaches the critical angle 0, = sin-‘(n&i,). When considering absorption of the evanescent wave, the quantity of evanescent power which can interact with the analyte is a critical parameter. In the case of optical-fibre EW sensors this quantity is closely related to the fraction, r, of the total guided power that resides in the cladding region, i.e., r = Pclad iPtot (3) This fraction is determined to a large extent by the fibre V-parameter, which is given by V&lii* I, (4) where a = fibre core radius, numerical aperture NA = (n,’ - n22)“2, n, = core refractive index and n2 = cladding (analyte) refractive index. The depen- dence of r on V for individual modes in optical fibres is shown in Fig. 1 [4]. From this Figure it is evident that: (i) substantial values ( >SO%) for r can be achieved in single-mode (V < 2.405) fibres; (ii) values of r are maximized for modes close to cut-off (which, in ray-optics terminology, are higher- order modes); (iii) for high V-number values (V $1) the average fractional power in the cladding is very low. Although the high sensitivity achievable with single- mode fibres is very useful in some applications, highly multimode fibres (V 9 1) are most frequently used in EW sensing because ‘of their much higher power throughput and ease of handling. Gloge [4] has shown that the average value of r is given by 4,,@3V for weakly guiding (n, N n2) multimode fibres in which all modes are propagating. Although many EW sensors do not comply with the weakly guiding condition and are not mode-filled, this expression gives an indication of the fraction of guided power available for EW sensing. For example, it yields a value of r = 0.0008 for light of wavelength 500 run in a typical silica (n, = 1.46) multi- mode fibre core (n = 300 )un) in an aqueous (n2 = 1.33) environment. Values of r less than 1% are typical in multimode EW sensors. Driver et al. [5], and other workers [ 6,7], have shown that EW absorption is linearly proportional to the bulk absorption coefficient, a, of the analyte under certain strict conditions. In particular, Driver et al. define an effective evanescent wave absorption coefficient, r, such that the expression I = 1, exp( - TL) (5) 0.6 0 2 4 b 8 IO I2 Fibre V-Parameter Fig. I. Plot of the dependence of the ratio P,,JP,,, on fibre V-parameter (taken from ref. 4). 31 describes the intensity transmitted through an EW ab- sorption region of length L, where 1, is the intensity transmitted in the absence of an absorbing species. They have also shown in the limit of weak absorption that where NA, is the numerical aperture of the launched radiation. This expression is true only for a uniform distribution of the launched radiation propagating at angles well away from the critical angle in the sensing region. It is significant that the V-’ dependence of EW absorption sensitivity predicted by Gloge [4] appears here again. Equation (6) allows one to determine the approximate attenuation for a given sensor system. If, however, the launched radiation approaches the critical angle, the dependence of evanescent absorption on ana- lyte concentration becomes non-linear. This can be understood in terms of the great discrepancy between the effective absorbing path lengths of high- and low- order modes. If fluorescent species are located within the evanescent volume, a fraction of the fluorescence excited by the evanescent wave is coupled back into guided modes of the optical fibre. The efficiency of collection of this fluorescence by the fibre has been the subject of a number of theoretical and experimental studies (8,9]. Marcuse [lo] has pointed out that fluorescent light in the cladding region couples to guided modes via the evanes- cent field tails of these modes. Consequently, modes near cut-off collect more light than more strongly guided modes. Therefore the collected signal should increase roughly linearly with V for fluorescent material dis- tributed throughout the cladding because the number of modes near cut-off increases linearly with V. If, however, the fluorescent material is located di- rectly at the core-cladding boundary, the collection efficiency should increase in proportion to the total number of modes, which is proportional to V2. This occurs because all modes have reasonably high field intensities at the core-cladding interface in step-index fibres. Love and Button [8] have shown experimentally and theoretically that the total fluorescent signal ob- tained from one end of a short length of fibre sur- rounded with fluorescent material can be described by If a &Z&z % ( > ‘sin* B,,,(NA)-4 where E is the absorption coefficient of the fluorescent material, I, is the optical source radiance, r,, is the radius of the launch spot, 0,,,,, is the maximum external angle of the launch and acceptance cone, and all other terms are as used previously in this text. It is important, however, to interpret this expression carefully with due awareness of the conditions under which some of the dependencies apply. For example, the eighth-power de- pendence of sin 8,, applies to launch numerical aper- tures that yield rays close to the critical angle. The theoretical considerations presented in this Sec- tion identify many of the critical parameters involved in optimizing the performance of EW sensors based on absorption or fluorescence. Most workers have concen- trated on those aspects which involve the launch/detec- tion optics and the fibre dimensions. The sensor numerical aperture (or the related V-number) appears in all the relevant expressions presented above. It is usual to accept that the analyte refractive index to- gether with that of the fibre core determines the sensor NA. It is possible, however, to overcome this apparent limitation. A novel approach is proposed here, which involves the use of porous glass coatings fabricated by the sol-gel process [ll]. By taking advantage of this technique it is possible to produce coatings with a range of refractive index values and porosities [12]. Accord- ingly, the potential exists to optimize the sensor re- sponse via the numerical aperture (or the related V-number) dependence in a manner that still allows the analyte to enter the evanescent sensing volume. In order for this approach to be justifiable, it is clear that the enhanced sensitivity must outweigh considerably the effects of reducing the sensing volume by the non- porous section of the coating. The use of such coatings also provides many other advantages, which will be emphasized in subsequent Sections. 4. The sol-gel process The sol-gel process is a method of preparing glasses and ceramics at low temperature by hydrolysis and polymerization of organic precursors [ 111. The process typically involves a metal alkoxide, water, a solvent and frequently a catalyst, which are mixed thoroughly to achieve homogeneity on a molecular scale. Chemical reactions (hydrolysis and condensation polymerization) lead to the formation of a viscous gel, which is an amorphous porous material containing liquid solvents in the pores. Low-temperature (typically < 100 “C) cur- ing expels most of the liquids and leaves the porous oxide. Further curing at higher temperatures leads to densification of the material. Silica, titania and silica- titania films and monoliths are routinely fabricated by this process. Porous silica sol-gel glass has been used to dip-coat a declad silica fibre [13]. The coating solu- tion was formed by precuring a TEOS (tetraethyl- orthosilicate) solution of pH _ 1 in the presence of alcohol for 20 h at 73 “C: Si(OC,H,), + H,O(pH w 1) + C2H50H 4 (-s-&a-), (8) The coated fibre was cured further to produce a hard porous coating bound intrinsically to the fibre. The sol-gel process is particularly attractive in that it affords the user a great deal of flexibility. By judicious alteration of the process parameters and precursors, it is possible to produce a range of values of material properties such as refractive index and porosity. 5. Sol-gel derived coatings for EW sensors The advantages of using porous sol-gel coatings on both reagent-mediated and direct spectroscopic evanes- cent wave sensors are presented here. 5. I. Reagent -mediated E W sensors EW sensors employing a reagent that responds opti- cally (e.g., absorption or fluorescence change) to an analyte can exhibit significantly higher sensitivity than direct spectroscopic sensors. Many methods for immo- bilizing such reagents on fibre surfaces have been re- ported [14]. The sol-gel process provides a means of coating an unclad optical fibre with a porous cladding within which the analyte-sensitive reagent is entrapped [IS]. In this case the reagent is added at the precursor stage (see eqn. (8)) and, under appropriate process conditions, becomes trapped in the nanometre-scale cages formed by the cross-linking -SiXJ--Si- units. This ensures that the reagent cannot be leached out, but smaller analyte molecules can still permeate the inter- connected microporous structure. This technique offers many advantages over other reported approaches: (i) Sensor fabrication is simple, involving straight- forward dip-coating of the unclad fibre followed by curing at low temperatures. (ii) Curing produces a tough, inert porous film which is considerably more resistant than, for example, polymer films in aggressive environments. (iii) The flexibility of the process enables a range of critical sensor parameters to be optimized. For exam- ple, the sensor response time is determined to a large extent by the coating thickness, which can be controlled accurately by the rate of withdrawal from the coating solution [12]. Moreover, as pointed out in Section 3, those parameters that enhance the evanescent wave interactions can be optimized by appropriate choice of process parameters. (iv) The technique is particularly suited to gas sens- ing because the high specific area (e.g., 100 m’/g) of the microporous structure enhances the sensitivity consider- ably. The viability of the technique was demonstrated re- cently with a pH sensor based on sol-gel entrapped fluorescein. In this application a short length of declad silica fibre was coated with a thin (0.3 pm) film of fluorescein-doped porous silica [ 131. This sensor was interrogated by evanescent wave radiation at 488 mn as shown in Fig. 2. The intensity of the resultant fluores- cence is pH sensitive. The response of the sensor to a range of pH buffer solutions is shown in Fig. 3. The sensitivity to pH is highest in the range pH 3.5-6.5. The response time of the sensor to a step change in pH was lower than 5 s, which is a considerable improvement on other optical pH sensors reported in the literature [ 161. The sensor performance in this non-optimized proof of principle was very encouraging and work is under way to optimize a range of similar sensors by capitalizing on the flexibility of the sol-gel process. Initial results from an oxygen gas sensor based on the same approach are impressive and will be reported at a later date. Other groups have reported sensor dye entrapment via the sol-gel process, but have used different sensor configurations [17, 181. Although the results reported here are for point sensors, the technique is particularly suited to distributed sensing whereby an optical fibre could be clad during the draw process with a low-loss porous coating containing a suitable concentration of reagent. Fig. 2. Sensor characterization system for dye-doped sol-gel sen- sor. A, argon-ion laser; B, beam expander; C, dichroic filter; D, prism; E, microscope objective; F, X, I’, 2 micropositioner; G, sol-gel coated fibre immersed in fluid; H, monochromator; I, photomultiplier tube; J, A/D converter; K, computer. ss()J 2 3 4 5 6 1 a PH Fig. 3. pH sensitivity of fluoresceindoped sol-gel sensor 33 5.2. Direct spectroscopic E W sensors Direct absorption spectroscopy of analytes may be performed via evanescent wave absorption on optical fibres. The typical approach to EW absorption sensors has involved the use of a completely (or partially) declad section of an optical fibre as an in-line ATR cell [19]. Here the fibre functions intrinsically as both the ATR probe and transmission line to and from the sensing location. The recent availability of good quality infrared-transmitting optical fibres has stimulated re- search interest in such sensors operating in the spectral region beyond 2 pm, which is rich in spectroscopic information. A considerable body of work has been published recently reporting the use of fluoride, chalco- genide and silver halide multimode fibres as in-line EW absorption sensors [20,21]. Single-mode fluoride fibres (Le Verre Fluore, France) are attractive for remote EW sensing in the 2-5 pm region by virtue of their high fractional power in the evanescent field and the absence of modal stability problems sometimes associated with remote multimode fibre EW sensors. We have investigated these fibres with a view to evaluating their potential for hydrocarbon gas sensing via absorption at the C-H stretching vibration band at 3.4 urn. The ability to detect the presence of flammable gases at concentrations below their LEL (lower explosive limit) is critical in many applications, such as in mines and on offshore rigs. In order the assess the viability of such an approach, point sensors based on polished semi-coupler technol- ogy were fabricated as shown in Fig. 4. This provided access to the evanescent field (interaction length 2~ 1 cm) on mechanically stable devices. Furthermore, such structures are useful in modelling the behaviour of D-shaped or eccentrically clad fibres that might be suitable for distributed sensing. The response of these devices to liquid isopropyl alcohol (IPA) was first mon- itored as a function of cladding overlayer thickness. The sensor attenuation data were then compared with the predictions of a planar waveguide model that was matched to the side-polished fibre sensor in terms of dimensions and refractive index values [ 221. The level of Fig. 4. Evanescent wave sensor formed by side-polishing an opti- cal fibre. agreement was sufficient to indicate that the model could be used to predict with reasbnable accuracy the sensitivity of such sensors to various analytes. The model predicts on attenuation of 0.006 dB cm-’ in an atmosphere of 100% methane, which corresponds to 0.00015 dBcm_’ at 50%LEL (i.e., 2.5% methane in air). Therefore a 30 m length of such a fibre would be required to cause a 10% change in transmitted intensity in the presence of 50% LEL methane. This very low sensitivity is due to the large refractive- index mismatch at the sensor-gas interface. This causes increased confinement of the guided mode and signifi- cantly reduces its evanescent field in the unclad region. Sol-gel-derived glass provides a potential solution to this problem in the form of a porous overlayer whose refractive index approaches that of the fibre core. In order for this approach to be beneficial, it is clear that the coating must provide both a reasonable refractive index match and a sufficient pore volume into which the analyte gas can diffuse. Titania (TiO,) has a bulk refractive index of 2.3 and films of this material are routinely prepared by the sol-gel process. The planar waveguide model was used to predict the response of side-polished EW sensors coated with a porous TiO, overlayer as shown in Fig. 5. The sensor attenuation data for 100% methane are plotted in Fig. 6 for a range of porosity values of the overlayer, which is presumed to be infinitely thick. In addition, the enhancement factors, calculated by comparison with the uncoated sensor, are presented. These data indicate that substantial enhancements can be achieved by coat- ing the sensor with an overlayer of the appropriate porosity. For example, a porosity of approximately 63% yields an attenuation of 0.5 dB cm’ in the pres- ence of 100% methane. This implies that a length of only 32 cm of such a coated fibre would be required to cause a 10% change in transmission in the presence of 50% LEL methane. This represents an enhancement of approximately 100 over the uncoated sensor. Work is Fig. 5. Side-polished evanescent wave sensor coated with porous TiOl overlayer. 34 d_ Fig. 6. Sensor response to 100% methane as a function of TiO, overlayer porosity. The enhancement achieved over uncoated sensors is displayed on the right-hand side. TiOz index = 2.3. now under way to verify experimentally these theoreti- cal predictions. The absence of well-established meth- ods for preparing porous infrared-transmitting glasses via the sol-gel process precludes the application of this idea to long lengths of fibre at the present time. How- ever, recent developments in the synthesis of fluoride glass by sol-gel methods suggest that such material will be available shortly [23]. 6. Conclusions The advantages that accrue from the use of sol-gel- derived porous glass coatings on evanescent wave chemical sensors have been demonstrated for both reagent-mediated and direct spectroscopic devices. This novel technique requires considerable development, but the potential for a generic approach to these enhanced sensors is evident. Such devices could be easily manu- factured and are particularly suited to disposable or distributed use. Acknowledgements The assistance of S. McCabe in running the planar waveguide model programme and in producing dia- grams is gratefully acknowledged. References 1 R. A. Lieberman, Recent progress in intrinsic fiber optic chemical sensing, SPIE Proc., Vol. 1368, 1990, pp. 15-24. 2 I. M. Walczak, W. F. Love, T. A. Cook and R. E. Slovacek, The application of evanescent wave sensing to a high sensitiv- II 12 13 10 D. Marcuse, launching light into fiber cores from sources located in the cladding, IEEE .I. Lightwave Technol., LT-6 (1988) 1273-1279. S. P. Mukherjee, Sol-gel processes in glass science and tech- nology, J. Non-Cry& Solids, 42 (1980) 477-488. C. J. Brinker and G. W. Scherer, Sol-gel Science, Academic Press, New York, 1990. B. D. MacCraith, V. Ruddy, C. Potter, B. O’Kelly and J. F. McGilp, Optical waveguide sensor using evanescent wave exci- tation of fluorescent dye in sol-gel glass, Electron. Lett., 27 (1991) 1247-1248. 0. S. Wolfbeis, Fibre optic fluorosensors in analytical and clinical chemistry, in S. J. Schulman (ed.), Molecular Lumines- cence Spectroscopy-Part II, Wiley, New York, 1988, pp. 129-281. 14 15 16 17 18 B. D. MacCraith, J. F. McGilp, B. O’Kelly and V. Ruddy, A waveguide sensor, PCT Patent Applicafion No. PCT/GB92/ 00428 ( 1992). M. S. Fuh, L. W. Burgess, T. Hinchfeld, G. D. Christian and F. Wang, Single fibre optic fluorescence pH probe, Analyst, 112(1987) 1159-1163. G. E. Badini, K. T. V. Grattan, A. W. Palmer and A. C. C. Tseung, Development of pH sensitive substrates for optical sensor applications, Proc. OFS ‘89, Paris, Springer Proceed- ings in I%yks, 1989, pp. 436-442. J. Y. Dine. M. R. Shahriari and G. H. Sine1 Jr.. Fibre ootic pH senso;8 prepared by sol-gel immobiiisatioi techniiue, EMron. Left., 27 (1991) 1560-1562. V. Ruddy, B. D. MacCraith and J. A. Murphy, Evanescent wave absorption spectroscopy using multimode fibers, J. Appl. Phys., 67 (1990) 6070-6075. V. Ruddy, S. McCabe and B. D. MacCraith, Detection of propane by IR-ATR in a Teflon-clad fluoride glass optical fiber, Appt. Spectrosc., 44 (1990) 1461- 1463. M. Rodrigues and G. H. Sigel Jr., Chalcogenide glass fibers for remote spectroscopic chemical sensing, SPIE Proc., Vol. 1591, 1991, pp. 225-235. B. D. MacCraith, V. Ruddy and S. McCabe, The suitability of single mode fluoride fibres for evanescent wave sensing, SPIE Proc., Vol. 1587, 1991, pp. 310-317. M. Poulain, University of Rennes, personal communication. 23 ity fluoroimmunoassay, Biosmors Bzoelectron., 7 (1992) 39- 48. 3 G. Stewart, D. F. Clark, B. Culshaw and I. Andonovic, Referencing systems for evanescent wave sensors, SPIE Proc., Vol. 1314, 1990, pp. 262-269. 4 D. Gloge, Weakly guiding fibers, Appl. Opt., 10 (1971) 2252- 2258. 5 R. 0. Driver, J. N. Downing and G. M. Leskowitz, Evanes- cent wave spectroscopy down infrared transmitting optical fibers, SPIE Pm., Vol. 1591, 1991, pp. 168-179. 6 M. Katz, A. Bornstein, I. Schnitzer and A. Katzir, Evanescent wave spectroscopy using chalcogenide glass fiber: theoretical analysis and experiments, SPIE Proc., Vol. 1591, 1991, pp. 236-245. 7 V. Ruddy, An effective attenuation coellicient for evanescent wave spectroscopy using multimode fiber, Fiber Integr. Opt., 9 (1990) 142-150. 8 W. Love and L. Button, Optical characteristics of fiber optic evanescent wave sensors, SPIE Proc., Vol. 990, 1988, pp. 175-180. 9 C. 0. Egalon and R. S. Rogowski, Theoretical model for a thin cylindrical fihn optical fiber fluorosensor, Opt. Eng., 31 (1992) 237-244.