Lifetime-Based pH Sensor System Based on a ... - ACS Publications

The metal complex binds irreversibly to the hydrophobic domains leaving the ... by Lifetimes of Luminescent Complexes Measured in the Frequency Domain...
0 downloads 0 Views 153KB Size
Download PDF
Anal. Chem. 2000, 72, 3468-3475

Lifetime-Based pH Sensor System Based on a Polymer-Supported Ruthenium(II) Complex Yvette Clarke, Wenying Xu, J. N. Demas,* and B. A. DeGraff*

Chemistry Department, University of Virginia, Charlottesville, Virginia 22904, and Chemistry Department, James Madison, Harrisonburg, Virginia 22807

A new luminescence lifetime-based pH sensor system is described. The system is based on [Ru(Ph2phen)2DCbpy]2+ (DCbpy ) 4,4′-dicarboxy-2,2′-bipyridine) immobilized in a mixed domain network copolymer utilizing hydrophobic regions in a hydrophilic, water-swellable, poly(ethylene oxide) matrix. The metal complex binds irreversibly to the hydrophobic domains leaving the pH-sensing COOHs projecting into an aqueous-rich poly(ethylene oxide) region. The complex shows a strong pH dependence of its lifetime (3-4-fold) and provides a usable pH range of about 3-5. The long (∼1 µs) excited-state lifetime and visible absorption of the sensor simplifies measurements. A model for the combined pH and oxygen-quenching sensitivity of the complex is provided; this allows use of the pH system over a wide range of oxygen concentrations. The combined polymer sensor is easy to prepare and requires no covalent chemistry. Further, the polymers enhance the luminescence of the complex and minimize interference from oxygen quenching. In recent years, the design and application of fiber-optic chemical sensors based on luminescence have been of considerable interest.1 Due to their sensitivity, specificity,2 small size, and flexibility,3 these sensors provide possibilities for blood gas measurements, waste- and groundwater monitoring, and industrial analysis. These types of sensors are especially important in biology, medicine, and industry for analyses of oxygen, pH, and pCO2. In particular, highly luminescent complexes of Ru(II), Os(II), and Re(I) have proved to be good sensor materials.4 The luminescence intensities and lifetime properties of these complexes are dependent on the analyte concentration. Probably the most well developed application of these complexes is as oxygen sensors because their intensity or lifetime is substantially quenched by molecular oxygen.5-7 (1) Lakowicz, J. R., Ed. Topics in Luminescence Spectroscopy. Probe Design and Chemical Sensing; Plenum Press: New York, 1994; Vol. 4. (2) Kneas, K. A.; Xu, W.-Y.; Demas, J. N.; DeGraff, B. A. Appl. Spectrosc. 1997, 51 (9), 1346-1351. (3) Jordan, D. M.; Walt, D. R.; Milanovich, F. P. Anal. Chem. 1987, 59, 437439. (4) Demas, J. N.; DeGraff, B. A.; Coleman, P. B. Anal. Chem. 1999, 71, 793A800A. (5) Price, J. M.; Xu, W.-Y.; Demas, J. N.; DeGraff, B. A. Anal. Chem. 1998, 70, 265-270. (6) Xu, W.-Y.; Schmidt, R.; Whaley, M.; Demas, J. N.; DeGraff, B. A.; Karikari, E. K.; Farmer, B. L. Anal. Chem. 1995, 67, 3172-3180.

3468 Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

pH sensing is another area of considerable interest. Typically, pH sensors have been based on either luminescence intensities or absorption.8 Luminescence intensities have the disadvantages of being influenced by signal instabilities caused by photobleaching, source fluctuations, or transmission changes in the optics.9,10 Absorption measurements are less practical because they do not provide the sensitivity required for the short absorption path lengths used in small fiber-optic probes.11 Therefore, it has become important to develop new pH sensors that avoid these problems. This leads naturally to sensors based on luminescence lifetimes.12 The simplest method of obtaining direct pH sensing is to attach a pH-sensitive group to a metal complex so that its emissive properties are altered as the pH changes. One class of pH-sensitive metal complexes is Ru(R-diimine)2L, where R-diimine has typically been a 2,2′-bipyridine, 1,10-phenanthroline, or substituted analogue and L is an R-diimine with pyridine,13,14 carboxylic acid,15,16 amine,17,18 or phenol substituents.5 The substituents can be electronically separated from the ligand by a methylene group13,19 or electronically coupled directly to the ligand.20-22 However, practical devices require a support for the sensor,23 and most of the prior work has focused on solution properties. We have been especially interested in simple techniques for immobilizing the metal complexes in practical sensors. (7) Kaneko, M.; Iwahata, S.; Asakura, T. Photochem. Photobiol. 1992, 55, 505509. (8) Baldini, F. SPIE Conference on Chemical, Biochemical, and Environmental Fiber Sensors X. Proc. SPIE-Int. Soc. Opt. Eng. 1998, 3540, 2-9 and references therein. (9) Szmancinski, H.; Lakowicz, J. R. Anal. Chem. 1993, 65, 1668-167. (10) Demas, J. N.; DeGraff, B. A. Coord. Chem. Rev., in press. (11) Demas, J. N.; DeGraff, B. A. J. Chem. Educ. 1997, 74, 690-695. (12) Kosch, U.; Klimant, I.; Werner, T.; Wolfbeis, O. S. Anal. Chem. 1998, 38923985. (13) Grigg, Robert; Norbert, W. D. J. A. J. Chem. Soc., Chem. Commun. 1992, 18, 1300-1302. (14) Licini, L.; Gareth William, J. A. Chem. Comm. 1999, 1943-1944. (15) Kohle, Oliver; Ruile, S.; Gra¨tzel, M. Inorg. Chem. 1996, 35, 4779-4787. (16) Elliot, C. Michael; Hershenhart, Elise, J. J. Am. Chem. Soc. 1982, 104, 75197526. (17) DeLaive, P. J.; Foreman, T. K.; Giannotti, C.; Whitten, D. G. J. Am. Chem. Soc. 1980, 102, 5627-5631. (18) Josceanu, A. M.; Moore, P.; Sheldon, P. Rev. Roum. Chim. 1998, 945955. (19) Grigg, R.; Holmes, J. M.; Jones, S. K.; Norbert, W. D. J. A. J. Am. Chem. Soc., Chem. Commun. 1994, 2, 185-187. (20) Kalyanasundaram, K.; Nazeeruddin, M. K.; Gra¨tzel, M.; Viscardi, G.; Savarino, P.; Barni, E. Inorg. Chim. Acta 1992, 198-200, 831-839. (21) Zheng, G. Y.; Wang, Y.; Rillema, D. P. Inorg. Chem. 1996, 35, 7118-7123. (22) Thompson, A. M. W. C.; Smailes, M. C. C.; Jeffery, J. C.; Ward, M. D. J. Chem. Soc., Dalton Trans. 1997, 5, 737-743. (23) Lobnik, A.; Oehme, I.; Murlovic, I.; Wolfbeis, O. S. Anal. Chem. Acta 1998, 367, 159-165. 10.1021/ac000111g CCC: $19.00

© 2000 American Chemical Society Published on Web 07/06/2000

Figure 1. Synthesis of D4TMI-PEG-Jeffamine network polymer.

We have shown that a Ru complex can be successfully used as an intensity-based pH sensor.11 This system consisted of [Ru(phen)2[(phen)(OH)2]]2+ and [Ru(Ph2phen)2[(phen)(OH)2]]2+ (Ph2phen ) 4,7-diphenyl-1,10-phenantholine, (phen)(OH)2 ) 4,7dihydroxy-1,10-phenanthroline, phen ) 1,10-phenanthroline) immobilized in a D4TMI-PEG-Jeffamine polymer network (Figure 1).5 Binding is based on a simple hydrophobic immobilization of the metal complex and required no covalent chemistry. However, due to low emission yields and short lifetimes in the unprotonated forms, these complexes were only suitable for intensity-based measurements with all the incumbent disadvantages. The complexes responded best in fairly acidic media. Our goal was to develop a lifetime-based system that was largely immune to the problems of intensity- and absorption-based sensors. We report on a new pH sensor system based on [Ru(phen)2DCbpy]2+ and [Ru(Ph2phen)2DCbpy]2+ (DCbpy ) 4,4′dicarboxy-2,2′-bipyridine) immobilized in a D4TMI-PEG-Jeffamine polymer network. Since these complexes emit well in both the protonated and unprotonated forms, they have the potential of being used as lifetime-based pH sensors.

A problem with metal complex sensors is that they are susceptible to oxygen quenching. Therefore, an additional goal of our research was to characterize the responses of these systems to mixed conditions of varying pH and oxygen concentrations. This is an important factor because it will allow measurements under conditions of varying oxygen concentrations, which are likely to occur in many applications. The previous work dealt only with a pH sensor under atmospheric oxygen conditions.5 We present here a new lifetime-based pH measurement system that is easily fabricated. A simple model for predicting the combined effects of pH and oxygen is developed. EXPERIMENTAL SECTION Preparation of [Ru(phen)2]Cl2 and [Ru(Ph2phen)2]Cl2. RuCl3‚3H2O from Avocado Research Chemicals Ltd. and 4,4′dicarboxy2,2′-bipyridine (DCbpy) were from Aldrich Chemical Co. 1,10-Phenanthroline (phen) and 4,7-diphenyl-1,10-phenanthroline (Ph2phen) were purchased from G. Frederick Smith. All compounds were used as received. Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

3469

[Ru(phen)2Cl2] and [Ru(Ph2phen)2Cl2] were synthesized by standard methods.24,25 Minor modifications were made to the procedure to ensure pure products. Column separation was used to purify the compounds. Basic alumina activity 1 was used as the column material with acetonitrile/ethyl ether mixtures as the eluent. Elution was started at 20% acetonitrile (v/v) and increased to 60% acetonitrile. Results were monitored by TLC (silica gel plates eluted with hexane/ethanol (90% v/v)). The dark violet solution was collected and dried to black microcrystals. Preparation of [Ru(phen)2DCbpy]Cl2 and [Ru(Ph2phen)2DCbpy]Cl2. Synthesis was a modification of published methods.7,15,20,26 We describe the synthesis of [Ru(Ph2phen)2DCbpy]Cl2. DCbpy was reacted with cis-(Ru(Ph2phen)2Cl2 (1.1: 1 mole ratio) in a methanol/water mixture (4:1) under N2. The reaction mixture was refluxed for approximately 16-24 h. The mixture was then reduced to one-third its original volume. NaCl-saturated water was dripped into the mixture to precipitate the complex. The solid was extracted with methanol to yield the soluble compound. The only difference for [Ru(phen)2DCbpy]Cl2, which is more soluble, was that the complex was precipitated with ether/ acetone solution (3:1; v/v). The solid was then dissolved in methanol and filtered, and benzene was added. The complex precipitated on standing overnight in the refrigerator. [Ru(phen)2DCbpy]Cl2 and [Ru(Ph2phen)2DCbpy]Cl2 were characterized by thin-layer liquid chromatography (TLC) and elemental analysis. Anal. Calcd for [Ru(Ph2phen)2DCbpy]Cl2: C, 63.40; N, 7.40; H, 4.10. Found: C, 63.66; N, 7.67; H, 4.28. We were unable to isolate pure [Ru(phen)2DCbpy]Cl2 because of it greater solubility, but it was pure as shown by TLC. Silica gel plates were used with ethanol/hexanes/methanol (4:3:3, v/v/v) as the eluant. Polymer Support. The insoluble cross-linked D4TMI-PEGJeffamine was synthesized as described earlier.27 The structure is given in Figure 1. The polymer consists of hydrophobic pockets comprised of a cyclic methyl siloxane ring with pendant isopropyl phenyl groups. The hydrophobic pockets are cross-linked with a very hydrophilic poly(ethylene oxide). Preparation of Sensor Solutions. Solutions of [Ru(phen)2DCbpy]Cl2 were prepared by dissolving 0.50 mg of the complex in 25 mL of distilled water to obtain a golden yellow solution. [Ru(Ph2phen)2DCbpy]Cl2 solutions were prepared by dissolving 0.64 mg of complex in 25 mL of a 15% ethanol/water mixture; the ethanol was necessary to solubilize the very hydrophobic [Ru(Ph2phen)2DCbpy]Cl2. A solution of 10 mM NaCl was added to all solutions to control the ionic strength. The pHs were adjusted by adding HCl or NaOH to unbuffered solutions. The pHs were measured with a Corning model 340 pH meter standardized with buffers. Preparation of Sensor Films. Because of the tendency of phen complexes to leach from the film, no film work was done with this complex. Sensor films of [Ru(Ph2phen)2DCbpy]Cl2 were prepared by swelling a 200-250-µm-thick film of the polymer with 10-15 mL of aqueous ruthenium complex (2 × 10-5 M) while (24) Sullivan, B. P.; Salmon, D. J.; Meyer, T. J. Inorg. Chem. 1978, 17, 12, 33343341. (25) Sprintschnik, G.; Sprintschnik, Hertha, W.; Kirsch, Pierre P. J. Am. Chem. Soc. 1977, 99, 4947-4953. (26) Chnig, E. T.; Szmacinski, H.; Malak, H.; Lakowicz, J. R. Biophys. J. 1995, 68, 342-350. (27) Xu, W. Y.; McDonough, R. C., III; Langsdorf, B.; Demas, J. N.; DeGraff, B. A. Anal. Chem. 1994, 66, 337-342.

3470

Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

shaking for 2 h. The final concentration in the films was estimated from the absorption spectrophotometer to be (3-5) × 10-4 M using the solution extinction coefficient. As with our earlier work with [Ru(Ph2phen)2(HO)2phen]Cl2, [Ru(Ph2phen)2DCbpy]Cl2 binds very tightly to the polymer. Initially the solution is brightly colored, but as the complex goes into the film, the solution becomes colorless and the film becomes highly colored. Visually the complex is homogeneously distributed in the polymer film. The films were then rinsed in water and stored in the dark in aqueous solution at near-neutral pH. Measurements were made in solutions at the appropriate pHs on films about 1 × 1 cm2 mounted reproducibly on the wall of a cuvette or in a test tube (lifetime measurements). The complex-loaded film was stable in solution for over two years even in 15% ethanol and showed no signs of leaching into solution. Absorption Measurements. Ultraviolet-visible absorption data were collected on a Hewlett-Packard 8452A absorption spectrometer. Luminescence Intensity Measurements. Luminescence spectra were measured using a Spex Fluorolog 2+2 spectrofluorometer. Excitation wavelengths were at the isosbestic points of the absorption spectra (∼475 nm). Solution data were collected with right angle detection. The pH dependence of the solutions was determined over a pH range from 1 to 12. Emission intensity data were evaluated at the maximum of the unprotonated and protonated forms. All data were taken at room temperature (22 ( 2 °C) in air-saturated and nitrogen-saturated solutions. A gas bubbler was used to saturate the gas stream with solvent. Luminescence Lifetime Measurements. A pulsed nitrogen laser (337 nm) decay system was used to measure lifetime (τ) data.28 Transients were recorded using a Tektronix TDS-540 digital oscilloscope. Emission decay curves were monitored at 630 nm. Each decay curve was the result of an average of 400 decays. Lifetime data for solution samples were acquired at room temperature and saturated with air and nitrogen. For the polymer samples, data were acquired in solutions purged with nitrogen, air (21% oxygen), 60% oxygen, and 100% oxygen. Data Analysis. All solution decay curves could be fit to a single- or double-exponential equation. Solution data are reported as preexponential weighted lifetimes. Polymer data could be fit to a three-exponential equation; however, the third component was very small and short-lived. Therefore, it contributed very little to the decay and was eliminated from the calculation of the preexponential weighted lifetimes. The preexponential weighted lifetimes, τpw, were determined from

τpw )

∑R τ /∑R j j

j

(1)

where the R’s and τ’s were measured by fitting the decay data to sums of exponentials using a Marquardt nonlinear least-squares algorithm. As we have shown, it is not necessary for there to be any physical significance to the fitting parameters.29,30 (28) Kneas, K. A.; Xu, W.-Y.; Demas, J. N.; DeGraff, B. A. Appl. Spectrosc. 1997, 51, 1346-1351. (29) Carraway, E. R.; Demas, J. N.; DeGraff, B. A. Anal. Chem. 1991, 63, 332336.

Figure 2. Luminescence spectra versus pH for [Ru(Ph2phen)2DCbpy]Cl2 in 15% ethanol and water under air. The pH of each spectrum is indicated on the Figure.

Figure 3. Luminescence intensity versus pH for [Ru(Ph2phen)2DCbpy]Cl2 in 15% ethanol/water under nitrogen monitored at the emission maximum of the basic form (triangles) and the acid form (diamonds). The solid lines are the global fit to both data sets using the sequential model (Scheme C).

The τpw versus pH curves were fit by nonlinear least squares to equations using models to be described later. Nonlinear least squares was done using PSIPlot (Poly Software International, Salt Lake City, UT). RESULTS Emission Data. Figure 2 shows the uncorrected emission spectra for the [Ru(Ph2phen)2DCbpy]2+ in solution as a function of pH. The behavior is similar for the polymer-supported complex. The emission of the protonated form is much more red shifted than the unprotonated form. However, both forms emit significantly, which is a requirement for a lifetime-based pH sensor. The data are similar for [Ru(phen)2DCbpy]2+ and are not reproduced here; the shift of the emission maximum is from 617 nm for the basic form to 668 nm for the acidic form. Solution Behavior. We first present the pH behavior for homogeneous solutions. Figure 3 shows emission intensity titration curves for [Ru(Ph2phen)2DCbpy]2+ monitored at the emission intensity of the fully protonated and totally unprotonated forms. The peaking of the emission is characteristic of 4,4′-DCbpy (30) Demas, J. N.; DeGraff, B. A. SPIE, Optically Based Methods for Process Analysis. Proc. SPIE-Int. Soc. Opt. Eng. 1992, 1681, 2-11.

Figure 4. Lifetime pH titration curve for [Ru(Ph2phen)2DCbpy]Cl2 in nitrogen-purged 15% ethanol/water. Experimental points: (- - -) fit to monoprotic acid model (Scheme 1A); (s) fit to diprotic and sequential models (Scheme 1B and C).

complex, but not of the 3,3′-DCbpy complexes.31 The [Ru(phen)2DCbpy]2+ data are very similar and are not reproduced. However, the similarities suggest that the excited-state properties are only minimally perturbed by the different phenanthroline ligands. Figure 4 shows the lifetime titration curves for [Ru(Ph2phen)2DCbpy]Cl2 in deoxygenated solutions. For oxygen quenching of [Ru(Ph2phen)2DCbpy]2+, the SternVolmer quenching lifetime plots for both complexes were linear at acid, neutral, and basic pHs. While the Ksv’s were highly dependent on pH, being much smaller under acid conditions, the oxygen bimolecular quenching rate constants, k2’s, were essentially constant. The bimolecular rate constants varied from 2.3 (pH 0.85) to 2.5 M-1‚ns-1 (pH 11.5) while the lifetime varied from 336 to 850 ns over the same range. The SV intensity and lifetime oxygen-quenching plots at pH 6.7 were virtually identical (k2 ) 2.2 and 2.5 M-1‚ns-1, respectively). This shows that oxygen quenching is purely dynamic as we would expect. Polymer Behavior. Only [Ru(Ph2phen)2DCbpy]Cl2 was measured in the polymer since our earlier results with a related complex had shown that the phen complex leached slowly from the polymer.5 In the current work, [Ru(phen)2DCbpy]2+ was even more hydrophilic and leached more rapidly than the phenol of the earlier work; such rapid leaching precluded the phen complex as being a useful sensor. The lifetime Stern-Volmer oxygenquenching plots were all linear at different pH’s. Since the lifetime varied substantially with pH, the Ksv’s also varied significantly. However, the k2’s were essentially independent of pH as shown in Figure 5. Complete lifetime versus pH and oxygen partial pressure data are shown in Figure 6. DISCUSSION The D4TMI-PEG-Jeffamine polymer was selected because it provides hydrophobic binding pockets scattered throughout a hydrophilic matrix. The poly(ethylene oxide) hydrophilic regions provide water swellability and easy transport of protons to the sensor molecule in the hydrophobic pockets. The isolated hydrophobic binding pockets (siloxane ring plus pendant isopropyl phenyls) prevent interactions between sensor molecules. Our (31) Xie, P.-H.; Hou, Y.-J.; Zhang, B.-W.; Cao, Y. J. Photochem. Photobiol. A: Chem. 1999, 122, 169-174.

Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

3471

To account for the oxygen quenching, we assumed simple Stern-Volmer quenching kinetics at all pH’s. This is supported by the linearity of the Stern-Volmer plots in solution and in the polymer support. The Stern-Volmer equations used to relate lifetime and luminescence intensities to the concentration of quencher for diffusional quenching are

Figure 5. Bimolecular oxygen-quenching rate constant versus pH for Ru(Ph2phen)2DCbpy]Cl2 in D4TMI-PEG-Jeffamine. The straight line is the average with a value of 0.54 ns-1.

Figure 6. Global fit of the luminescence lifetime for [Ru(Ph2phen)2DCbpy]Cl2 on in D4TMI-PEG-Jeffamine network polymer versus pH and oxygen concentration. The percent saturation relative to 1 atm of pure oxygen is given next to each curve.

earlier studies indicated that sensor molecules with a hydrophobic portion and a polar region would bind in the polymer’s hydrophobic regions via a hydrophobic interaction. Further, the polar, pH-sensing portion of the sensor molecule would be left projecting into the aqueous region where it could respond to pH changes.5 The current work supports a similar binding mode for [Ru(Ph2phen)2DCbpy]2+. The failure of [Ru(phen)2DCbpy]2+ to bind irreversibly is also consistent with this model. Our results also show that we have realized a prototype luminescence lifetime-based pH sensor utilizing a long-lived Ru(II) complex. The lifetime changes for the polymer-supported complex are large (3-fold in air-saturated solutions and almost 4-fold in deoxygenated solutions) for a pH change of ∼2. The lifetime is long enough to be easily measured, and the complex can be readily excited with blue LEDs. The system exhibits mixed sensitivity in that its response is affected by both pH and oxygen; we will now discuss the modeling used to quantitatively account for the response. Fitting Model. Our basic model assumes that we can treat the molecules as emitting from all of the different forms present at any given pH; the composition is determined by the pH equilibria. Further, each species is quenched in the normal bimolecular fashion by oxygen. The observed decays result from the simultaneous contribution of both of these effects. 3472

Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

τ0/τ ) 1 + KSV[Q]

(2a)

I0/I ) 1 + KSV[Q]

(2b)

KSV ) k2τ0

(2c)

where τ’s and I’s are the lifetimes and luminescence intensities, respectively, as a function of quencher concentration. The subscript “0” represents the measurement in the absence of quencher. The Stern-Volmer quenching constant is KSV and the bimolecular quenching constant is k2. Accounting for the pH sensitivity was more complex. There are two sites on the complex that can be protonated. One can envision three models as shown in the schemes. In Scheme 1 A, only one of the protonatable sites is active. In Scheme 1B, a single dual-proton reaction is assumed. In Scheme 1C, two sequential one-proton processes are assumed. All three schemes were evaluated as described below. In all three schemes, we assumed that there was a rapid equilibrium between the different forms. Using the steady-state approximation with a rapid exchange in the excited-state manifold, the relevant lifetime equation for Scheme 1A is

τ)

1 + [H+]/K* kA + kHA[H+]/K*

kA )

1 + KSVHA[O2] τA

kHA )

1 + KSVHA[O2] τHA

K* )

[H+][A-*] [HA][HA*]

(3)

where [H+] is equal to 10-pH and the parameters kA and kHA are the excited-state decay constants for A-* and HA*, respectively. K* is the acid dissociation constant. If there were no exchange between the two forms, the reciprocal of these rates would be the observed lifetimes for each of the species. kA and kHA are dependent on oxygen concentration and obey the Stern-Volmer relationship shown in eq 2. The simultaneous two-proton reaction of Scheme 1B yields

τ)

1 + [H+]2/K* kH2A + kA[H+]2/K*

(4)

Again kA and kH2A are related to oxygen concentrations by a SternVolmer relation.

Scheme 1. Protonation Models

The sequential model of Scheme 1C yields the following equation: +

τ)

I)

+ 2

1 + [H ]/K1* + [H ] /K2*K1* kA + kHA[H+]/K1* + kH2A[H+]2/K2*K1*

(5)

where there are analogous Stern-Volmer equations relating kA, kHA, and kH2A to the oxygen concentration. Using a similar kinetic model, one can also write equations for the luminescence intensity. The equation for scheme C is shown below.

I)

IA + IHA([H+]/K1*) + IH2A([H+]2/K2*K1*) 1 + [H+]/K1* + [H+]2/K2*K1*

Scheme 1B is

(6)

where the I’s are intensities that would be seen for each of the pure components at the monitored wavelength. This equation reduces to Scheme 1A by eliminating the second K2* term and replacing K1* with the two-proton K*. The intensity equation for

IA + IH2A([H+]2/K1*) 1 + [H+]2/K1*

(7)

where the terms are defined above. Solution Data. In solution, the luminescence intensity and lifetimes of both complexes show pronounced pH sensitivity over the 2-5 pH range. The lifetime and intensity data have similar pH profiles, although the magnitude of the changes differ. This difference is just a manifestation of the differences in the quantity measured. For the intensity data, we are not comparing quantum yields, but emission intensity at a single emission wavelength where the different species have different emission spectra. The intensity titration curves show steep rises and small but distinct peaks. The single-proton step of Scheme 1A cannot duplicate the rapid rise. Neither Scheme 1A nor B can give a peak. However, of these two schemes, the Scheme 1B step does do a much better job of modeling the sharp rise of the data. The sequential two single-proton model of Scheme 1C is chemically Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

3473

the most plausible model and also gives the best fit, which is shown in Figure 3. There are two roughly equivalent protonatable sites in the complex. Assuming the sequential model, the two pK*’s are similar and close enough to each other to preclude seeing a double-titration curve. The fitting results were similar for the phen complex with the sequential two-proton model being required to fit the various features. These results are similar to those for [Ru(Ph2phen)2DCbpy]2+, which shows that the acid/base chemistry is very similar for the two complexes and is controlled largely by the aromatic COOHs with little interaction with the remaining R-diimine ligands. The solution lifetime data for [Ru(Ph2phen)2DCbpy]Cl2 are shown in Figure 4 along with fits to the three schemes. Scheme 1B or C gave the best fits, and it was not possible to clearly differentiate between these two. A sequential two-proton model, however, is physically more reasonable than the two-proton model of Scheme 1B and would be consistent with the intensity data. As noted in the Results, for [Ru(Ph2phen)2DCbpy]Cl2 in solution, the Ksv’s varied substantially between the acid and base forms; but the bimolecular oxygen-quenching rate constants were virtually the same. Thus, despite large differences in lifetimes and structures, the unprotonated and protonated forms of the complex are oxygen quenched at essentially the same rate. In contrast to the solution data, the best fits for [Ru(Ph2phen)2DCbpy]Cl2 on the polymer were obtained with Scheme 1A or C. Scheme 1B produced distinctly inferior fits because it forced much too rapid a rise in the lifetime versus pH curve. Both parts A and C of Scheme 1 gave excellent fits to each of the individual oxygen concentration data sets versus pH. The K’s for the individual fits at different oxygen concentrations agreed reasonably well with each other. However, a more systematic way of obtaining the best overall fit was desired. To this end, we carried out global fits on all of the different lifetime oxygen concentration data sets. The varied parameters were the K’s, the unquenched τ’s of the different species, and the bimolecular quenching constants. Since k2 was shown to be independent of species, we assumed a single k2 for both acid and basic forms. Parts A and C of Scheme 1 clearly gave the best results with a standard deviation in the total fit of 9.3 and 9.1 versus 17 ns for Scheme 1B. Therefore, there is no mathematical reason for choosing (C) over (A) although (C) is chemically more reasonable. The parameters for Scheme 1A are pK* ) 3.19 ( 0.04, τA ) 3282 ( 25 ns, τHA ) 979 ( 31 ns, and k2 ) 0.55 ( 0.01 M-1‚ns-1. However, for analytical modeling, Scheme 1A is clearly much simpler than (C) since it requires two fewer parameters. Even though Scheme 1C is probably more rigorously correct, Scheme 1A is much easier to implement for a reliable calibration curve. Indeed, the necessary fitting parameters for Scheme 1A can be obtained with as little as one pH titration under nitrogen-saturated conditions to provide K and the unquenched τs and one SternVolmer measurement, which supplies k2. The optimum place to measure oxygen quenching would be in the basic region where quenching is greatest. Then, with just these four parameters, calibration curves can be computed for any oxygen concentration and pH. The fits in Figure 6 are all done with the global Scheme 1A parameters. 3474

Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

Polymer Implications. There are dramatic changes in the photophysical properties on binding the complex to the polymer. The unquenched excited-state lifetimes increase dramatically (∼3-4) on binding to the polymer. The oxygen-quenching constants decrease by a factor of 5. We attribute these changes to the beneficial shielding of much of the complex from water, which is a known deactivator of MLCT excited states.32 The polymer also provides a more rigid environment, which may further decrease radiationless processes. The decrease in k2 on the polymer is almost certainly a consequence of steric shielding of the excited state from diffusional quenching by oxygen. This effect is analogous to that observed on inclusion of metal complexes into cyclodextrins.33 Because of interpretation complexities, it is not easy to compare the solution and polymer pK*’s. However, the inflection points in the solution and polymer data are nearly indistinguishable at ∼3.75. This implies little change in pK* on binding to the polymer, which is a result consistent with our interpretation that the hydrophilic COOHs project into the aqueous phase of the polymer. Analytical Requirements. It is useful to determine the accuracy necessary in the oxygen determination in order to achieve a given accuracy in the pH determination. From the lifetime versus pH and oxygen partial pressure and the fitting parameters, we can estimate the maximum sensitivity of 1300, 760, and 240 ns/pH unit under nitrogen, air, and pure oxygen, respectively. From the oxygen quenching at each of these conditions, we calculate a sensitivity to the oxygen measurement of 0.0035 pH unit/Torr uncertainty in the oxygen tension, 0.0035 pH/Torr, and 0.0024 pH/Torr for nitrogen, air, and pure oxygen, respectively. Since oxygen tension can probably be measured to significantly better than 10 Torr, the errors associated with oxygen determination would then be well below 0.03 pH unit near the optimum range. CONCLUSION We have described a new lifetime-based pH sensor using a long-lived ruthenium complex. The lifetime is long enough to make measurements simple. The complex is easily excited with a blue LED. A calibration equation with a minimum number of parameters allows simple calibration of the sensor as a function of pH and oxygen concentrations. This work shows the generality of the cyclic siloxane polymer for a hydrophobic support for pH sensors. The binding to the polymer is completely hydrophobic and avoids the need for any covalent chemistry. Yet the hydrophilic portion of the polymer allows easy access of solvent to the pH-sensing portion of the sensor molecule. Other types of aqueous borne analytes are easily envisioned. The support serves the very nice function of simultaneously lengthening the excited-state lifetime of the sensor molecule, increasing its luminescence efficiency, and reducing its sensitivity to oxygen quenching. This once again points out the importance of proper selection and design of the sensor support. While the pulsed lifetime measurement method used here is adequate, it is cumbersome since complete decay curves must (32) Hauenstein, B. L., Jr.; Dressick, W. J.; Buell, S. L.; Demas, J. N.; DeGraff, B. A. J. Am. Chem. Soc. 1983, 105, 4251-4255. (33) Sacksteder, L.; Lee, M.; Demas, J. N.; DeGraff, B. A. J. Am. Chem. Soc. 1993, 115, 8230-8238.

be collected and reduced. This could be greatly simplified by using the rapid lifetime determination method where only the area under a small number of portions of the decay curve need be evaluated.34 Even though the assumed fitting model would be incorrect, it would reduce the complex decays to a single, easily obtained number that could be used for evaluating the pH. Alternatively, a phase shift method could be used for evaluating an effective lifetime. Again, no significance would be placed on any computed lifetimes; it would merely provide a parameter that varied monotonically with pH and could be used for calibration. Finally, the fact that the oxygen concentration must be known to determine the pH is not an overly serious problem. Many (34) Sharman, K. K., Periasamy, A.; Ashworth, H. H.; Demas, J. N.; Snow, N. H. Anal. Chem. 1999, 71, 947-952.

systems require simultaneous pH and pO2 measurements, so dual sensors would be common. The pO2 data could then be used to obtain the pH from the sensor. Of course, a single reliable calibration curve at one oxygen concentration (e.g., air saturated) could be obtained. One could then use our model but include only a fixed oxygen concentration. ACKNOWLEDGMENT The authors thank the National Science Foundation (Grants CHE-94-19074 and CHE 97-26999) for support. Received for review February 3, 2000. Accepted May 16, 2000. AC000111G

Analytical Chemistry, Vol. 72, No. 15, August 1, 2000

3475