Open Air Calibration with Temperature Compensation of a

Anal. Chem. 2007, 79, 3173-3179

Open Air Calibration with Temperature Compensation of a Luminescence Quenching-Based Oxygen Sensor for Portable Instrumentation Alberto J. Palma,† Javier Lo´pez-Gonza´lez,‡ Luis J. Asensio,† Marı´a Dolores Ferna´ndez-Ramos,‡ and Luis Fermı´n Capita´n-Vallvey*,‡

Department of Electronics and Computer Technology, and Solid Phase Spectrometry Research Group, Department of Analytical Chemistry, Campus Fuentenueva, Faculty of Sciences, University of Granada, E-18071 Granada, Spain

This report addresses the task of calibrating an optical sensor for oxygen determination. Detailed analyses of the functional dependences from our measurement system results have been carried out with the additional aim of temperature compensation. As a result, an empirical calibration function has been successfully derived for the luminescent quenching-based oxygen sensor included in a self-designed portable instrument. This function also compensates for the temperature influence on the quenching luminescence process in the range from 0 to 45 °C. Moreover, the calibration procedure is extremely simple because only a single standard is needed. In fact, the oxygen measurement system can be calibrated with exposure to an open air atmosphere, and therefore, neither laboratory standards nor trained personnel are required. The method has been applied to a set of 11 units of the mentioned sensor (up to 24% oxygen concentration) giving an overall deviation between our calibrated system results and the laboratory standards of 0.3% oxygen concentration (calculated with 95% confidence level). The proposed calibration function has shown itself to be applicable for different sensing film thicknesses and luminophore concentrations using the same fittings parameter. Additionally, this function has been successfully applied to other oxygen dyes. Good agreement has also been found when the performance of the instrument was compared to a commercially available portable instrument based on an electrochemical sensor. We believe that this work could be an interesting finding for spreading the use of optical sensors for atmospheric oxygen determination in commercial measurement equipment for different purposes in confined working atmospheres, such as mines, undergrounds, warehouses, vehicles, and ships. Optical oxygen sensing is widely used for a wide variety of applications including real-time clinical monitoring, environmental sampling, food packaging technology, and surface air pressure * Corresponding author. E-mail: [email protected]. † Department of Electronics and Computer Technology. ‡ Solid Phase Spectrometry Research Group, Department of Analytical Chemistry. 10.1021/ac062246d CCC: $37.00 Published on Web 03/17/2007

© 2007 American Chemical Society

distributions (pressure-sensitive paints). An optical oxygen sensor works on the basis of the dynamic quenching of the luminescence of a luminophore, usually dispersed in a permeable polymeric matrix by molecular oxygen. The degree of quenching is related to the oxygen concentration of the sample, or pressure, via a Stern-Volmer type equation. The usual calibration of quenching-based oxygen sensors relies on a Stern-Volmer type equation, which is usually nonlinear due to the microheterogeneity of the luminophores in polymeric matrixes.1 The full characterization of Stern-Volmer behavior means using a range of standard gas mixtures in defined proportions.2 The Stern-Volmer algorithm requires at least two standards: the first at 0% oxygen concentration and the second at the end of the dynamic range used.3,4 Alternatively, the calibration of oxygen optical sensors can be performed by a single standard using the exponential dilution method.5 Multivariate calibration by means of principal component analysis has been used in the case of dual-luminophore sensing for surface pressure mapping.6 The quenching process of most polymer-based oxygen sensors is influenced by temperature, due to several factors including the temperature dependence of the radiationless decay rate and the temperature dependence of the triplet yield of the luminophore and also the diffusivity of oxygen through polymer.7,8 Different strategies have been developed to correct for temperature dependence both in intensity-based methods,9,10 with its wellknown drawbacks,11 and in lifetime-based methods.12-14 In all (1) Demas, J. N.; DeGraff, B. A. Sens. Actuators, B 1993, 11, 35-41. (2) Chuang, H.; Arnold, M. A. Anal. Chim. Acta 1998, 368, 83-9. (3) Trettnak, W.; Kolle, C.; Reininger, F.; Dolezal, C.; O’Leary, P.; Binot, R. A. Adv. Space Res. 1998, 22, 1465-74. (4) Ocean Optics. Theory of operation: oxygen sensor. 2001. (5) Choi, M. M. F.; Xiao, D. Anal. Chim. Acta 2000, 403, 57-65. (6) Koese, M. E.; Omar, A.; Virgin, C. A.; Carroll, B. F.; Schanze, K. S. Langmuir 2005, 21, 9110-20. (7) Schanze, K. S.; Carroll, B. F.; Korotkevitch, S.; Morris, M. J. AIAA J. 1997, 35, 306-10. (8) Gouin, S.; Gouterman, M. J. Appl. Polym. Sci. 2000, 77, 2815-23. (9) Ocean Optics. FOXY Fiber Optic Oxygen Sensor System Manual, 2002. (10) Koese, M. E.; Carroll, B. F.; Schanze, K. S. Langmuir 2005, 21, 9121-9. (11) Borisov, S. M.; Wolfbeis, O. S. Anal. Chem. 2006, 78, 5094-101. (12) Coyle, L. M.; Gouterman, M. Sens. Actuators, B 1999, B61, 92-9. (13) Hradil, J.; Davis, C.; Mongey, K.; McDonagh, C.; MacCraith, B. D. Meas. Sci. Technol. 2002, 13, 1552-7.

Analytical Chemistry, Vol. 79, No. 8, April 15, 2007 3173

cases, the temperature determination from the intensity or lifetime of the luminophore makes it possible to select the temperaturecorrected calibration curve. These strategies include the use of a second-order polynomial algorithm as an oxygen calibration function including temperature dependence, through the inclusion of a temperature sensor in the fluorescence intensity-based instrument.9 The thermostatization of a flow cell can be sufficient for a low gas flow rate;15 however, the most common way to compensate for the temperature effect is via a second independent measurement, by means of the addition of a non-oxygen quenched, temperature-dependent luminophore to the oxygen sensor. Very different approaches have been described based on this dual-luminophore scheme. These include the following: (i) the use of separate layers for each luminophore;16,17 (ii) the compartmentalization of the oxygensensitive and the temperature-sensitive dyes in the binder, through the dispersion of both luminescent dyes, each encapsulated in different polymer microspheres or nanospheres, in an oxygenpermeable polymer binder;10,14 and (iii) the dispersion of a bichromophoric luminophore in the polymeric membrane.18 Most of the previous calibration procedures require standard gas mixtures prepared in laboratory by trained personnel and facilities. A method that can bypass these requirements with a single standard calibration is needed, and if this single standard is in open air, the advantages would be notables. In this article, previous calibration strategies have been applied to our experimental data and we point out some drawbacks. Our goal was to obtain a simple and reliable calibration algorithm suitable for implementing in a portable instrument19 for oxygen determination. Then, we report on a calibration procedure that compensates for the temperature dependence in a gaseous oxygen measurement system. This system is made up of a portable electronic instrument with an optochemical sensor previously described.20 The microcontroller-based instrument uses a light-emitting diode (LED) as optical excitation and a binary output photodetector, coated with the oxygen sensing film, for collecting the luminescent emission. The sensing film is based on the dye platinum octaethylporphyrin complex immobilized in a polystyrene membrane and stabilized with the heterocyclic amine DABCO. The feasibility of single standard calibration in open air atmospheres for this portable instrumentation has been tested, validating the results against a commercial portable instrument EXPERIMENTAL SECTION Reagents and Sensor Film Preparation. Platinum octaethylporphyrin (PtOEP) was obtained from Porphyrin Products Inc. (Logan, UT), 1,4-diazabicyclo[2.2.2]octane 98% (DABCO), tetrahydrofuran (THF), polystyrene (PS, average MW 280 000, Tg 100 (14) Borisov, S. M.; Vasylevska, A. S.; Krause, C.; Wolfbeis, O. S. Adv. Funct. Mater. 2006, 16, 1536-42. (15) Trettnak, W.; Gruber, W.; Reininger, F.; Klimant, I. Sens. Actuators, B 1995, 29, 219-25. (16) Woodmansee, M. A.; Dutton, J. C. Exp. Fluids 1998, 24, 163-74. (17) Klimant, I.; Holst, G. U.S. Patent 6,303,386, 2001. (18) Ji, H. F.; Shen, Y.; Hubner, J. P.; Carroll, B. F.; Schmehl, R. H.; Simon, J. A.; Schanze, K. S. Appl. Spectrosc. 2000, 54, 856-63. (19) Capita´n-Vallvey, L. F.; Palma, A. J.; Fernandez-Ramos, Maria Dolores, Lo´pezGonza´lez, J.; Asensio, L. J. Patent PCT/ES 2006/000384. (20) Palma, A. J.; Lo´pez-Gonza´lez, J.; Asensio, L. J.; Fernandez-Ramos, M. D.; Capita´n-Vallvey, L. F. Sens. Actuators, B 2007, 121, 629-38.

3174

Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

°C, GPC grade) was obtained from Sigma-Aldrich Quı´mica S.A. (Madrid, Spain). The oxygen sensing-coated photodetectors were prepared by casting an oxygen-sensitive membrane on a solid-state photodetector using a coating technique. The deposited film contains the dye PtOEP encapsulated in PS and stabilized with the heterocyclic amine DABCO. Mixtures for the preparation of oxygen-sensitive devices were typically made by dissolving 0.5 mg of PtOEP and 12 mg of DABCO in 1 mL of a solution of 5% (w/v) PS in freshly distilled THF in a 2-mL glass vial. The membranes were cast by placing successive volumes of 5 µL of the mixture on the active face of a photodetector with the aid of a micropipet. After each addition, the device was left to dry in darkness. We prepared three different types of coated photodetectors with a total volume deposited of 5, 10, and 15 µL. Once the coating was finished, the photodetector was dried in a dryer with saturated THF atmosphere for 24 h at room temperature and in darkness. The obtained coated photodiodes show the bulb covered with a homogeneous, transparent, and pink film with an estimated average thickness of about 12, 23, and 36 µm, respectively, and an approximate PtOEP concentration of 14.4 mM. Additionally, we prepared coated photodetectors with different concentrations of PtOEP, namely, using cocktails containing 0.3 and 0.7 mg of PtOEP and maintaining the rest of the components the same. For the indicator molecule/polymer matrix system PtOEP/ PS, excitation at the metalloporphyrin Q-band at 534 nm causes luminescence emission at 647 nm, which is dynamically quenched by oxygen, thus causing changes in both phosphorescence intensity and lifetime. The spectroscopic properties of this PtOEP membrane in the presence of DABCO did not change, although the decay time increased slightly (104.4 µs at 0% O2 and 13.7 µs at 100% O2 for membranes with 25% DABCO) compared to membranes with no DABCO (99.9 µs at 0% O2 and 10.8 µs at 100% O2). The incorporation of DABCO to the membrane meant a substantial increase in photostability, with a mean drift of 0.02% oxygen concentration per day. Portable Instrumentation and Signal Processing. A lowpower portable instrument for oxygen measurement was designed and fabricated where the optical excitation is sourced by means of a green ultrabright LED with an emission maximum at 525 nm (110104, Marl International Ltd., Ulverston, UK).21 The oxygen sensing film was deposited directly onto the photodetector active face as explained above. The photodetector (IS486, Sharp), with a built-in Schmidt trigger circuit, gives a binary output depending on an illuminance threshold, named Ith. A temperature sensor (DS1820, Maxim, Sunnyvale, CA), located next to the coated photodetector, was included to measure the temperature in order to compensate for its influence on the quenching phenomenon. In addition, logic circuits for the analog/digital conversion and power management circuits were designed. User interface was made up of components such as a keyboard for measurement configuration, an alarm for indicating risky oxygen concentrations, an LCD display for presenting the measurement, and an RS-232 port for computer communication. The instrument is managed by a low-cost microcontroller (PIC16F873, Microchip, Chandler, AZ) to perform the abovementioned hardware control and to (21) Capitan-Vallvey, L. F.; Asensio, L. J.; Lopez-Gonzalez, J.; Fernandez-Ramos, M. D.; Palma, A. J. Anal. Chim. Acta 2007, 583, 166-73.

include the signal processing algorithm and the calibration function. More details are given in ref 20 and 21. With these optoelectronics and electronics components, the prototype shows the following main features: (i) one optical signal channel without filters and no alignment support, (ii) fully digital signal processing, (iii) hand-held with low size and low weight, (iv) low-energy consumption, which allows battery operation, v) temperature measurement, vi) communication with computer for data saving and easy reconfiguration, and vii) built-in menu for operation mode selection. The signal processing is described in Figure 1. The internal variable that encloses the oxygen concentration information is N, the count number used as an analytical parameter. Briefly, the procedure can be explained as follows: 1. After LED excitation, the coated photodetector collects the phosphorescence emission decay, I(t) from the sensing film layer. The photodetector output (PO) remains high while this decay is above the photodetector illuminance threshold (Ith). When I(t) < Ith, the PO signal changes to low level. As is well-known, I(t) depends on analyte concentration; therefore, the time that the PO signal is high, tN, varies with this magnitude. 2. The next step is to accurately quantify tN. To do that, it is multiplied for a high-frequency signal, in our case, the microcontroller clock signal (CLK). 3. Finally, this pulsed train is codified, counting the periods of CLK contained in the time interval tN with a built-in counter of the microcontroller. It should be stated that this measured parameter, tN, is closely related to a lifetime measurement. An approximate calculation helps us to demonstrate this. Let us consider a simple exponential decay for the emitted phosphorescence, I(t) ) I(0)e-t/τ, where I is the luminescence intensity, τ the lifetime, and t the time, as shown in Figure 1. Considering Ith, the photodetector illuminance threshold, the pulse duration ratio can be written as

tN0 t

N

)

τ0 lnI0 - lnIth τ lnI - lnIth

(1)

where parameters with 0 subscript correspond to the absence of oxygen and the rest to the presence of oxygen. Given that intensity is affected by the logarithmic function, the lifetime is dominant in the parameter extracted by the measurement procedure implemented in our portable instrument. The signal processing is completed by improving the signalto-noise ratio by means of repeated measurements and subsequent averaging. The microcontroller sends a control signal with a frequency of 2 kHz and 50% duty cycle to the LED polarization circuit. Therefore, each 0.5-ms optical excitation is produced and the photodetector output is processed with the explained method. This was repeated 2000 times during 1 s, and the results were averaged. The last step is to calculate and include in the microcontroller a calibration function that relates the oxygen concentration to the analytical parameter, N. Moreover, this function must compensate for the thermal influence in the luminescent emission. In this work, our aim has been to find the function with which the oxygen concentration can be extracted from the temperature, T, and the analytical parameter N.

Figure 1. Schematic representation of the instrument signal processing. A luminescence decay, I(t), is plotted and superposed on the photodetector illuminance threshold, Ith. The time interval I(t) > Ith, named tN, is later quantified (by multiplying with the signal clock system, CLK) and finally codified by the microcontroller counter (µC Counter), resulting in the analytical parameter: the count number N.

Calibration Setup. For the calibration procedure, a set of 11 coated photodetectors was prepared using the process detailed above with 10 µL of cocktail of 500 mg/L in PtOEP. Each photodetector was separately characterized, including it in the portable instrument for an oxygen concentration range between 0 and 24% and from 0 to 45 °C. The count number N was calculated by an average of 10 replicas for each concentration of oxygen and temperature. The gases O2 and N2 used were of high purity (>99%) and were supplied in gas cylinders by Air Liquid S.A. (Madrid, Spain). The standard mixtures of oxygen were produced using nitrogen as the inert gas component by controlling the flow rates of oxygen and nitrogen gases entering a mixing chamber using a computercontrolled mass flow controller (Air Liquid Espan ˜a S.A.) operating at a total pressure of 760 Torr and a flow rate of 500 cm3‚min-1 with a mean precision of 0.2%. The oxygen measurements were performed after equilibration of the atmosphere of the instrument with the gas mixture. Data were saved in a computer through the serial port connection. All the controlled-temperature experiments were performed in a thermostatic chamber with a lateral hole for connection to a computer, which made it possible to maintain a constant temperature between -50 and +50 °C with an accuracy of (0.5 °C. RESULTS AND DISCUSSION Comparison with Previous Models. The first question to answer is whether the sensing configuration and the analytical parameter proposed behave like earlier similar quenching-based oxygen sensors. To do this, earlier theoretical and empirical models were applied to our sensing system. In parallel, from these Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

3175

discussions of previous models, a calibration procedure is obtained to extract the oxygen concentration from the analytical parameter N, compensating for the temperature influence. Regarding the Stern-Volmer model, the immobilization of the PtOEP complex in a PS membrane results, measuring lifetime, intensity, or count number, in a reproducible and characteristic negative deviation (at [O2] > 12%) from the linearity predicted by the Stern-Volmer equation.20,21 In addition, the analytical parameter N shows a variation in the temperature coefficient from -0.5%/°C, at low oxygen concentrations and low temperatures, to -2.5%/°C at high concentrations and high temperatures in the above-mentioned ranges. The variation of the thermal slope with the oxygen concentration indicates a non-negligible cross-sensitivity of the sensor. This trend agrees with previous studies on the thermal dependence of the lifetime of the luminescence of this complex and similar ones in the presence of oxygen. It has been attributed to a thermally activated nonradiative decay.22-24 Considering the Arrhenius-like model for calibration purposes and accepting the count number as a nearly lifetime magnitude, it can be written as12,25

∆E 1 ∆E 1 ) ) ko + k1 exp + k1 exp RT RT N(T) N(T ) 0)

(

)

(

)

(2)

where ko is the temperature-independent term, k1 the preexponential factor, ∆E the activation energy, T the absolute temperature, and R the gas constant. Evaluating eq 2 for a reference temperature Tr, dividing and taking natural logarithms yields

ln

[ (

)]

N(T) N(T ) 0) - N(Tr) N(Tr) N(T ) 0) - N(T)

)

( )

1 ∆E 1 R T Tr

(3)

Extrapolation of our results on N to low temperatures shows that N(T ) 0) is much larger than N(T) and N(Tr); thus, [N(T)0) N(Tr)]/[N(T)0) - N(T)] is approximately the unity, hence

[ ]

ln

( )

N(T) ∆E 1 1 ) R T Tr N(Tr)

(4)

From eq 4, the activation energy can be extracted, as shown in Figure 2, where very good agreement with the model has been achieved. However, regardless of the compact relation calculated, different activation energies were obtained at different oxygen concentrations. Moreover, data from a reference temperature (different from the measurement temperature) are necessary for thermal compensation with this model. For these reasons, this option was discarded as calibration function, although it provides theoretical support for the oxygen sensing device. The thermal dependence of this kind of sensing membrane has been modeled by a second-order polynomial function. This model has been used in two different contexts. On the one hand, (22) Lee, S. K.; Okura, I. Anal. Sci. 1997, 13, 535-40. (23) Lupton, J. M.; Klein, J. Chem. Phys. Lett. 2002, 363, 204-10. (24) Tsuboi, T.; Murayama, H.; Penzkofer, A. Thin Solid Films 2006, 499, 30612. (25) Liebsch, G.; Klimant, I.; Wolfbeis, O. S. Adv. Mater. 1999, 11, 1296-9.

3176

Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

Figure 2. Temperature dependence of the count number ratio at different oxygen standards. Lines represent fits to the Arrhenius eq 4.

Figure 3. Fits of our instrument analytical parameters (symbols) to the second-order polynomial function given in eq 5 (lines) plotted for different oxygen standards and temperatures.

it has been theoretically derived to justify the downwardly curved Stern-Volmer plots, assuming the coexistence of dynamic and static quenching processes in intensity-based oxygen sensors.26 On the other hand, it has been argued as an empirical approximation for calibration purposes.9 Therefore, we tested the following function:

N0 ) 1 + K1[O2] + K2[O2]2 N

(5)

In this equation K1 and K2 are two empirical coefficients. In Figure 3, data from the coated photodetector number 1 are successfully fitted to eq 5 for all tested temperatures. To include thermal compensation, the K1 and K2 coefficients and N0 have also been fitted to a second-order function and a linear function in temperature, respectively. For calibration, this second-order polynomial algorithm requires at least three standards of known oxygen concentration. The first standard must have 0% oxygen concentration, and the last standard must have a concentration in the high end of the concentration range in which measurement will be done. Despite the excellent agreement with this model, the calibration function with thermal compensation requires a great number of parameters, specifically eight, for accurate calibration. Single Standard Calibration Procedure. The earlier applied models demonstrate the following: (i) our measurement system (26) Lee, S. K.; Okura, I. Spectrochim. Acta. A 1998, 54, 91-100.

Table 1. Fit Parameters and Correlation Coefficients of the Proposed Calibration Function (eq 6) for the Set of Eleven Characterized Photodetectors photodetector

A0

A1

A2

A3

A4

r2

1 2 3 4 5 6 7 8 9 10 11

-16.39 -18.12 -11.44 -14.09 -14.64 -14.34 -16.83 -17.08 -17.12 -15.33 -16.94

0.162 0.158 0.106 0.120 0.128 0.128 0.161 0.143 0.137 0.134 0.137

7275 8204 4979 6213 6548 6241 7304 7935 8004 6858 7747

-84.24 -93.27 -54.78 -64.58 -75.11 -59.75 -84.76 -86.90 -87.83 -77.00 -77.56

0.172 0.271 0.189 0.166 0.295 -0.005 0.213 0.293 0.363 0.284 0.188

0.999 62 0.999 43 0.999 46 0.999 77 0.999 76 0.999 82 0.999 80 0.999 01 0.998 85 0.999 61 0.999 24

Table 2. Linear Relationships between the Coefficient A0 and the Remaining Coefficients of Eq 6, in the Form yi ) miA0 + di yi

slope (mi)

intercept (di)

r2

A1 A2 A3 A4

-7.690 × 10-3 -5.026 × 102 6.044 -2.075 × 10-2

1.698 × 10-2 -8.455 × 102 1.780 × 101 -1.043 × 10-1

0.713 98 0.974 98 0.873 57 0.165 50

presents characteristics similar to other comparable sensing configurations previously described, and (ii) they require a great number of parameters and standards to achieve good calibration and thermal compensation. From the above discussion, the functional dependences of the analyte concentration as a function of the temperature and the count number can be deduced. Thus, the thermal dependence of the oxygen concentration fits a quadratic function quite well. From eq 5, the functional dependence of the oxygen concentration as a function of the analytical parameter N is the inverse of the square root, at constant temperature. Moreover, cross-sensitivity terms should be included according to the above discussion. With these considerations and with the aim of simplifying the calibration procedure, the relationship between the oxygen concentration and the analytical parameter (I, intensity; τ, lifetime; or N, count number) by means of an empirical function is proposed, including individual parameter dependences in addition to cross-correlated terms:

{ } -1/2

[O2] ) C0 + C1

{

I C0 ) A0 + A1T τ-1/2 with C ) A + A T + A T2 1 2 3 4 N-1/2 (6)

where T is expressed in Celsius grades and [O2] is obtained in volume percentage. This function has been applied to experimental data (N) from the 11 coated photodetectors, and the results of the fittings are summarized in Table 1, showing excellent correlation coefficients, r2. Figure 4 shows good agreement between the oxygen concentration calculated from eq 6 with the parameters N and T of one of the analyzed photodetectors (symbols) and the gaseous standards applied to the instrument (plotted as lines for clarity). With this function, the number of parameters has been considerably reduced from the eight in the previous model to five.

Figure 4. Comparison of the oxygen concentration calculated with the empirical calibration function given in eq 6 from the analytical parameters of one photodetector (symbols) with the laboratory oxygen standards applied (shown as lines for clarity). Temperatures are as in Figure 3.

However, Table 1 provides different coefficients for each photodetector. It would be worthwhile to find a single calibration function with the minimum number of independent parameters from one photodetector to other. With this goal, and after a careful inspection of the data in Table 1, correlations were found among them that could simplify the proposed model even more. In fact, if Ai (i ) 1-4) are plotted as a function of A0, and fitted to linear functions, Ai ) miA0 + di (i ) 1-3), acceptable agreements are found as shown in Table 2. Only the linear fit between A4 and A0 was not very good. But as the A4 term is the least significant in eq 6, the final comparison between this model and the experimental results are quite acceptable, as we will show below. Accepting the results of Table 2, the A1, A2, and A3 coefficients can be calculated from A0. Hence, the calibration function is unified for all the photodetectors and it can be written with only one free parameter, A0. Therefore, the oxygen concentration can be extracted from eq 6 as a function of temperature, count number N, and A0, i.e., [O2] ) f (N, T, A0). This finding is quite remarkable because a single standard calibration method can be achieved. To do so, the fits shown in Table 2 are included in eq 6, obtaining for A0

A0 )

[O2] - (d1T + d2N-1/2 + d3TN-1/2 + d4T2N-1/2) 1 + m1T + m2N-1/2 + m3TN-1/2 + m4T2‚N-1/2

(7)

Evaluating eq 7 for a known temperature, an oxygen concentration, and its corresponding count number N, we are able to calculate the remaining Ai coefficients by means of Table 2, because the parameters mi and di are common to all the photodetectors. Therefore, the calibration function can be easily reconstructed with a single oxygen concentration at a measured temperature. In particular, the calibration process can be carried out with an open air sample ([O2] ) 20.95%), requiring no more laboratory-prepared standards. To sum up, the proposed calibration method consists of the following steps: (1) exposition of the portable instrument to an open air atmosphere, (2) acquisition of the temperature and the count number by the instrument; (3) evaluation of eq 7 with [O2] Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

3177

Figure 5. Differences between the laboratory standards and the readings of our system of oxygen concentrations shown for one of the photodetectors studied as a function of temperature. Lines are only a guide for the eye.

Figure 7. Comparison between the calibration function (lines) and experimental data (symbols) for sensing films with different compositions. ([) Intensities from PtOEP in PVC; (9) count numbers from PtOEP in PVC; (2) intensities from PtOEP in silicone; and (4) lifetimes from PtOEP in PS;27 intensities from (0)N-926 in PS29 and from (+) (x) Os(dpp) in PS 368.5 at 630 and 650 nm, respectively;30 (]) lifetimes from Ru(dpp) in PVC.28 Table 3. Oxygen Determination with the Portable Instrument with the Proposed Single Standard Calibration Function Using an Available Commercial Instrument as a Reference described instrument

Figure 6. Comparison between the proposed calibration function (lines) and experimental data (symbols) for the sensing film with several thicknesses and different concentrations of PtOEP in PS. (]) 12 µm with 500 mg/L; (O) 23 µm with 500 mg/L; (4) 36 µm with 500 mg/L and (9) 300 mg/L and 23 µm; ([) 700 mg/L and 23 µm.

) 20.95% by the built-in microcontroller; (4) calculation of the coefficients of the temperature-compensated calibration function (eq 6) with Table 2; (5) now, the instrument is ready to measure the oxygen concentration. Therefore, the main advantage of this procedure is that a single standard from open air is only needed to obtain a thermal compensated calibration of the portable instrument away from laboratory conditions and without trained personnel. Validation of the Calibration Procedure. The calibration algorithm explained above was tested exhaustively. We applied it to the set of coated photodetectors under study, for a temperature range from 0 to 45 °C and oxygen concentrations up to 24%. We investigated if there was any poorly calibrated instrument operation range. Let us consider the absolute deviation, ∆, as the absolute value of the difference between the standard experimental oxygen concentration, [O2]std and the reading from our calibrated portable instrument, [O2]cal:

∆ ) |[O2]std - [O2]cal| 3178

Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

(8)

reference instrument

[O2] (%)

X

S

X

S

Pval (%)

24 20 16 12 8 open air (20.95%)

24.2 20.3 17.0 12.3 8.2 21.0

0.058 0.289 0.500 0.289 0.289

24.1 19.9 16.1 11.9 7.8 20.9

0.006 0.115 0.115 0.265 0.115

72.5 11.1 7.6 12.8 9.0

Figure 5 represents the typical behavior of this absolute deviation as a function of the temperature for several oxygen concentrations for one of the studied photodetectors. We can observe that there is no operation range where the calibrated instrument gives worse readings. An overall absolute error of 0.3% (oxygen concentration) was obtained with a confidence level of 95% in the set of photodetectors studied. It should be noted from Figure 5 that zero deviation is obtained for open air oxygen concentration at room temperature (22 °C), as expected. The calibration function was also applied to data with photodetectors with different amounts of coating sensing film, i.e., different sensing film thicknesses and different luminophore concentrations. In the first case, three groups with three photodetectors each were prepared, depositing 5, 10, and 15 µL of sensing film. The resulting average thicknesses were of about 12, 23, and 36 µm, respectively. In the second, two groups with three photodetectors were coated with sensing films (23-µm average thickness) with 300 and 700 mg/L PtOEP, respectively. The calibration procedure was applied to these sets of devices using the same coefficients in Table 1. The maximum overall absolute error (defined as above) was 0.35%, thus, showing very good agreement and demonstrating the independence of the pro-

posed model on thickness and luminophore concentration in the ranges analyzed. Figure 6 summarizes this analysis, where the experimental data (symbols) are fitted to eq 6 (lines) at room temperature. Data from other sensing film compositions were analyzed with eq 6 at room temperature. In Figure 7, the data of the sensing film prepared in our laboratory with different polymers are displayed and fitted to the proposed function. Good agreements were obtained for phosphorescence intensity and count number measurements from PVC membranes and for intensity data from silicone membranes. Figure 7 also shows experimental data from the literature coming from different oxygen dyes and analytical parameters, I and τ, which were successfully fitted to eq 6; namely, lifetime data from an organic light-emitting device based on PtOEP in PS27 and from a membrane of ruthenium(II)-tris-4,7-diphenyl1,10-phenanthroline (Ru(dpp)) in PVC;28 and otherwise, intensity measurements from iridium(III)-bis(2-phenylpyridinyl)-N,N,N,Ntetramethyl-(4,4-diamine-2,2_-bipyridinyl) chloride (N-926) in PS29and from osmium(II)-tris-4,7-diphenyl-1,10-phenanthroline chloride (Os(dpp)) in PS 368.5 at 630 and 650 nm.30 To test the usefulness and validity of the work prototype, several gaseous standards with different oxygen percentages covering the dynamic range were measured. The portable instrument results were compared with those of the microPac Plus commercial compact personal gas detection equipment from Dra¨ger based on an electrochemical sensor. Table 3 shows the results as the mean values from three determinations of each gaseous standard, the standard deviation of the replicates measured, and the probability value (Pval) of the test used for comparison of the results obtained from both methods. The results indicate that there is no significant difference between the means obtained from both methods (Pval > 5%). CONCLUSIONS A single standard calibration procedure with temperature compensation has been successfully proposed and applied to (27) Shinar, R.; Zhou, Z.; Choudhury, B.; Shinar, J. Anal. Chim. Acta 2006, 568, 190-9. (28) Andrzejewski, D.; Klimant, I.; Podbielska, H. Sens. Actuators, B 2002, 84, 160-6. (29) Fernandez-Sanchez, J. F.; Roth, T.; Cannas, R.; Nazeeruddin, Md.; Spichiger, S.; Graetzel, M.; Spichiger-Keller, U. E. Talanta 2007, 71, 242-50. (30) Xu, W.; Kneas, K. A.; Demas, J. N.; DeGraff, B. A. Anal. Chem. 1996, 68, 2605-9.

calibrate an optical sensor of PtOEP coating a solid-state photodetector. This coated photodetector forms part of a complete portable instrument for the determination of the oxygen concentration. An exhaustive study with 11 units of this sensor was carried out to find the functional dependences of the different magnitudes involved. Different previous calibration strategies were tested, showing good agreement with these models but needing several temperature or oxygen concentration standards. To simplify the method, an empirical calibration function has been proposed with only one free parameter, which also allows temperature compensation. With this characteristic it is possible to make the instrument calibration with a single standard, which can be obtained with exposure to the open air. The consequences of this fact are clear: (i) laboratory standards are not required and (ii) trained personnel for calibration are not necessary. Those simplifications make a wider usage of this kind of sensors possible for oxygen determination. An overall deviation of 0.3% in oxygen concentration was obtained after a full comparison with a laboratory standard in the whole operation range. Moreover, good fittings were obtained with different film compositions and thicknesses and other luminophores. Finally, our calibrated measurement system was validated by comparison with an available commercial oxygen detector based on an electrochemical sensor. Good agreement between results from both instruments was achieved at room temperature. This possibility of open air calibration provides more advantages to our measurement system, because interferences from other noncompensated magnitudes may be reduced. For example, if the instrument has to be used in an environment with different pressure or humidity than a previous calibration, a new calibration prepares the instrument for proper operation in these new conditions. ACKNOWLEDGMENT We acknowledge financial support from the Ministerio de Educacio´n y Cultura, Direccio´n General de Ense˜nanza Superior (Spain) (Projects CTQ2005-09060-CO2-01 and CTQ2005-09060CO2-02). Received for review November 27, 2006. Accepted February 19, 2007. AC062246D

Analytical Chemistry, Vol. 79, No. 8, April 15, 2007

3179