Trace Gas Measurement with an Integrated Porous ... - ACS Publications

Our present interest is in small, integrated trace gas measurement systems in which the gas of interest is collected by diffusion/permeation into a se...
0 downloads 0 Views 150KB Size
Download PDF
Anal. Chem. 2003, 75, 4050-4056

Trace Gas Measurement with an Integrated Porous Tube Collector/Long-Path Absorbance Detector Kei Toda,* Ken-Ichi Yoshioka, and Shin-Ichi Ohira

Department of Environmental Science, Faculty of Science, Kumamoto University, 2-39-1, Kurokami, Kumamoto 860-8555, Japan Jianzhong Li and Purnendu K. Dasgupta*

Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061

Porous membrane tubes filled with an absorbing solution that change colors upon selective reactions with specific gases provide high sensitivity inexpensive gas sensors. These can be routinely used for ambient monitoring in a fully automated manner. We consider both stopped and continuous flow operations and show the superiority of the stopped flow mode theoretically and experimentally. Light throughput through various membrane tubes is presented, and superior performance of such tubes over Teflon AF is shown. Sensors for NO2 and for O3 were based on Griess-Saltzman and indigotrisulfonate chemistries, respectively. A computer-controlled two-LED absorbance measurement system (one wavelength monitors the signal, the other references the system) that also governs automated reagent refilling was implemented. Sub-parts-per-billion-volume detection limits are attainable within a few minutes for both gases. Comparative data with a commercial UV-photometry-based ozone monitor showed excellent agreement with the response pattern of the present instrument. Low cost, ready applicability to the measurement of different gases by merely changing the light source and chemistry, and high sensitivity makes this instrument attractive for both pedagogic and practical purposes. Liquid core waveguides (LCWs) have recently emerged as an attractive tool for making simple but sensitive absorbance and fluorescence measurements.1 An LCW permits efficient propagation of light in a core liquid bounded by a transparent tube of a refractive index (RI) lower than that of the liquid, thus functioning as a liquid-filled optical fiber. Stone et al. originally proposed the concept three decades ago.2 At that time, there was no convenient material that had a RI less than that of water (1.33). A new amorphous fluoropolymer (Teflon AF, hereinafter called AF) with an RI as low as 1.29 became commercially available by the early * Corresponding authors. E-mail addresses: [email protected] (K. Toda), [email protected] (P. K. Dasgupta). (1) Tsunoda, K.; Umemura, T. Bunseki 2001, 668-673. (2) Stone, J. Appl. Phys. Lett. 1972, 20, 239-243. Walrafen, G. E.; Stone, J. Appl. Spectrosc. 1972, 26, 585-589. Stone, J. J. Chem. Phys. 1978, 69, 4349-4356.

4050 Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

1990s.3 The interest in and the applications of water-core LCWs have increased greatly ever since.4 Our present interest is in small, integrated trace gas measurement systems in which the gas of interest is collected by diffusion/ permeation into a selective chromogenic reagent contained in a tube, and the optical absorption is measured axially in the same tube. AF not only exhibits a low RI that makes possible a waterfilled LCW, its amorphous structure and high free volume makes for higher gas permeability than other types of Teflon.5 Earlier attempts have thus been made toward gas sensors based on an AF tubes.6 Nevertheless, there are disincentives, as well. On an absolute scale, the permeability of AF for many gases is quite small. In a recent study to measure H2S by a fluorescence method,7 the measured transfer rate for ∼100-µm-thick AF, silicone, and Nafion membranes was 16, 200- and 9300 nmol/cm2/s/atm, respectively. The actual membrane selected for use in that study, an expanded porous poly(tetrafluoroethylene) (ePTFE) material, exhibited an even greater transfer rate and permitted the measurement of parts-per-trillion-volume levels of the gas. In an absorptiometric gas measurement device containing a stationary liquid, the sensitivity increases linearly with both the tube length and the gas transfer rate. If a tube does not function as an LCW, it must be used in relatively short lengths (e.g., 1-5 cm) to prevent the detector from becoming light-starved. However, it will still produce the same sensitivity as a 50-cm-long AF tube if the gas transfer rate is 50-10 times greater). These considerations become even more relevant in fabricating affordable devices. At a cost of several times that of gold, AF is the most expensive commercially available polymer tube. Although AF may be the (3) http://www.dupont.com/Teflon/af/index.html. (4) Byrne, R. H.; Yao, W.; Kaltenbacher, E.; Waterbury, R. D. Talanta 2000, 50, 1307-1312. Waterbury, R. D.; Yao, W.; Byrne, R. H. Anal. Chim. Acta 1997, 357, 99-102. Dress, P.; Belz, M.; Klein, K.; Grattan, K. T. V.; Franke, H. Sens. Actuators, B 1998, 51, 278-284. Datta, A.; Eom, I.; Dhar, A.; Kuban, P.; Manor, R. M.; Ahmad, I.; Gangopadhyay, S.; Dallas, T.; Holtz, M.; Temkin, H.; Dasgupta, P. K.; IEEE Sens. J., in press. (5) Pinnau, I.; Toy, L. G. J. Membr. Sci. 1996, 109, 125-133. Yu, A.; Yampolskii, Y. P.; Shantarovich, V. P.; Nemser, S. M.; Plate´, N. A. J. Membr. Sci. 1997, 126, 123-132. (6) Milani, M. R.; Dasgupta, P. K. Anal. Chim. Acta 2001, 431, 169-180. Dasgupta, P. K.; Genfa, Z.; Poruthoor, S. K.; Caldwell, S.; Dong, S.; Liu, S. Anal. Chem. 1998, 70, 4661-4669. (7) Toda, K.; Dasgupta, P. K.; Li, J.; Tarver, G. A.; Zarus, G. M. Anal. Chem. 2001, 73, 5716-5724. 10.1021/ac0341719 CCC: $25.00

© 2003 American Chemical Society Published on Web 07/04/2003

only choice in some applications, there may be better alternatives for gas sensing. Many fluorocarbons have refractive indices close to that of water. The RI of FEP Teflon (fluorinated ethylene propylene copolymer) is 1.34. Depending on the morphology of the pores, the presence of air-filled pores can cause the effective RI value to be lower than that of the matrix. Although bulk silica has a RI of 1.45, nanoporous silica can have RI values as low as 1.2.8 Peterson and Dasgupta showed that water-filled porous PTFE and polypropylene tubes may not behave as LCWs but can still conduct usable amounts of light over several cm.9 The absolute brightness of inexpensive solid state light-emitting diode (LED) sources has also increased greatly in recent years and all indications are that it will continue to increase further in the foreseeable future. With brighter sources, the precise light-conducting efficiency of a gas collection membrane tube becomes less important. In this work, we show the attractive performance of LED/ photodiode-interrogated absorptiometric porous membrane-based gas sensors for two criteria pollutant gases, nitrogen dioxide and ozone. PRINCIPLES In the presently proposed geometry, a cylindrical liquid-filled porous tube of length L cm and outer diameter do cm is disposed concentrically within a jacket tube. The air sample flows parallel to the axis of the tube at a rate of Q cm3/min around the porous tube. This geometry has previously been studied.10,11 At very low flow rates, the collection efficiency, f, is essentially quantitative, and the mass of the analyte collected, m, linearly increases with the sampling rate, Q, with zero intercept. At higher flow rates, the system follows a generic Gormley-Kennedy12 (G-K) behavior,

1 - f ) Ae-BDL/Q

(1)

much lower than the ideal G-K value. For most practical devices and operating conditions, the fraction collected is small enough such that as a first approximation, the entire length of the device can be assumed to be exposed to the same analyte gas concentration, Cg. Under these conditions, the total mass of analyte collected by the device often shows little or no dependence on the gas flow rate because the decrease in f with increasing Q makes fQ almost constant.10 Second, the mass collected in time t is proportional to the exposed surface area of the device πdoL,

m ) kπdoLCgt

where the constant k is dependent on the diffusion coefficient of the gas and the parameters that govern the membrane sink efficiency. The analyte concentration in the solution, Cs, is obtained by the collected mass divided by the internal volume V of the tube (this is the same as the effective liquid volume without liquid flow, referred to as the stopped flow (SF) mode).

Cs ) m/V

(8) Horvath, R.; Pedersen, H. C.; Larsen, N. B. Appl. Phys. Lett. 2002, 81, 21662168. (9) Dasgupta, P. K.; Peterson, K. Spectroscopy 1987, 2, 50-51. (10) Dasgupta, P. K.; Dong, S.; Hwang, H.; Yang, H.-C.; Genfa, Z. Atmos. Environ. 1988, 22, 949-963. (11) Dasgupta, P. K. Automated Diffusion Based Collection and Measurement of Atmospheric Trace Gases. In Sampling and Sample Preparation Techniques for Field and Laboratory; Pawliszyn, J., Ed.; Wilson and Wilson’s Comprehensive Analytical Chemistry Series, Vol. XXXVII; Elsevier: New York, 2002; pp 97-160. (12) Gormley, P. G.; Kennedy, M. Proc. R. Ir. Acad. Sci. 1949, 52A, 163-169.

(3a)

In the continuous flow (CF) mode, the effective volume in which the analyte dissolves is Ft, where F is the continuous liquid flow rate.

Cs ) m/Ft ) kπdoLCg/F

(3b)

For the SF case, if the membrane wall is thin relative to the membrane diameter, the outer and inner diameters can be approximated to be the same parameter, d, such that eq 3a becomes

Cs ) 4kCgt/d where A and B are positive constants (0 < A < 1) and D is the diffusion coefficient of the sampled gas, such that DL/Q is a dimensionless quantity. It may be observed that f decreases with increasing Q, and at high values of Q, f attains a limiting constant value of 1 - A (0.18 for the G-K equation). The G-K equation itself applies to an analyte gas flowing through a cylindrical tube where the interior surface of the tube is a perfect sink for the gas. This represents an ideal maximum. In the case of the present system, the analyte gas flows through an annular space where only the inside surface of the annulus is a sink for the gas, the sink efficiency depending on the fractional porosity, pore size, pore tortuosity, wall thickness, the interior liquid composition, and the interior liquid uptake efficiency for the analyte gas. Overall, the combination of these factors tends to make the value of 1 - A

(2)

(4a)

In both cases, the observed absorbance A is related by Beer’s law to the product of the absorptivity, ; optical path length, L; and Cs; for SF and CF modes, we respectively get

A ) 4kLCgt/d

(5a)

A ) kπdoL2Cg/F

(5b)

From these principles, at constant Cg, other parameters being the same, the response is proportional to L and L2 in the SF and CF modes, respectively. The mathematical identity of the two modes can be easily verified by equating F and V/t. The SF approach is generally superior in terms of performance characteristics. Operationally, it does not require a liquid phase pump that must maintain a stable low flow rate. In practical systems, this can be the most expensive and bulky system component. A SF system is easily automated by having a gravityfed reagent reservoir, with a solenoid valve at the liquid inlet or outlet that opens for a short period of time for reagent replacement. This can occur either at fixed periods or at set absorbance levels that indicate reagent exhaustion. Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

4051

In SF systems, the absence of liquid flow makes liquid phase diffusion the only motive force for mixing. The overall sensitivity is improved by a reduction in diameter of the device, as most readily apparent from eq 4a. The characteristic radial mixing time will be d2/4DL, where DL is the liquid phase diffusion coefficient (the limiting value will be typically determined by the larger reagent molecule, as opposed to the smaller analyte molecule). Assuming some typical values, for example, 10-5 cm2/s for DL, and allowing a characteristic mixing time of 60 s, the upper limit of the tube diameter one computes is ∼0.5 mm. Considerations on light throughput largely set the lower limits on d. At too small a tube diameter, the light throughput becomes the limiting factor, because the detector noise becomes too high. Other than d, the parameters that govern the detector noise include device length, the LED intensity, the nature of the detector, and the electronic design. Thus, an optimum value of d cannot be computed a priori. However, practical issues, such as total reagent requirements, etc., also dictate that actual devices are built with d values in the 0.252.5-mm range. The desirability of a smaller diameter device stems from the same factors that call for higher surface/volume ratios in a membrane collector of any geometry,13 because Cs increases with an increase in this ratio. EXPERIMENTAL SECTION Reagents. Sodium hydroxide, acetic acid, and sulfanilic acid (Nacalai Tesque), N-1-naphthylethylenediamine dihydrochloride (NEDH, Wako), sulfuric acid, and potassium indigotrisulfonate (ITS, Aldrich Chemical) were used without further purification. All were reagent grade. The Griess-Saltzman (GS) reagent for the determination of NO2 consisted of 5.0 g of sulfanilic acid; 50 mL of glacial acetic acid; 50 mg of NEDH; and 1 mL of Zonyl FSN (Dupont), a fluorosurfactant, in 1.0 L of solution. Distilled deionized water was used throughout. The ITS reagent used for ozone determination was 20 µM in concentration and adjusted to pH 2 with sulfuric acid; this reagent was stored in an Al-foilwrapped bottle to avoid photobleaching. NO2 Collector/Detector. The collector/detector is shown in Figure 1a. The NO2 detector was based on a porous PTFE tube (Poreflon,14 Sumitomo Electric Fine Polymer, 50% porosity, 2-mm i.d., 3-mm o.d.) MT. The tube MT was connected between two plastic tees, T, and 1.5-mm-diameter acrylic optical fibers SF and DF were inserted into the tee ports to about the center of the tee. The entire assembly was placed in a poly(vinyl chloride) (PVC) block B containing a 6-mm-wide and -deep groove that surrounds MT. Block B was sealed at the top with a PVC cover plate, C, and a silicone gasket, S (0.5 mm), in between. The gas outlet/inlet (GO/GI) connections were provided on the cover plate as shown. Apertures at optical fiber and tee exits were sealed with silicone adhesive. At the source fiber end, the 1.5-mm fiber was coupled to two 0.5-mm-diameter individual fibers that went, respectively, to a 525-nm LED (DG3803X, Stanley Electric, 1200 mcd at 20 mA) and a 875-nm LED (HSDL4230, Agilent, 16 mW at 50 mA), driven at 20 and 35 mA, respectively, via MOSFET switches controlled by a notebook computer. The detector fiber DF was connected to a photodiode PD containing an integrated operational amplifier-based current-to-voltage converter (OPT301, (13) Ohira, S.; Toda, K.; Ikebe, S.; Dasgupta, P. K. Anal. Chem. 2002, 74, 58905896. (14) http://www.sei-sfp.co.jp/english/product/index.html.

4052 Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

Figure 1. (a) (upper) Schematic diagram of the NO2 collector detector system. Liquid was made to flow by gravity and flow on/off was controlled by solenoid valve, SV. RB, reagent bottle; SL, soda lime column; TT, Tygon tube (0.8 mm i.d. × 2.4 mm o.d.); T, plastic tee; MT, 2 mm i.d. × 3 mm o.d. Poreflon tube; W, waste; GI, gas in; MF, miniature (M5) fittings; GO, gas out; SM, silicone sealant; LED1,2, green and IR LEDs; OC, optical fiber coupler; SOF, small optical fiber (core 0.5 mm); SF, large optical fiber (core 1.5 mm), source fiber; DF, detector fiber; B, PVC block; C, PVC cover plate; PD, photodiode/ op-amp. (b) (lower) Ozone detection system. Symbols same as a, plus VT, charcoal vent trap; F, 25-mm laptop PC style suction fan; ST, 30-cm3 syringe body, all was housed in an opaque plastic enclosure.

Texas Instruments) operated at a gain of ∼5 × 108 V/A with all necessary components mounted on a single circuit board for compactness and minimum noise. The reagent bottle, RB, was protected from intrusion of NO2 by a soda lime trap, SL. The reagent flowed by gravity through the device when normally opened solenoid valve, SV, was off. SV was always off in the CF mode; in the SF mode, SV was turned on during the accumulation period. Ozone Sensor Construction. The ozone sensor is schematically shown in Figure 1b. The following differences from the NO2 sensor are noteworthy. A 25-mm-diameter fan, F, operated at 4.2 V (intended for cooling laptop PCs, TT-25, Eagletech Computers, Pearl River, NY) provides the necessary suction (23 L/min) for sampling air. The fan, F, sits at one end of a tube, ST, made of a 30 cm3 disposable plastic syringe barrel, 21.5-mm i.d., the whole enclosed inside a black plastic enclosure that protects the system from ambient light. The reagent bottle is protected from O3 with activated carbon as a vent trap (VT); it is also wrapped in Al foil to prevent photobleaching. Membrane tubes MT used in this case were ePTFE (0.005-in. wall, 0.040-in. i.d, Zeus15) and Accurel PP (1.75-mm i.d., porous polypropylene, Enka). The optical fibers (1mm core) came up to the membrane tube termini inside the tees (inserted inside the membrane for membrane inner diameters >1.0 mm) and resulted in significantly better light throughput as (15) http://www.zeusinc.com/product_sheets/ePTFE/ePTFE_tubing.html.

compared to the NO2 detector geometry, even with the smaller diameter optical fibers and the membrane tubes. The LEDs used in this case had center emission wavelengths of 600 nm (manufacturer nominal specification is 592 nm, EFY5366X, 2800 mcd at 20 mA) and 850 nm (DN305, 10 mW at 50 mA); both were from Stanley Electric and driven at 20 mA. Excess plastic was removed from the LEDs, and they were coupled together to the same fiber by optical grade epoxy without the use of a splitter. As test gas, NO2 from a cylinder-based standard and ozone generated by a low-pressure Hg lamp were diluted as needed with zero air using mass flow controllers; details have been given previously.13,16 With both detectors, custom software provided for data acquisition and computation of absorbance based on log[(Vr - Vb)/(Vs - Vb)], where Vr, Vs, and Vb, respectively, represent the detector voltage with the reference (higher wavelength) LED on, the signal LED on, and neither of them on. In auxiliary experiments, optical fiber-based spectrometers (S2000, Ocean Optics, Dunedin, FL or CDI-PDA, Control Development Corp., South Bend, IN) were used as detectors. An incandescent light source coupled to a monochromator (MiniChrom, PTR Optics, Waltham, MA) was used to measure the light throughput through individual membrane tubes as a function of wavelength. RESULTS AND DISCUSSION Choice of Tube for Collector/Detector. Light transmission data for different water-filled tubes are shown in Figure 2a as a function of length and in Figure 2b as a function of wavelength. There is no question that AF tubes transmit much more light. Of the porous tubes shown, in comparable diameters, the Accurel PP tube has the best light transmission. The larger diameter Poreflon tube is a close second and, in fact, transmits more light than the Accurel tube at wavelengths lower than 500 nm. The increased light transmission at higher wavelengths for porous tubes is related to the increased inability of the longer wavelength photons to perceive the pores as discrete structures. The poor performance of the porous poly(vinyledenefluoride) tube is likely due to the relatively large pores in this membrane. Note also that the dip in the transmittance spectra at ∼750 nm observed in all the spectra is not an artifact; it is due to the absorption band of water in the deep red (contrary to descriptions in many textbooks, H2O is not colorless but blue; D2O is, indeed, colorless17) made more easily noticeable here due to the increased path length. We chose the Accurel PP and the larger Poreflon tubes as the respective test platforms to demonstrate applicability to the determination of O3 and NO2. That the porous tubes in general provide better sensitivity for the desired application is amply demonstrated in the comparison between the results observed with an AF tube (17 cm long, 0.6 mm i.d., 100-µm wall) compared to an Accurel PP tube (5 cm long, 1.75 mm i.d., 320-µm wall) filled with the same reagent and exposed to 240 ppbv O3 (Figure 2c). Beer’s Law Behavior in Porous Membrane Tubes as a Function of Length. It should be noted that when adherence to Beer’s law is studied with respect to the tube length, the effect of stray light (which includes any mismatch between the LED emission band and the chromogen absorption band), specular (16) Li, J.; Dasgupta, P. K. Anal. Chem. 2000, 72, 5338-5347. (17) (a) Braun, C. L.; Smirnov, S. N. J. Chem. Educ. 1993, 70, 612-615. (b) http://www.dartmouth.edu/∼etrnsfer/water.html.

Figure 2. Light transmittance characteristics through water-filled tubes. (a) Light attenuation as a function of tube length: (, Teflon AF 2400 (1.064 mm × 1.270 mm); b, Poreflon (2 × 3 mm); 9, Poreflon (1 mm × 2 mm); 2, ePTFE (1.016 mm × 1.270 mm). (b) Light transmission as a function of wavelength for 100-mm lengths of various porous membrane tubes. Tested tubes were 1 Accurel PP (1.75 × 2.40 mm), 2 Poreflon (2 × 3 mm), 3 Poreflon (1 × 2 mm), ePTFE (1.016 × 1.270 mm), and pPVDF (1 × 1.5 mm). (c) Despite poorer light throughput, porous membrane tubes provide better performance than Teflon AF as a result of their substantially superior gas transport properties. Dashed line indicates the start of exposure to 240 ppbv O3. ITS concentration was 2.5 µM.

reflection along the walls, etc. can lead to an absorbance less than that computed on the basis of the length alone. Obviously, the effect becomes more apparent when the overall absorbance is high and light throughput is low. For example, in the NO2 measurement system, for a 17-cm tube filled with an absorbing reagent containing 0.5 µM NO2- and exhibiting an apparent absorbance of ∼0.4 represents 85% of what would be calculated On the basis of observations in a 1-cm tube. Choice of Signal and Referencing Sources and the Referencing Strategy. Figure 3a shows the absorption spectra of the GS reagent with 0, 10, and 20 µM NO2- added to it with the emission spectra of the green and the near-infrared (NIR) LEDs used. The reagent blank does not have significant absorbance above 400 nm. With addition of NO2- (this forms the same product that results upon the absorption of gaseous NO2), the broad absorption due to the analytical product is centered at 540 nm. Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

4053

Figure 3. Absorbance spectra of the product/initial reagent and signal/reference LED emission spectra. (a) Solid lines indicate absorbance spectrum of Griess-Saltzman reagent with 0, 5, and 10 µM NO2-. Dashed lines are green (signal) and IR (reference) LED emission spectra. (b) Solid lines indicate absorbance spectrum of acidic potassium indiogotrisulfonate reagent, dashed lines are orange (signal) and IR (reference) LED emission spectra.

The emission band of the green LED shows excellent overlap with the analyte absorption and is well-suited for monitoring the analyte. The emission band of the 875-nm NIR LED used did not overlap with the analyte product absorbance. Figure 3b shows essentially the same situation for ozone measurement. In this case, the analytical reaction involves loss of color; ITS is selectively bleached by ozone. The absorption band of the reagent blank is wellmatched by the emission band of the signal LED and the NIR LED used for reference emits at a nonoverlapping wavelength. Numerous strategies are possible to separately acquire the signals of the signal and reference LEDs. Piezo fiber modulators18 are not very practical in laboratory experiments. Ultimately, we chose the simplest alternative, turning on the signal and the reference LEDs alternately for a fixed period (2-5 s) each. In some experiments, a dark current measurement period of 5 s was incorporated after each 10 s period of alternate LEDs being turned on. In others, the algorithm called for measuring the dark current during the reagent refilling time. Compared to the previous referencing with the second detector on the back of LED,6 the present reference system with switching two LEDs and one photodiode detector can compensate for changes in light throughput due to physical changes in the tube. The same ends as a sensor based on a broadband light source and multiwavelength detector19 are more easily achieved with an independent LED. Response to NO2 in CF and SF Modes. As discussed in the Principles Section, the response is expected to increase initially (18) http://www.piezojena.com/start.html. (19) Li, Q.; Morris, K. J.; Dasgupta, P. K.; Raimundo, I. M., Jr.; Temkin, H. Anal. Chim. Acta 2003, 479, 151-165.

4054 Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

Figure 4. Typical response curves obtained in (a) SF and (b) CF modes. Poreflon tube, 2 × 3 × 50 mm, gas flow 0.4 SLPM, SF mode 3 min stop time, CF mode 300 µL/min. (c) Performance of the SF mode at low levels.

linearly with the gas sampling rate and then reach a plateau value. Exactly this is observed in the present case, that is, the response was proportional to the flow rate until 0.3 standard liters per minute (SLPM) and was constant over the flow rate for NO2. On the basis of these results, the sampling rate chosen for further experiments was 0.4 SLPM. At this flow rate, the mean gas velocity around the membrane tube is computed to be 23 cm/s. Figure 4 shows data from both SF and CF mode operations. In the SF case, triplicate experiments were conducted at each concentration. The liquid flow was stopped for 3 min, with an attendant increase in the absorbance. The liquid was then replenished by opening solenoid valve SV. Trace 4a (SF mode) thus shows peaklike responses. In the CF mode, a steady state is achieved in the analyte concentration on the basis of the analyte input rate (concentration) and the solution flow rate. Plateau responses are thus observed in the duplicate experiments shown in trace 4b. In both modes, the amplitude of the peak or plateau response is proportional to the analyte concentration. Obviously, the SF mode is more sensitive. Trace 4c shows response at low levels of NO2 in the SF mode, with an accumulation period of 3 min. We infer an LOD of 0.4 ppbv on the basis of these data, which is adequate for routine monitoring of ambient air for establishing compliance to air quality regulations. Linear calibration curves were obtained with tubes from 10 to 168 mm in both SF and CF modes. The higher response behavior in the SF compared to the CF mode is directly attributable to

Figure 5. Tube length dependence on (a) SF and (b) CF mode response. The NO2 concentrations were O, 10 ppbv; 9, 20 ppbv, ], 40 ppbv; 1, 60 ppbv; 4, 80 ppbv; and b 100 ppbv. Note that the ordinate of the panel b is the root of absorbance.

Figure 6. Ozone detection system. Response at three different concentrations with automated 30-s-long reagent refill (a) as a set absorbance (∼0.3 AU) is reached and (b) after a preset time of 9.5 min. 1.75 × 2.40 × 50 mm Accurel PP tube. SF mode. Flow velocity ∼100 cm/s.

differences in the residence times of the liquids. Figure 5a and b shows the relationship between the response and the tube length. The response increases with the length of the tube in a manner accurately predicted by eqs 5a and 5b, respectively. Note that the response in the CF mode was proportional to L2. The slight departure from linearity in the SF mode at long tube lengths is a consequence of the stray light and like effects, as previously discussed. Response to Ozone. Since the SF mode was not only expected theoretically to be more attractive but was indeed observed in the NO2 case to be so, we present only SF data for ozone. The flow rate used corresponds to a velocity of 100 cm/s. A higher sampling rate was used to avoid flow-limited response, since the ozone uptake was perceived to be faster than that of NO2. Figure 6a shows results from an automated arrangement in which the device was exposed to varied ozone concentrations, and liquid refill was programmed to occur soon as a preset low

absorbance value (ca. 0.3 AU, using a simple voltage comparator based circuit). It is interesting to note that in this arrangement, the refill frequency increases with increasing analyte concentration, because the interval of time before refill occurs is directly proportional to the cumulative exposure (CE) of the device (e.g., in parts-per-billion times minute) during that interval. Obviously, the sum of these intervals over any given period is also proportional to the CE or to the time-weighted average (TWA) concentration over that period. When operated in this mode, the amount of waste liquid generated during any given sampling period is proportional to the average concentration in that period. In addition, since the sampling flow was not turned off during the refill period, the starting absorbance decreases as the sampled ozone concentration is increased. Figure 6b shows the results from an arrangement in which the liquid refill occurs on a constant periodic basis, the chosen interval being based on maximum anticipated concentration of Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

4055

Figure 7. Ozone detection system. Ozone concentrations as measured with an U.S. EPA equivalent UV photometric instrument (dashed line) compared to the absorbance slope produced by the present instrument (solid line). Lubbock, TX, January 18-19, 2003. Other conditions as in Figure 6.

ozone such that this concentration will not result in saturation of the device’s response (in this case, a 9.5-min sampling period was followed by a 0.5-min refill period.). In this case, the TWA/CE during each interval is proportional to the absorbance change (absorbance just after refill minus that just before the next refill) during each interval. In both cases, the instantaneous concentration is proportional to the slope of the absorbance change, that is, to -dA/dt. Figure 7 shows ozone concentration in ambient air in Lubbock, TX, being monitored by a commercial UV photometric ozone instrument in comparison to the first derivative of the absorbance signal being produced by the present instrument operated in the constant periodic replacement mode. Considering low ozone levels in our semirural location, the ability of the present instrument to follow the temporal ozone profile is remarkably good and amply demonstrates the power of this very simple and inexpensive approach. It is worthwhile to recognize that the transition between linear to an exponentially decaying response slope is due to the relatively large diameter membrane tubes used in these experiments and slow liquid-phase diffusion. It takes a long time for the gas coming through the surface of the membrane to penetrate to the very center of the tube, where the last of the unbleached dye remains. This diffusive approach is an exponential process that we shall discuss in detail elsewhere.

The limit of detection (LOD, S/N ) 3) is dependent on the accumulation time. For a period as short as 0.5 min, the LOD is 2.5 ppbv. With integration times of 1.5, 2.5, 3.5, 4.5, and 5.5 min, the LOD improves stepwise to 1.2, 0.8, 0.6, 0.5, and 0.4 ppbv. After this, the improvement is marginal: by 9.5 min, an LOD of 0.3 ppbv is reached.

(20) Levaggi, D. A.; Feldstein, M. Am. Ind. Hyg. Assoc. J. 1964, 25, 64-66. Lambert, J. L.; Chiang, Y. C. Anal. Chem. 1983, 55, 1829-1830. Selavpathy, P.; Pitchai, R.; Ramakrishna, T. V. Talanta 1990, 39, 539-544. (21) Mottola, H. A. Personal communication, 1995.

Received for review February 20, 2003. Accepted May 27, 2003.

4056

Analytical Chemistry, Vol. 75, No. 16, August 15, 2003

CONCLUSIONS Excellent performance can be attained with color reactionbased gas sensors that rely on a porous tubular membrane as a gas collector. Inexpensive but fully automated sensing systems are possible that are capable of monitoring ambient levels of criteria pollutant gases at a short time resolution. It should be possible to apply the same principle to the measurement of numerous other gases of interest, for example, CO via the formation of Ag or Pd sols,20 Cl2 via its reaction with tetramethylbenzidine,6 H2S and mercaptans via their reaction with metallofluorescein derivatives,7 SO2 via the formation of Fe(II)(o-phenanthroline)3 from Fe(o-phen)3,21 etc. We plan to soon report on these and related developments, implementation of the same principles in a microfabricated format, and more computational simulations that describe in more detail the response behavior.

AC0341719