Anal. Chem. 2006, 78, 1890-1896
Hybrid Fluorometric Flow Analyzer for Ammonia Natchanon Amornthammarong,† Jaroon Jakmunee,‡ Jianzhong Li, and Purnendu K. Dasgupta*
Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061
We describe a robust, highly sensitive instrument for the determination of ambient ammonia. The instrument uses two syringe pumps to handle three liquids. The flow configuration is a hybrid between traditional flow injection (FI) and sequential injection (SI) schemes. This hybrid flow analyzer spends ∼87% of its time in the continuous flow FI mode, providing the traditional FI advantages of high baseline stability and sensitivity. The SI fluid handling operation in the remaining time makes for flexibility and robustness. Atmospheric ammonia is collected in deionized water by a porous membrane diffusion scrubber at 0.2 L/min with quantitative collection efficiency, derivatized on-line to 1-sulfonatoisoindole, and measured by fluorometry. In the typical range for ambient ammonia (0-20 ppbv), response is linear (r2 ) 0.9990) with a S/N ) 3 limit of detection of 135 pptv (15 nM for 500 µL of injected NH4+(aq)) with an inexpensive light emitting diode photodiode-based detector. Automated operation in continuously repeated, 8-min cycles over 9 days shows excellent overall precision (n ) 1544 pNH3 ) 5 ppbv, RSD ) 3%). Precision for liquid-phase injections is even better (n ) 1520, [NH4+(aq)] ) 2.5 µM, RSD ) 2%). The response decreases by 3.6% from 20 to 80% relative humidity. Since its development in 1975, flow injection analysis (FIA)1 has been widely used. Sequential fluid handling strategies, typically based on a single bidirectional pump coupled to a multiport distribution valve, were introduced 15 years later2 to provide lower reagent consumption/waste generation and lower maintenance. Sequential injection strategies cannot compete with FIA, however, in sample throughput and, especially, attainable limits of detection (LODs) because FIA provides a continuous reagent blank as a detector background. Gaseous ammonia has long been known to play a key role in atmospheric and biogeochemical processes. It is by far the dominant atmospheric base, responsible for the neutralization of particulate atmospheric acidity,3 such that the major fraction of atmospheric fine particulate matter consists of sulfate, ammonium nitrate, or both.4 Paradoxically, ammonia deposition in soils and subsequent nitrification leads to acidification and may cause long* Corresponding author. E-mail:
[email protected]. † Permanent address: Department of Chemistry, Mahidol University, Bangkok, Thailand 662-2015120. ‡ Permanent address: Department of Chemistry, Chiang Mai University, Chiang Mai, Thailand 50200. (1) Ruzicka, J.; Hansen, E. H. Anal. Chim. Acta 1975, 78, 145-157. (2) Ruzicka, J.; Marshall, G. D. Anal. Chim. Acta 1990, 237, 329-345. (3) Kean, A. J.; Harley, R. A. Environ. Sci. Technol. 2000, 34, 3535-3539.
1890 Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
term imbalances of Ca, K, and Mg; ammonia is also partially converted to N2O, a greenhouse gas.5,6 In recent years, the measurement of gaseous ammonia has become important because of altogether separate needs. A recent review of ammonia sensing techniques is available.7 Low- to subppbv NH3 measurements are needed in semiconductor fabrication facilities to maintain device yield;8-11 there is considerable interest in removing ammonia in clean rooms to levels as low as possible.12 Higher levels of ammonia are of interest in large-scale animal operations because of hygiene, odor, and esthetic concerns.13 In chronic dialysis patients, the efficiency of the dialysis procedure the patient undergoes is presently measured by the blood urea nitrogen assay of the pre- and postdialysis blood samples. Narasimhan et al.14 have shown that the progress of dialysis can be monitored in real time by measuring the ammonia content of the patient’s exhaled breath with a CO2 laser-based photoacoustic spectrometer, which can of course also be used for other ammonia measurement applications.15 Others have also explored laser-based photoacoustic spectrometry for NH3 measurement, with varying degrees of success.16,17 For some time now, this laboratory has had an interest in developing sensitive, reliable, and robust methods for measuring ammonia.18-24 (4) Tolocka, M. P.; Solomon, P. A.; Michell, W.; Gemmill, D.; Weiner, R. W.; Homolya, J.; Natarajan, S.; Vanderpool, R. W. Aerosol Sci. Technol. 2001, 34, 88-96. (5) Committee on the Environment and Natural Resources, Air Quality Research Subcommittee. National Oceanic and Atmospheric Administration. Atmospheric Ammonia: Sources and Fate (A Review of Ongoing Federal Research and Future Needs); NOAA Aeronomy Laboratory; Boulder, CO. 2000. http:// www.al.noaa.gov/AQRS/reports/ammonia.pdf. (6) United States Department of Agriculture. Agricultural Research Service. Action Plan: Component II: Ammonia and Ammonium Emissions. http://www.ars.usda.gov/research/programs/programs.htm?np_code) 203&docid)320, United States Department of Agriculture, July 15, 2005. (7) Timmer, B.; Olthuis, W.; van den Berg, A. Sens. Actuators, B 2005, 107, 666-677. (8) Park, J. C.; Bae, E. Y.; Park, C. G.; Han, W. S.; Koh, Y. B.; Lee, M. Y.; Lee, J. G. Jpn. J. Appl. Phys. 1995, 34, 6770-6773. (9) Hase, U.; Matsuyoshi, Y. NEC Res. Dev. 1998, 39, 95-100. (10) Matsuyoshi, Y.; Satoh, Y.; Shinozaki, T.; Suzuki, E.; Nagata, N. NEC Res. Dev. 2000, 41, 93-97. (11) Satoh, Y.; Matsuyoshi, Y. NEC Res. Dev. 2001, 42, 314-318. (12) Lapkin, A.; Bozkaya, B.; Mays, T.; Borello, L.; Edler, K.; Crittenden, B. Catal. Today 2003, 81, 611-621. (13) Pinder, R. W.; Strader, R.; Davidson, C. I.; Adams, P. J. Atmos. Environ. 2004, 38, 3747-3756. (14) Narasimhan, L. R.; Goodman, W.; Patel, C. K. N. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 4617-4621. (15) Pushkarsky, M. B.; Webber, M. E.; Baghdassarian, O. Narasimhan, L. R.; Patel, C. K. N. Appl. Phys. B 2002, 75, 391-396. (16) Schilt, S.; Thevenaz, L.; Nikles, M.; Emmenegger, L.; Huglin, C. Spectrochim. Acta A 2004, 60, 3259-3268. (17) Schmohl, A.; Miklos, A.; Hess, P. Appl. Opt. 2002, 41, 1815-1823. 10.1021/ac051950b CCC: $33.50
© 2006 American Chemical Society Published on Web 12/27/2005
Figure 1. Diffusion scrubber (DS) and sampling arrangement: SS, stainless steel tube; T, polypropylene tee; G, glass outer jacket; M, microporous polypropylene membrane tube. The inset shows the details of the LED-photodiode-based flow-through fluorescence cell: SP, solution pathway (FEP Teflon tube); M, metal jacket; LED, UV-light emitting diode; OPT301, photodiode-operational amplifier; OF, blue plastic optical filter; PD, photodiode for monitoring source light intensity.
Some of these strategies have been used by others in remote marine areas such as the Tenerife islands to assess the effect of cloud processing on the marine budget of reduced nitrogen compounds25 and also developed further into instruments with excellent LODs.26,27 Classically, ammonia has been colorimetrically determined by Berthelot’s reaction (indophenol blue, LOD 0.6 µM) or Nessler’s reaction (LOD 1.2 µM).28 In 1971, Roth29 introduced the reaction of ammonia, or primary amino acids, with o-phthaldialdehyde (OPA) and borohydride or mercaptoethanol to produce strongly fluorescent compounds; this has been widely used in chromatographic separation and quantitation of amino acids. Other dialdehydes were also developed for the purpose.30,31 In 1989, we reported the superiority of the OPA-sulfite-NH3 reaction for fluorometric NH3 determination in terms of its higher sensitivity and greater selectivity over amino acids,19 and this has remained our favorite chemistry. There are, however, two remaining problems that remain to be solved for long-term maintenance(18) Dasgupta, P. K.; Dong, S. Atmos. Environ.1986, 20, 565-570. (19) Genfa, Z.; Dasgupta, P. K. Anal. Chem. 1989, 61, 408-412. (20) Genfa, Z.; Dasgupta, P. K.; Dong, S. Environ. Sci. Technol. 1989, 23, 14671474. (21) Genfa, Z.; Uehara, T.; Dasgupta, P. K.; Clarke, T.; Winiwater, W. Anal. Chem. 1998, 70, 3656-3666. (22) Li, J.; Dasgupta, P. K. Anal. Chim. Acta 1999, 398, 33-39. (23) Li, J.; Dasgupta, P. K. Genfa, Z. Talanta 1999, 50, 617-623 (24) Genfa, Z.; Dasgupta, P. K. Anal. Chem. 2000, 72, 3165-3170. (25) Milford, C.; Sutton, M. A.; Allen, A. G.; Karlsson, A.; Davison, B. M.; James, J. D.; Rosman, K.; Harrison, R. M.; Cape, J. N. Tellus 1990, 52B, 273-289. (26) Harrison, R. M.; Msibi, I. M. Atmos. Environ. 1994, 28, 247-255. (27) Sørensen, L. L.; Granby, K. Nielsen, H.; Asman, W. A. H. Atmos. Environ. 1994, 28, 3637-3645. (28) Standard Methods for Examination of Water and Wastewater, 16th ed.; American Public Health Association: Washington, DC, 1985. (29) Roth, M. Anal. Chem. 1971, 43, 880-882. (30) Roach, M. C.; Harmony, M. D. Anal. Chem. 1987, 59, 411-415. (31) de Montigny, P.; Stobbaugh, J. F.; Givens, R. S.; Carlson, R. G.; Srinivasachar, K.: Sternson, L. A.; Higuchi, T. Anal. Chem. 1987, 59, 1096-1101.
free operation: (a) accomplish the chemistry in a small footprint affordable format while maintaining a stable detector baseline without the use of tubing-based, maintenance-intensive peristaltic pumps; (b) improve reagent stability so that reagent replacement is not needed for at least two-week-long periods. We describe here a new instrument and reagent formulation that meet these goals and present intercomparison data with an independent technique. EXPERIMENTAL SECTION Reagents. Reagent A. OPA (P-1378, Sigma), 2.01 g, was dissolved in 250 mL of methanol and then made up to 1 L with 0.1 M pH 11 aqueous phosphate buffer (add ∼2 M NaOH to 26.81 g of Na2HPO4 (Aldrich) dissolved in 900 mL until pH is 11; make up to ∼1 L with deionized (DI) water). The solution is 15 mM in OPA. Reagent B. Sodium bisulfite-formaldehyde addition compound, sodium hydroxymethanesulfonate (Aldrich), 804 mg, was dissolved in water to make 1 L of a 6 mM solution. Both reagents A and B were stored at room temperature. Ammonium standards were made from a 0.1 M NH4Cl stock solution and fresh DI water. Exposure of reagents and standards to ambient air must be minimized as it is very easy to cause contamination from ambient NH3. All chemicals were reagent grade. One week of continuous operation (8-min cycles) requires 2.5 L of water and 1.25 L of each reagent Diffusion Scrubber and Sampling Manifold. The diffusion scrubber (DS)32,33 and the sampling arrangement is shown in Figure 1. A microporous polypropylene membrane tube (Accurel (32) Dasgupta, P. K. In Wilson and Wilson’s Comprehensive Analytical Chemistry Series; Pawliszyn, J. Ed.; Elsevier: New York, 2002; Vol. XXXVII, pp 97160. (33) Toda, K. Anal. Sci. 2004, 20, 19-27.
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
1891
PP Q3/2, pore size 0.2 µm, wall thickness 200 µm, inner diameter 600 µm, Membrana, Wuppertal, Germany) was the active element of the DS; this type of membrane has not previously been used for sampling NH3. The membrane tube is washed thoroughly with methanol prior to assembly to obtain high and reproducible collection efficiency (we recommend the interior of the jacket and the membrane be washed with methanol each time reagents are replaced, as routine preventive maintenance). One end of the methanol-washed membrane (initially 30 cm long) is forcibly inserted to a depth of 3-4 mm into an L-shaped stainless steel (S) tubing and secured with a minimum amount of fast-acting epoxy adhesive. These L-shaped tubes are constructed by first inserting a metal wire (0.7-mm o.d.) through the lumen of the hypodermic needle tube (1.47-mm o.d., 1.19-mm i.d.; Small Parts Inc., Miami Lakes, FL). Once bent to form the L, the wire is removed and the tubing cut to provide arm lengths of ∼20 × 5 mm. A 6-mm-i.d. glass tube (G) was first lined inside with a snugly fitting PTFE inner tube, 3.5-mm in i.d., and provided with polypropylene tees T (Ark-Plas, Flippin, AR) at each end. The free end of the membrane tube is inserted through the perpendicular arm of one tee and brought out through the other. After affixing the stainless tube L in T with hot-melt adhesive, the membrane is pulled taut, excess membrane is cut off, the free membrane end is inserted into the second L and epoxied to secure the joint, and then the L is affixed in the T with hot-melt adhesive. The active membrane length was 21 cm. The air exiting the DS may be saturated with moisture and can also contain particles. To avoid any damage from condensation of liquid water or particle deposition on the thermal mass flow sensor element, the DS exit air was passed sequentially through a water trap containing glass wool and a glass fiber filter (Whatman GF/B, 2.4 cm), Whatman) and then passed over a 40Ω, 10-W power resistor to which 5 VDC was applied, resulting in ∼0.6 W of heating; this assured that water will not condense in the flow sensor (Honeywell AWM3000). Calibration, Humidification, and Zero Gas. Efficient removal of ambient NH3 is necessary to prepare zero air and standards to reliably and reproducibly measure low levels of NH3. Purified house air is passed through a 2.5 × 100 cm column the bottom half of which is filled with acidic silica gel (prepared by mixing 50 mL of 6 N H2SO4 with ∼200 g of 20-mesh silica gel and drying overnight at 70 °C) while the top half is filled with standard silica gel. The air then passes through a second similar column filled with regular silica gel. This “NH3-free” air was continuously metered by a mass flow controller (MFC) at 0.6 standard liters per minute (SLPM) first into a 6 × 500 mm copper coil maintained in a 30 °C thermostated enclosure for thermal equilibration and then through a permeation chamber also in the same enclosure containing a wafer-type NH3 permeation device (VICI Metronics Inc., Poulsbo, WA), gravimetrically calibrated to be emitting 15.1 ng of NH3/min. The output NH3 stream was diluted by a second zero-air stream controlled by a second MFC (MFC2). The DS sampled from this diluted NH3 standard stream, and the balance was vented. To humidify the sample air, a third MFC (MFC 3) directs zero air through a microporous membrane tube (Accurel PP Q8/2, Membrana, 0.5 × 65 cm) immersed in water in a closed vessel. The relative humidity (RH) was varied by the airflow ratios 1892
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
through MFC3 and MFC2 and was measured in the mixed stream with a humidity probe (model RH390, ExTech Instruments, Waltham, MA). LED Photodiode-Based Fluorescence Detector. The configuration of the LED photodiode-based fluorescence cell is shown in the inset of Figure 1. A 1.6-mm bore hole is drilled straight through the center of one face of a 1.5-cm black Delrin cube and provided with 1/4-28 threaded flat seat termini. A fluorinated ethylene propylene copolymer (FEP) Teflon tube (1.07-mm i.d., 12-mm length, 300-µm wall, type 18 TW; Zeus Inc., Orangeburg, SC) passes through this hole and at each end is forcibly inserted into a opaque black PTFE tube whose ends were drilled out to accommodate the FEP tube and fixed in place at each end with a standard 1/4-28 threaded male nut and ferrule. As shown in the Figure 1 inset, from one face perpendicular to the long axis of the tubular flow passage, an InGaN LED (NSHU550E, Nichia USA, λpeak 375 nm, half bandwidth 12 nm), illuminates the FEP tubing through a 5.2-mm aperture and is affixed in place with a 1/4-28 threaded male nut. Perpendicular to the LED source, a photodiode-operational amplifier (OPT301, Texas Instruments), with a blue plastic filter OF on its face (type 856; Edmund Scientific., Gloucester, NJ, serving to filter stray LED light) measured the emitted fluorescence, through a 2.4 × 6.6 mm rectangular aperture. A silicon photodiode (PD; BPW 34, Siemens AG) was placed on the opposite face of the LED through a 1/4-28 threaded port and held there by a nut to measure the intensity of the light source. The overall circuitry for the detector that provides secondary gain and offset is provided in Supporting Information (Figure S1). Hybrid Flow Analyzer (HFA). All components of the HFA were housed in a “minitower” personal computer case. The fluidic system is shown in Figure 2. It consists of two syringe pumps (P/N 54022), one equipped with a three-way distribution valve (P/N 99884) and the second equipped with a six-way distribution valve (P/N 17619); Kloehn, Las Vegas, NV) and a solenoid valve (12 VDC, 30 PSI, NResearch Inc.). Each pump is equipped with a 5-mL capacity zero dead volume syringe (P/N 24691). Table 1 summarizes one analytical cycle. Initially P1 aspirates deionized water through port B from bottom water container W1 via mixedbed ion exchange column MB1 (Dowex MR3, 200 mesh, Sigma, in a 5 × 100 mm PTFE tube) thus purifying the water in-line. During this period, P2 alternately aspirates 50 µL of reagents A and B (ports C and D, respectively) 20 times, aspirating 1 mL of each reagent. This alternate aspiration of the reagents in thin layers in the wide-bore syringe already begins to cause significant mixing. Complete mixing is ensured in the next step when P2 addresses port B, which contains an open-ended coil MC (1.07 × 120 mm). P2 alternately dispenses and withdraws 100 µL into MC three times, mixing the two reagents completely in the process. (Although we have not conducted extensive experiments, it appears that this step, certainly three alternations, may actually not be necessary.) During this period, SV switches to NC position, P1 switches to port A, and water from container W2 is aspirated through mixed-bed ion exchange column MB2, through the lumen of the diffusion scrubber DS into the sample holding coil SC (0.66 × 1550 mm, 530 µL), a coil made in the Serpentine-II fashion34 to reduce dispersion. This action causes the DS (the internal volume of the DS is 68 µL) to be refilled with fresh water,
Figure 2. Hybrid flow analyzer for ammonia determination: W1, W2, water; FL, fluorescence detector; DS, diffusion scrubber; SV, solenoid valve (NC, NO, normally closed and open ports); SC, sample holding coil; MB1, MB2, mixed-bed resin columns; MC, open-ended mixing coil (acid silica vent trap at end not shown); AP, air pump; bottles A, B, reagents A, B. Table 1. One-Cycle Operation of HFA for Ammonia Determination step
time (min)
P1
1 2 3
0.00-0.02 0.02-0.04 0.04-1.20
4
1.20-1.70
5
1.70-8.00
aspirate 1480 µL W at 3 mL/min continue step 1 continue step 1 until 0.49 min when aspiration is complete, then remain idle aspirate 520 µL DS contents at 1.05 mL/min pump contents to heated coil and fluorescence detector at 0.3 mL/min
and the ammonia originally collected in the DS liquid moves to SC. It has been previously observed20 that, at typical values of atmospheric CO2 and NH3, water behaves as a perfect absorber. In this work also, acidic absorbers showed no advantage. The flow rate of P1 is reduced to minimize higher flow-induced dispersion. Finally, in step 5, the solenoid valve switches to NO, P1 and P2 both pump through respective ports A, in conjunction, both at low flow rates, with the NH3 sample now residing in SC, the latter behaving as a virtual sample loop. Note that the water originally drawn into P1 in steps 1-3 provides the baseline after the sample response is completed. Thus, in step 5, the system is completely analogous to a FIA system, the pump effluent flows through a 0.7 × 2000 mm PTFE knit reactor (tR 77 s) potted in a low-melting Bi-Sn alloy and maintained at 70 °C.35 The formation of 1-sulfonatoisoindole occurs in the reactor, and the resulting fluorescence is registered by the detector. The program cycles repeats indefinitely or for a preprogrammed number of cycles or until manually shut down. To allow free withdrawal of liquids from the liquid reservoirs, the liquid reservoirs must be provided with a vent hole. But intrusion of ambient ammonia can occur through these vent apertures, so small columns filled with acidic silica gel (34) Waiz, S.; Cedillo, B. M.; Jambunathan, S.; Hohnholt, S. G.; Dasgupta, P. K.; Wolcott, D. K. Anal. Chim. Acta 2001, 428, 163-171. (35) Li, J.; Dasgupta, P. K.; Genfa, Z.; Hutterli, M. A. Field Anal. Chem. Technol. 2001, 5, 2-11.
P2
SV
aspirate 50 µL A at 3 mL/min aspirate 50 µL B at 3 mL/min repeat steps 1 and 2, 19 times
no no no
dispense 100 µL into MC, then withdraw, 3 mL/min, three times pump contents to heated coil and fluorescence detector at 0.3 mL/min
nc no
(vide supra) were provided on each bottle vent. The control software for HFA was written in-house on the Softwire 3.1 platform (Graphical Programming for Visual Basic, SoftWIRE Technology, Middleboro, MA). RESULTS AND DISCUSSION Reagent Composition and Stability. The original recipe for carrying out the reaction of interest involved aqueous unbuffered OPA as one reagent and sulfite buffered at pH 11 with phosphate as the other.19 Because of the oxidative instability of S(IV), especially in alkaline solutions, the reagent was prepared daily. In a subsequent application20 to use this method for atmospheric NH3 measurement over a longer period, the sulfite reagent was modified by the addition of HCHO in a 2:1 molar ratio of S(IV)/ HCHO, with the belief that HCHO will stabilize S(IV) through the formation of the hydroxymethanesulfonate (HMSA).36 No real stabilization was observed. In subsequent work, the authors went back to a no-HCHO S(IV) reagent formulation and the recommendation of daily S(IV) reagent preparation was reinstituted.23 Daily reagent preparation is obviously not very practical for running a field instrument. Moreover, the previous effort was flawed not only in that a substoichiometric amount of HCHO was (36) Dasgupta, P. K.; DeCesare, K.; Ullrey, J. C. Anal. Chem. 1980, 52, 19121922.
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
1893
added, from published values37 of the HCHO-HSO3- equilibrium constant of ∼107 and the acid dissociation constant38 of HSO3- of ∼10-7, in a solution that contains 1.5 mM HCHO and 3 mM S(IV) at pH 11, one computes that only ∼35% of the S(IV) will be present as HMSA. In revisiting this problem, we first attempted stabilization with a sacrificial oxidant, ascorbic acid. This was only partially successful; the signal decreased gradually over a period of 1 week, by 10%. Next we attempted what is now the recommended strategy in this paper. This calls for incorporating the alkaline phosphate buffer in the OPA reagent while the S(IV) reagent is stabilized with essentially a stoichiometric amount of HCHO, or as recommended here, the sodium bisulfite-formaldehyde addition compound is directly used. The S(IV) is protected by the HCHO and, as is well known, virtually indefinitely stable in solution. HMSA itself is not a weak acid;39 a dilute solution of OHCH2SO3Na is essentially neutral. For a 6 mM neutral solution of OHCH2SO3Na, one computes from available equilibrium constants that 99.7+% of the compound is present as the adduct. The adduct, however, will not by itself participate in the OPA reaction. When the alkaline heavily buffered OPA reagent is added to the adduct solution, at pH 11, the equilibrium shifts, generating significant concentrations of free sulfite. This is also aided by the elevated temperature as the equilibrium constant for the adduct formation decreases by a factor of ∼50 from room temperature to 70 °C.40 Relative to the situation for a neutral 6 mM solution of OHCH2SO3Na, for a solution of 3 mM OHCH2SO3Na at pH 11 and 70 °C (formation constant 2.1 × 105 at 70 °C40), as obtained after mixing of the two reagents, 94+% of the S(IV) is liberated into the free form. Further, the decomposition kinetics of HMSA has been studied; the first-order rate constant at pH 5 and 25 °C is of the order of (3.5-10) × 10-6 s-1 and increases 1 order of magnitude with each unit rise in pH.41,42 At a pH of 11 at 25 °C, the half-time for decomposition will already be in milliseconds; the equilibrium level of free S(IV) will be generated even faster in the heated reactor. The concentrations of OHCH2SO3Na and the OPA concentrations stated in the Experimental Section, 6 and 15 mM, respectively, were chosen based on the optimization experiments shown in Figure 3. DS and Collection Efficiency. The geometry of the DS permits air flow without directional changes and thus minimizes particle deposition within the DS43 and protects sample integrity. The DS is maintained in a vertical configuration to avoid gravityinduced particle deposition. Air flows upward at 0.2 SLPM, and when water is drawn through the DS, it flows down from the top. At our test location (680 mmHg, 22 °C), 0.2 SLPM constitutes an actual volumetric flow rate of 4 cm3/s. The collection efficiency of the DS was measured by a serial scrubber method44 with two (37) Dong, S.; Dasgupta, P. K. Atmos. Environ. 1986, 20, 1635-1637. (38) Rhee, J.-S.; Dasgupta, P. K. J. Phys. Chem. 1985, 89, 1799-1804. (39) Dasgupta, P. K.; DeCesare, K.; Ullrey, J. C. Anal. Chem. 1980, 52, 19121922. (40) Deister, U.; Neeb, R.; Helas, G.; Warneck, P. J. Phys. Chem. 1986, 90, 32133217. (41) Blackadder, D. A.; Hinshelwood, C. J. Chem. Soc. 1958, 2720-2734. (42) Kok, G. L.; Gitlin, S. N.; Lazrus, A. L. J. Geophys. Res. 1986, 94, 28012804. (43) Genfa, Z.; Dasgupta, P. K.; Cheng, Y.-S. Atmos. Environ. 1991, 25A, 27172729. (44) Dasgupta, P. K.; Dong, S.; Hwang, H.; Yang, H. C.; Genfa, Z. Atmos. Environ. 1988, 22, 949-963.
1894 Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
Figure 3. Optimization of S(IV) and OPA concentrations. 6 mM S(IV) and 15 mM OPA were chosen based on these results.
Figure 4. Experimental vs theoretical ammonia collection efficiency of the diffusion scrubber. The circles indicate experimental values, the triangles indicate values computed according to ref 45 assuming the outer perimeter of the membrane as the sink, and the diamonds indicate computed values assuming the effective diameter of the membrane is the inner diameter times the fractional porosity.
identical DS units. Recently, a collection efficiency calculator has been developed for annular diffusion-based gas collectors.45 Although such calculations strictly apply to situations when the collection surface is an ideal sink, it is known that the calculations can provide reasonable results even when the sink surface is imperfect. In Figure 4, we plot the observed collection efficiencies versus the sampling rate (note that these are given in L/min; under our operating conditions, 1 SLPM ∼ 1.2 L/min). We also show collection efficiencies for the DS dimensions given and (45) Berg, J. M.; James, D.; Berg, C. F.; Dasgupta, P. K. personal communication, 2005.
DNH3 ) 0.226 cm2/s at 22 °C (estimated from the data given by Wintergerst46 at 20 °C), estimated according to ref 45. However, the present tubular porous membrane is a relatively thick wall capillary tube with a wall/inner radius ratio of 0.67. The extant work provides no guidance on whether the inner or the outer diameter of the membrane is the appropriate parameter to use as the inner bound of the annulus. As seen in Figure 4, using the outer membrane diameter clearly overpredicts the collection efficiency, and indeed, this is not a reasonable assumption: This is not a hydrophilic membrane, and the outer surface is not a sink. Even assuming the inner diameter as the inner annular sink may not be appropriate because (a) diffusion through the membrane wall cannot be as facile as in free air and (b) only 70% of the surface is porous. We also show the predictions assuming the inner annulus bounds are properly given for this smalldiameter membrane Effect of Humidity. Previous experience on porous membrane DS devices for the collection and analysis of ammonia showed that there is very little effect of humidity; at high humidities, the response decreases by a few percent, possibly due to the formation of a gas-phase aggregate with water and the resulting decrease in diffusion coefficient.20 The instrument was therefore shipped to a collaborating laboratory without thorough testing on the effect of humidity. When it was tested for humidity effects at this location, large humidity effects were observed, with signals being dramatically higher at higher RH levels. The instrument was brought back to our laboratory, and using a similar test setup, much to our consternation and puzzlement, we also observed such an effect. We state this here in some detail to prevent others from making the same mistakes that we and our collaborators made. Further experiments have shown that an apparent increase in the signal upon humidification is entirely an artifact that arises from (a) significant levels of NH3 that are always present in a typical laboratory, (b) the natural tendency of an experimenter to connect and disconnect the humidification source when testing the effect of humidity, (c) the inevitable presence of moisture at the exit of the humidification source, and (d) uptake of laboratory NH3 in the moisture at the exit of the said source, which subsequently evaporates upon connecting the humidification source. Such experiments must be conducted over a prolonged period without disconnecting the humidification source. Detailed experiments were conducted in the RH range of 25.2-72.9% at six different RH levels; the RSD of all responses together was 3.8% (N ) 18). For 20 ppbv NH3, the relative response changed from 7.78 (0.02 at 25.2% RH to 8.04 (0.07 at 72.9% RH. Long-Term Stability and Performance. Figure 5 shows a typical calibration series. The system and the reagent proved stable for long-term use. The system was calibrated before and after a three-week-long period of continuous use; neither the calibration slope nor the intercept was statistically different between the pre- and post-test calibrations at the 95% confidence level. Figure 6 (top) shows continuous data trace for a ∼10-day period at a gaseous ammonia level of 5.4 ppbv. Note that this performance is a composite result of not only the liquid-phase reagent stability/product formation but also photometric stability of a single-beam fluorescence detector and fluctuations in the trace gas generation source. Ninety-five percent of the results fall within (46) Wintergerst, E. Ann. Physik 1930, 4, 323-51.
Figure 5. Typical system output, gas-phase samples: 0.00, 1.78, 4.15, 5.95, 8.19, and 10.75 ppbv NH3, 8 min per peak. These data were obtained at a higher gain than that used to obtain eq 1, to clearly show the extent of the blank response.
Figure 6. Stability of response over 10 days to a 5 ppbv ammonia source. The inset shows intercomparison data with an ion chromatograph-based instrument (GPIC) over 2.5 days. The solid line is bestfit line with zero intercept.
the (3% (one relative standard deviation) of the mean: this will be deemed acceptable for most purposes. Some significant part of the observed deviation does arise from the stability of the gas standard generation system. The relative standard deviation for a comparable period of solution-phase (2.5 µM) NH4Cl injection Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
1895
was lower, at 1.7% (N ) 1544). These data do indicate excellent system robustness under continuous operation. The system also demonstrates excellent linearity over the typical range of ambient NH3; the calibration equation for 0-19.8 ppbv was
peak height (V) ) 0.613 ( 0.005 × NH3, ppbv + 0.160 ( 0.065,
r2 ) 0.9990,
n ) 15 (1)
Similarly for aqueous NH3/NH4+, the calibration equation for 0-20 µM was
peak height (V) ) 0.412 ( 0.0035 × NH3, µM - 0.228 ( 0.041,
r2 ) 0.9990,
n ) 17 (2)
Based on the observed noise level (variation of the blank), the S/N ) 3 LOD is estimated to be 135 pptv for NH3 and 15 nM for NH4+ (500-µL injection). True zero level blanks for either gaseous or aqueous NH3 are very difficult to attain in practice. In all our previous efforts, we have observed that the zero gas still contains some NH3. Were this blank level to be reduced, with accompanied reduction in blank variation, the attainable LOD can be improved further. Ammonia is a relatively sticky gas. Switching between blank to 5 ppbv and vice versa or between 5 and 18 ppbv and vice versa indicates that in all cases the full steady-state response requires two cycles, i.e., 16 min after switching. Changes in RH accompanying concentration changes may further affect response time. Intercomparison Experiment with a Gas-Particle Ion Chromatograph (GPIC). The GPIC is an instrument recently developed in this laboratory that measures gas and particle composition.47 It collects ammonia and other water-soluble gases in a wettable membrane-based parallel plate denuder in which the liquid remains stationary and is withdrawn every 40 min for (47) Ullah, S. M. R.; Takeuchi, M.; Dasgupta, P. K. Environ. Sci. Technol. In press. (48) Kenski, D.; Pushkarsky, M.; Webber, M. E.; Patel, C. K. N.; Dasgupta, P. K. Proc. NADP National Acid Deposition Program Ammonia Workshop; Washington, DC, October 2003.
1896
Analytical Chemistry, Vol. 78, No. 6, March 15, 2006
analysis. For ammonia determination, the denuder effluent bearing NH4+ is concentrated on a cation exchanger preconcentrator and eluted as NH3 using NaOH. The liberated NH3 passes through a membrane device with a pure water receptor on the other side, and quantitation is achieved by conductometry. The principles of determination are thus very different. A predecessor of this version of the GPIC has been compared in the field with the photoacoustic spectrometer and shown to generally agree within 0.2 ppb.48 Unfortunately, the GPIC provides a substantially poorer time resolution (40 min), and the results can only be compared after the results from the present instrument are averaged over five cycles. The two instruments were set up to sample ambient air outside our laboratory in Lubbock, TX. A common ∼1-m-long, 6.3-mmi.d. PTFE inlet was used that bifurcated at the instrument end, 1.5 and 0.2 SLPM, respectively, being sampled by the GPIC and the HFA. The results from a 2.5-day-period covering 9/29/0510/1/05 are shown in the inset of Figure 6. Note that the two instruments used independent calibration sources (gas-phase NH3 for HFA and solution-phase NH4Cl for GPIC), and very different integration periods and rather low levels of ammonia were measured. ACKNOWLEDGMENT N.A. was supported by the Royal Golden Jubilee Ph.D. Program and Postgraduate Education and Research Program in Chemistry (PERCH), Thailand, for his stay at Texas Tech University, which was extended by support from the U.S. Environmental Protection Agency through STAR grant RD 83107401-0. However, the manuscript has not been reviewed by the Agency and no endorsement should be inferred. He thanks Dr. Andrea Kirk for editorial help and S. M. Rahmat Ullah for the GPIC NH3 comparison data. N.A. expresses his deep indebtedness to D. Nacapricha, for being an ideal mentor.
Received for review November 27, 2005. AC051950B
October
31,
2005.
Accepted
