Measurements of Gaseous NH 3 and Particulate NH 4 + in the Atmosphere by Fluorescent Detection after Continuous Air – water Droplet Sampling

Phase partitioning of NHx (gaseous NH3 and particulate NH4) in the atmosphere is crucial for the lifetime of NHx during atmospheric transport. Reliable data for gaseous NH3 and NH4 in aerosols are necessary to understand phasepartitioning and atmospheric cycles of NHx. A semi-continuous microflow analytical system (MF system) was developed for measuring gaseous NH3 and particulate NH4 in the atmosphere. Two inlet lines were used to differentiate total amounts of NHx and particulate NH4 after gaseous NH3 were removed by phosphoric acid coated denuder from the sample air stream. Small water droplets were mixed with sample air and separated for liquid phase analysis in the MF system. The NH4 concentration in the liquid was measured using sensitive fluorescence detection after reaction with ophthalaldehyde and sulfite. Based on air sampling at a flow rate of 1 L/min with stripping water at a flow rate of 100 μL/min, the MF system can analyze down to 3 nmol/m of atmospheric NH3 concentration at 15 min intervals. Comparison with data based on the annular denuder method for gaseous NH3 and particulate NH4 concentrations indicated reasonable agreement with the MF system. Field tests of the MF system for one month showed good agreement with NH4 concentrations of fine particles collected daily on PTFE filters at the site. The MF system can monitor gaseous NH3 and particulate NH4 concentrations at 30 min intervals, thereby providing short-term phase partitioning data of NHx.


INTRODUCTION
Ammonia (NH 3 ), the dominant volatile base in the atmosphere, plays an important role in atmospheric chemistry: it neutralizes precipitation, cloud water, and acidic atmospheric aerosol particles such as sulfate, nitrate, and organic acids.Modification of the aerosol chemical composition by NH 3 neutralization engenders alteration of hygroscopicity and optical properties of aerosols (Seinfeld and Pandis, 2006).Furthermore, NH 3 can enhance new particle formation on a regional scale (Weber et al., 1998;McMurry et al., 2005) and on a laboratory scale (Ball et al., 1999), although NH 3 enhancement of ternary nucleation processes (H 2 SO 4 -H 2 O-NH 3 ) is controversial (Riipinen et al., 2007;Yu and Turco, 2008).
Atmospheric NH 3 is emitted primarily from livestock waste, volatilization from NH 4 + containing fertilizer, and other natural (birds, animals, ocean, etc.) and anthropogenic (fuel consumption, biomass burning, etc.) sources, all at the earth's surface (Bouwman et al., 1997;Galloway et al., 2004;Sutton et al., 2008).The NH 3 concentrations in the atmosphere near the ground range from < 0.01 mol/m 3 in remote regions (Ayers and Gras, 1980;Quinn et al., 1988;Norman and Leck, 2005;Johnson et al., 2008) to approximately 4 mol/m 3 or more (Theobald et al., 2006;Blackall et al., 2007;Cao et al., 2009;Hsieh and Chen, 2010), largely depending on the proximity to emission sources and deposition processes.However, typical concentrations of sub-micrometer particulate NH 4 + in the ambient air are < 0.01 mol/m 3 and approximately 0.3 mol/m 3 or more (Warneck, 1999;Sutton et al., 2008).Phase partitioning that occurs between gaseous NH 3 and particulate NH 4 + varies with environmental conditions (temperature and humidity) and acidity of the counteracting aerosols (Stelson et al., 1979;Allen et al., 1989;Mozurkewich, 1993).The complex behavior of gaseous NH 3 and particulate NH 4 + (hereinafter, NH x denotes the total amount of gaseous NH 3 and particulate NH 4 + ) hampers precise simulation of their temporal and spatial distributions in chemical transport models.
To understand the lifetime and behavior of NH x in the atmosphere, reliable measurements of gaseous NH 3 and particulate NH 4 + are needed without modification of their phase partitioning in the atmosphere.However, such measurements, especially those for low NH 3 concentration, are difficult because of (1) emissions from measuring personnel, (2) adhesive characteristics of NH 3 molecules, and (3) volatilization from labile particulate NH 4 + (existing as NH 4 NO 3 , NH 4 Cl, etc.) at the inlet and gas-particle separation parts.Contamination, including that by human emissions, of the measurement system from the surrounding atmosphere can be minimized to automate sample handling and analysis.Stickiness of NH 3 molecules to the wall engenders "inlet problems" such as slower response time and higher detection limits (Yokelson et al., 2003), and "calibration problems" of producing a gaseous standard at sub-ppbv concentration levels.Sampling artifacts from non-ideal separation between gaseous NH 3 and particulate NH 4 + present another "inlet problem".Several reports of comparisons for measuring atmospheric NH3 have been published (e.g., Wiebe et al., 1990;Williams et al., 1992;Mennen et al., 1996;Fehsenfeld et al., 2002;Schwab et al., 2007;Norman et al., 2009;von Bobrutzki et al., 2010).Among these measurements, diffusion denuder techniques (Ferm, 1979) have been widely used for sampling and stripping the gas phase NH 3 from the air stream.Most simple diffusion denuders consist of a glass tube coated inside with acidic reagents such as phosphoric and oxalic acids.As a sampling method, the diffusion denuder method presents advantages: it is a simple and low-cost method used for collecting atmospheric NH 3 .However, this sampling method has disadvantages: (1) it is time consuming; (2) it has low temporal resolution because of necessary handling and accumulation of sufficient amounts of NH 3 to analyze; and (3) it is labor intensive when high-frequency (e.g.hourly) measurements are needed.Moreover, manual handling, including sample preparation, wet-chemical analysis, and sample storage increase the likelihood of sample contamination.Nevertheless, diffusion denuder techniques have been used to remove the gas phase NH 3 from the air stream, leaving particulate NH 4 + to be analyzed (Bae et al., 2007;Huang et al., 2009;Thomas et al., 2009).
It is widely recognized that ammonium (NH 4 + ) is a major component of sub-micrometer particulates in terms of mass.Usually, collection of aerosol particles onto filters or impactor substrates has been conducted for off-line NH 4 + analyses.However, sampling artifacts of NH 4 + resulting from the adsorption and absorption of NH 3 onto deposited particles might occur on filter-based samples (Stelson et al., 1979;Harrison and Kitto, 1990;Kitto and Colbeck, 1999).Recently, several techniques have been developed for real time analysis of sub-micrometer aerosol particles.For instance, Bae et al. (2007) compared three semi-continuous NH 4 + measurement methods.In their analyses, agreement among the three techniques was almost satisfactory, but several issues, especially separation of gas and particles, have been pointed out related as possibly causing discrepancies.
Reliable data for gaseous NH 3 and NH 4 + in aerosols are necessary to elucidate atmospheric cycles of NH x .To obtain such data, gaseous NH 3 and particulate NH 4 + should be measured using the same standard materials and calibrations for both species without modifying their phase-partitioning in the atmosphere.As described herein, we propose a semi-continuous monitoring technique of gaseous NH 3 and particulate NH 4 + by switching analytical lines for NH x and particulate NH 4 + concentrations after removal of NH 3 using a simple diffusion denuder.

METHODS
To accomplish sensitive NH x analysis, we employed fluorescent detection of the o-phtalaldehyde (OPA)sulfite-NH 3 reaction product (Genfa and Dasgupta, 1989;Genfa et al., 1989) with a microflow (MF) system (Maruo et al., 2001;Kimoto et al., 2003a).As described in this paper, the NH 3 concentration is calculated as the difference between total (NH x : gaseous NH 3 plus particulate NH 4 + ) and NH 4 + concentrations in liquid samples by application with or without a phosphoric acid denuder at the inlet.Atmospheric NH x was dissolved in ultrapure water using a continuous air-water droplet sampler (Kimoto et al., 2003b).Herein, we describe the system improvements and performance checks used for measuring NH 3 and NH 4 + in the atmosphere.

Inlet and Air-Water Droplet Sampler
Fig. 1 depicts the continuous air-water droplet sampler connected to the MF analytical part.Two physically identical air-liquid lines were prepared for subtracting total NH x -particulate NH 4 + in the atmosphere.At the inlets, coarse particles were removed using Nuclepore filters (pore size: 5 m, 25 mm diameter) having aerodynamic diameter of ca. 2 m for 50% cut-off at an air flow rate of 1 L/min (John et al., 1983).Then 50-cm-long 3 mm i.d.glass tubes etched inside were connected after the inlet.One glass tube (Line 1 in Fig. 1) was coated inside with 3% H 3 PO 4 and 5% glycerine in 48% methanol and 44% pure water by weight.To develop laminar flow, 5 cm of the uncoated part was left at one end of the tube.The glass tube of the Line 2 was not coated with reagents, but was placed to ensure identical flow conditions.Assuming that the coated wall behaves as a perfect sink of gaseous NH 3 , the collection efficiency of this denuder is estimated as > 99% at the air flow rate of 1 L/min according to formulae presented by Dasgupta (1993).However, the actual collection of gaseous NH 3 is expected to be less efficient because of non-ideal conditions in use.That is, the collection efficiency depends on the relative humidity, NH 3 concentration, and other parameters.Because the denuder performance deteriorates after use over a long period, we changed the inlet denuder every week.These coated and uncoated glass tubes were connected by short silicon tubes to PTFE T-shape tubes having different diameters (1/8 and 1/16 inch) at the ends.At a flow rate of 100 L/min, ultrapure water drops were added at the Tshape tube into the sample air stream, and transferred to a Fig. 1.Schematic diagram of inlet system with acidic denuder (Line 1) and without acid denuder (Line 2).Aside from the coating, the acidic reagent (3% H 3 PO 4 and 5% glycerine in 48% methanol and 44% water by weight), lines 1 and 2 are physically identical.Ultra-pure water droplets were supplied using a syringe pump at 100 L/min.Nuclepore filter, pore size, 5 m, 25 mm diameter; MFM, mass flow meter; ALS, air-liquid separator made by glass; Sample air flow rate is 1 L/min.The dotted line represents the observation room wall.PTFE tube coil (1/16 inch i.d., 5 m length).Sample air of 1 L was washed using 100 L water every minute.Preliminary experiments showed that water is sufficiently effective as an absorber for measuring ambient levels of NH 3 (Genfa et al., 1989;Kimoto et al., 2003b).The length of the washing PTFE coil was necessarily greater than 3 m (Kimoto et al., 2003b).In our study, the air inlets, glass tubes, and the droplet mixers were placed outside within a white-colored weather shield to minimize heating by sunlight.At the ends of the washing PTFE coil, liquid phase droplets were separated by glass air-liquid separators in a temperature-controlled box.Sample air flow rates were controlled and monitored using mass flow meters.Liquid samples were delivered to the porous degassing tubes by a peristaltic pump in the MF system.The sample lines were switched by computer controlled three-way valves.The liquid sample was loaded into a 200 L sample loop on a six-port valve (Fig. 2) and injected to the MF system.

Microflow Analysis
The basic system of microflow analysis is almost identical to that reported by Maruo et al. (2001).Fig. 2 portrays a schematic diagram of the MF system.The R1 solution was composed of 30 mM of o-phthalaldehyde (biochemical grade; Wako Pure Chemical Industries Ltd.) in 70% methanol (fluorometric grade, Luminasol; Dojindo Laboratories, Japan).The R2 solution contained 10 mM of disodium sulfite (analytical grade; Wako Pure Chemical Industries Ltd.) and 50 mM of potassium dihydrogen phosphate (analytical grade, Wako Pure Chemical Industries Ltd.) with 5 mM trisodium citrate (analytical grade, Wako Pure Chemical Industries Ltd.) at pH 11.0 adjusted by adding 1-M sodium hydroxide.The R1, R2 and ultrapure carrier water were supplied to the reaction coil (RC) by precise syringe pumps at a flow rate of 10 L/min for the R1 and R2, and 100 L/min for the carrier water.The reaction coil was curled in a heating block maintained at 85 ± 1°C.A porous degassing tube (Gore-Tex) was connected between the reaction coil and the fluorescence detector that uses excitation at about 360 nm and emission greater than 420 nm.Using a six-port valve with a 200 L sample loop, a constant volume of the liquid sample was injected to the MF system.The volume of sample loop can be reduced for higher atmospheric concentrations.The time required for one sample analysis was 15 min in this study.Switching analytical lines 1 and 2 at every 15 min, a pair of data for NH x and particulate NH 4 + in the atmosphere was obtained for 30 min intervals.Three-port and six-port valves, syringe and peristaltic pumps, and other system components were all controlled using a built-in computer in the MF system which also records data of air flow rates, fluorescent detector voltages, temperatures in the equilibrator box, etc.All digital data can be monitored on a PC screen using a LAN connection with the MF system.

Calibration
For the liquid-phase standard, 1 g/kg NH 4 + standard solution (ion chromatography grade, Wako Pure Chemical Industries Ltd.) was used as a primary stock solution.Working standard solutions were prepared by diluting the stock solution down to 0.05 M immediately before measurements.Responses of the fluorescent detector of 0.05-1.09M are depicted in Fig. 3 + solution indicated standard deviation ( ) of peak areas as 0.173.Assuming constantindependence from concentration levels, the limit of quantification (Miller and Miller, 2010) of the analyte (x Q ) can be estimated as where b is the slope of the regression line.In this case, we obtained x Q = 0.03 M, corresponding to 3 nmol/m 3 of atmospheric concentration, assuming perfect collection of gaseous NH 3 into the liquid.Regarding reproducibility associated with the MF system, 10 consecutive replicate measurements of 1.13 M NH 4 + solution indicated standard deviation ( ) of concentration as 0.04.In other words, the relative standard deviation was 3.5%.
We also performed comparisons with the gas phase NH 3 standard.The NH 3 standard (48.58 ppmv in N 2 ; Takachiho Chemical Industrial Co.Ltd.) was diluted using a two-step dilution system: changing mass flow ratio and microorifices (Fig. 4).The first dilution was performed using two mass flow controllers (MFC1 and MFC2 in Fig. 4) for dilution from 1/92 to 1/4000 of the original gas concentration.In fact, the NH 3 gas adheres to the wall of the apparatus.For that reason, use of mass flow controller should be avoided for dilution at sub-ppbv concentration levels.Therefore, micro-orifices were used at the second stage to produce sub-ppbv standards as a fixed 1/200 flow ratio under constant temperature and back-pressure conditions.All parts after the first stage were produced using PTFE.
Fig. 5 portrays the relation between gas standards and the MF response calibrated using the liquid standards, as depicted in Fig. 3(b).The regression line slope was 0.86, which suggests that (1) collection efficiency of NH 3 in this system was 86% on average, (2) part of NH 3 gas was lost in the dilution system, or (3) differences might exist in between primary gas and liquid standards used.According to test results for changing the air flow speed from 0.5 to 1.5 L/min at a constant gas concentration (ca.30 nmol/m 3 ), the collection efficiency was equal from 0.5 to 1.2 L/min; it decreased slightly at 1.5 L/min (not shown; similar flow dependency was reported in Kimoto et al. (2003b)).In contrast, a strong linear relation was obtained for gas standards, which implies that the loss of diluted NH 3 is unlikely to be the cause of the difference.For our study, we primarily used liquid standards for calibration because of traceability.Gaseous NH 3 standard was secondarily used for checking the system performance.

Comparison with Annular Denuder Method at an Urban Site
The MF system developed in this study was compared with the annular denuder method designed for sampling NH 3 and NH 4 + in the atmosphere, similar to basic configurations used in Matsumoto and Okita (1998).The annular denuder system consists of a cyclone separator (2000-30EH; URG Corp.) cutting particles larger than 2.5 m in diameter, an annular denuder (28 mm OD × 242 mm length, -2000-30x242-3CSS; URG Corp.) coated with 2% oxalic acid in methanol and glycerol for collecting NH 3 , a PTFE filter (nominal pore size, 1.0 m; filter size, 47 mm diameter; Advantec Toyo Kaisha Ltd.) for collecting aerosol particles, and a backup oxalic acid impregnated filter (impregnated 300 L of 0.01M oxalic acid in a 16/84 glycerol/methanol solution by volume, 47 mm diameter) for collecting gaseous NH 3 evaporated from particles on the PTFE filter.We used data from the annular denuder as gaseous NH 3 and the sum of PTFE and the oxalic acid impregnated filter as particulate NH 4 + .The sample flow rate was 16.7 L/min for the annular denuder system.It was monitored using a mass flow meter (SEF-51; STEC Inc.).+ and NH x of the denuder system were slightly higher than that in our MF system.Differences in cut-off diameters between the cyclone (> 2.5 m) and the Nuclepore filter (> 2.0 m) at the inlets might engender the differences observed for the denuder and the MF systems.During the comparison of NH 4 + in aerosols, difficulty in producing identical inlet conditions was noted also in Bae et al. (2009).

Field Testing at a Background Site
The MF system was used for field work conducted from mid-March to mid-April, 2008.During the field campaign, standard solution (approximately 1 M) was measured automatically every day.Fig. 7 presents response variation of the standard solution from 22 March to 17 April.Response of the standard solution decreased gradually from 1.12 to 0.90 M during 4 weeks, possibly because of the degradation of reagents stored at room temperature and accumulation of stain within the detector and PTFE tubing in the MF system.Biological activity in the tubing might also have reduced the response of the standard solution, as discussed later in detail.According to various examinations of the MF system after the campaign, degradation of the OPA reagent showed the highest potential for decreasing response during the measurement.Using data presented in Fig. 7, atmospheric NH 3 and NH 4 + concentrations were corrected for the response variation.
The field campaign was conducted at Cape Hedo Atmosphere and Aerosol Monitoring Station (CHAAMS), Okinawa, Japan (26.87°N, 128.25°E, 60 m a.s.l., Takami et al., 2007;Takami et al., 2010).The station is located at the northern end of Okinawa Island, far away from populated areas of this island.Westerly winds prevail during winter to spring.Therefore, this station has been used to study the outflow of pollution from continental Asia.At this station, because numerous parameters of atmospheric aerosols and gases have been monitored (http://www.nies.go.jp/asia/hedomisaki/outline_e.html), it is suitable to test our system for a longer time period.In addition, stacked filter pack samples were collected daily using an automated sampler (GS-10; Tokyo Dylec Corp.) with a typical air flow rate of 23 L/min.The stacked filter pack consisted of an impactor, a Nuclepore filter (pore size: 8 m), and a PTFE backup filter for collecting particles having aerodynamic diameters of > 7, > 1.5, and < 1.5 m, respectively.We compared NH 4 + concentrations in fine particles of the stacked filter pack samples.
Fig. 8 shows the wind direction (green), wind speed (black), and the results (red lines) of the MF system with the NH 4 + concentration (blue line) of fine particles obtained from the stacked filter pack.Although large short-term variations were apparent in NH 4 + concentrations measured using the MF system, the MF data averaged for the duration of the filter pack samples agreed well with filter pack data (Fig. 9).Large variation was also found for NH 3 concentrations, but most large spikes in NH 3 concentrations were out of phase in peaks of NH 4 + concentrations.Comparison with local meteorological parameters reveals some high NH 3 peaks that occurred during periods of wind from directions from 90 (E) to 240 (SSW), as indicated by arrows at the top of Fig. 8.At these directions from the site, many farms had been cultivating sugar cane.March is the season of harvesting and planting new sugar cane near the site.Natural fertilizers such as poultry manure were used for planting new sugar cane.For that reason, NH 3 and amines might be evaporated from fertilized farmland.However, primary amines were not detected on chromatograms of cation analysis for filter packs during this period.In addition, as Kérousel and Aminot (1997) reported, interference from primary amines was < 0.5% for the reaction of NH + /NH x increased from nearly zero to > 0.9 during this period concomitantly with increased NH 4 + concentrations.This example demonstrates one advantage of the MF system.It is unrealistically labor intensive to use the denuder method manually for such a large variation of NH 4 + /NH x within a short (< 12 h) duration.For that reason, the MF system is more useful to observe short-term variation of NH 4 + /NH x .High NH 4 + concentrations are often found in the air masses transported from China (Takami et al., 2007(Takami et al., , 2011)).Under conditions of continental outflow with high NH 4 + concentrations from the Asian continent, NH 3 concentrations were low.Local wind directions during the continental outflow were mostly westward to northward.Therefore, the influence of NH 3 emanating from local farmland was expected to be small.

Potential Effects of Bacterial Activity in the MF System
After several field and laboratory tests of the MF system, responses of gaseous standards, especially for inlet line 1 (NH 3 denuded line coated with phosphoric acid and glycerol), decreased gradually.After washing the sampling line with methanol, the response reverted to its earlier level.Some aerobic microorganisms such as Nitrosomonas and Nitrococcus can convert ammonia to nitrate.Gradual growth and accumulation of these microorganisms might decrease NH 4 + concentrations in the sampling line, especially in the air-liquid separator.It is particularly interesting that the degree of NH 4 + suppression for the other line without the coating reagent was not so much and slower, which implies that combined activities might enhance nitrification from NH 4 + such as Nitrosomonas associated with Nitrobacter (Spieck and Bock, 2005).Considering this potential activity of nitrifying microorganisms at the inside walls of the PTFE tubes and the air-liquid separators, methanol was added as 2.5% by volume to the ultrapure sampling water.Addition of methanol to the water provides good system performance for long-term operations.

CONCLUSIONS
A semi-continuous microflow analytical system (MF system) was developed for measuring gaseous NH 3 and particulate NH 4 + in the atmosphere.The MF system can quantify concentrations as low as 3 nmol/m 3 of atmospheric NH 3 using an air sampling rate of 1 L/min.Comparison with annular denuder data for gaseous NH 3 and particulate NH 4 + concentrations agreed reasonably well with the MF data.One month of field testing of the MF system proved the long-term capability of semi-continuous monitoring.Data on gaseous NH 3 and particulate NH 4 + concentrations at 30 min interval provide short-term partitioning data to elucidate the NH x cycle in the atmosphere and to evaluate the performance of chemical transport models.
(a).Fig. 3(b) portrays a calibration plot of data depicted in Fig. 3(a).Strong linearity was obtained for the plot.The 43 consecutive replicate measurements of 0.05 M NH 4

Fig. 6
presents results of atmospheric measurements at Nagoya University on February 29, 2008.Although the comparison data were few, concentrations of gaseous NH 3

Fig. 6 .
Fig. 6.Comparison with the annular denuder method (stepwise blue lines) and the MF system (open red circles) observed in Nagoya, 29 February 2008.Meteorological data at Nagoya were obtained from the Japan Meteorological Agency.
4 + with OPA and sulfite.Consequently, high NH 3 peaks during wind directions from 90 (E) to 240 (SSW) were attributed to NH 3 emission from fertilized farmland, engendering changes in NH 4 + / NH x (particle fraction of NH x ), for example, the gradual change of NH 4 + /NH x from nearly 1 to almost 0 during March 22-23.In contrast, the drastic change of NH 4 + /NH x from nearly 0 to almost 1 during March 23-24 might be attributed from rapid and large variation of NH 4 + .Rain (several millimeters per hour at maximum) was observed during 12-19 h on March 23.During 17-19 h, the NH 4 + concentrations were almost zero, but they increased to 400 nmol/m 3 at 4 am on March 24.The minimum concentration of NH 3 was about 10 nmol/m 3 at around 16 h for this period.The NH 3 concentrations were almost constant at about 30 nmol/m 3 from 19 h on March 23 to 07 h of the next day.Consequently, values of NH 4

Fig. 8 .Fig. 9 .
Fig. 8. Results of field measurements at Cape Hedo, Okinawa, Japan obtained during March-April, 2008.Red lines show results for the MF system.No data were obtained during 31 March-2 April because of malfunctions.The blue line shows NH 4 + concentrations in fine (< 1.5 m) particles obtained by stacked filter packs.Wind data were obtained at the station.