Method for the Automated Measurement of Fine Particle Nitrate in

Average nitrate values from the automated system were compared to traditional ... the possible relationships between specific chemical constituents an...
0 downloads 0 Views 117KB Size
Download PDF
Environ. Sci. Technol. 2000, 34, 907-914

Method for the Automated Measurement of Fine Particle Nitrate in the Atmosphere MARK R. STOLZENBURG AND SUSANNE V. HERING* Aerosol Dynamics Inc., 2329 Fourth Street, Berkeley, California 94710

An integrated collection and vaporization cell has been developed to provide automated, 10-min resolution monitoring of fine particle nitrate in the atmosphere. Particles are collected by a humidified impaction process and analyzed in place by flash vaporization and chemiluminescent detection of the evolved nitrogen oxides. Particle collection efficiency was measured between 95% and 100% for particles above 0.1 µm. Evaporative losses for ammonium nitrate were 2 ( 4%. Average nitrate values from the automated system were compared to traditional denuder filter measurements in three cities, with correlation coefficients of greater than 0.97 and regression slopes ranging from 0.96 to 1.06. The detection limit is governed by the field blank, which for the Los Angeles area was 0.4 µg/m3. The system operated unattended for days at a time, with data recovery of 97%.

Introduction Airborne particles with diameters below 2.5 µm, called PM2.5, contribute to the atmospheric haze in our national parks (1); they play a role in radiative forcing and influence global climate (2); and they may have implications for human health (3). While the mass of PM2.5 is the parameter subject to regulation by the U.S. Environmental Protection Agency (4), data on particle chemical composition is needed to understand particle sources, atmospheric transformation, and the possible relationships between specific chemical constituents and health. Automated monitoring methods, as are available for specific gaseous pollutants such as ozone and carbon monoxide, have made it possible to obtain high time resolution, continuous ambient concentration data for these gases at reasonable cost. In contrast, particle constituents are usually measured using manual filter-based methods. This approach is costly, and often results are not known until some months after the sample was collected. As long as 20 years ago, particulate sulfate was measured continuously using flame photometric detection of sulfur compounds (5-8). Particulate organic and elemental carbon have been measured semi-continuously by the gradual, insitu heating of a particle sample collected by filtration (9) or impaction (10). “Black” carbon concentrations have been assayed by the decrease in optical transmission through a filter deposit (11, 12). More recently, inorganic constituents have been measured with on-line ion chromatography using wet-walled vapor denuders coupled with a steam-injection particle collector (13, 14) or a steam-jet collector (15). Other * Corresponding author phone: (510)649-9360; fax: (510)649-9260; e-mail: [email protected]. 10.1021/es990956d CCC: $19.00 Published on Web 01/29/2000

 2000 American Chemical Society

ion chromatography approaches include the wet frit particle collector (16) and liquid film impactor (17). Recent advances in Simon and Dasgupta’s method reduce interference from NO2 (18). In this paper, we present a new method for the automated measurement of fine particle nitrate based on a flash vaporization analysis. The method is designed to provide automated measurement of PM2.5 nitrate concentrations with a time resolution of 10 min. The instrument operates unattended, and results are available immediately. The goal is a rugged, affordable instrument suitable for routine monitoring. Presented below are laboratory and field tests, including evaluation of particle collection efficiency, analytical response, vaporization losses, and interferences. Also shown are example data from three cities, with comparison to parallel measurements with traditional, denuder-filter methods.

Experimental Methods Instrument Description. The automated nitrate monitor uses an integrated collection and vaporization cell (ICVC; 19) whereby particles are collected by a humidified impaction process and analyzed in place by flash vaporization. The approach is similar to the manual method that has been used for over 20 years to measure the size distribution of sulfate aerosols (20, 21). The difference is that the particle collection and analysis have been combined into a single cell, allowing the system to be automated. Particles are humidified prior to impaction to eliminate the rebound of particles from the collection surface without the use of grease (22, 23). Interference from vapors such as nitric acid is minimized by use of a denuder upstream of the humidifier. Analysis is done by flash vaporization into a nitrogen carrier gas with quantitation by a chemiluminscent NOx analyzer, similar to that described by Yamamoto and Kosaka (24). The flow system is configured such that there are no valves on the aerosol sampling line. The typical collection period is 8 min, and the time for analysis is about 90 s. The system automatically alternates between collection and analysis to provide on-line values for nitrate every 10 min. A schematic of the instrument is shown in Figure 1. The system consists of a pre-impactor, a multicell denuder, a humidifier, the collection and vaporization cell, a nitrogen source, a chemiluminescence nitric oxide detector, vaporization electronics, data system, and various sensors, flow meters, and valves. The pre-impactor excludes particles above 2.5 µm. The denuder is an activated carbon-impregnated ceramic honeycomb manufactured by Corning. The humidifier is a 300 mm long, 2.2 mm i.d. tube of Nafion surrounded by a jacket of water (Perma Pure MH110). The ICVC interior measures 32 mm in diameter by 19 mm high. It has a single 0.37 mm diameter orifice that operates under sonic flow conditions. A stainless steel strip mounted 1.9 mm below the orifice exit collects the particles. The strip is held in place by metal mounting posts that provide the electrical contact. The pre-impactor, denuder, humidifier, and collection-analysis cell are housed in an environmental enclosure that is ventilated with outdoor air by means of an exhaust blower. Temperature and relative humidity of the sampled air are measured between the humidifier and the cell. The flash vaporization electronics unit contains a 0.315 farad capacitor bank that is charged to 8.45 ( 0.05 V and then discharged through the collection strip under computer control using a silicon-controlled rectifier switch. The nitrogen oxide detector is a commercial chemiluminescence analyzer that measures NO based on its chemiluminescent VOL. 34, NO. 5, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

907

FIGURE 2. Schematic of aerosol testing system for the automated nitrate monitor. CNC, condensation nucleus counter; LAS-X, laser optical particle counter. LP-OPC is a second laser optical particle counter operated at low pressure below the cell. Nitrogen lines for analysis are not shown.

FIGURE 1. Schematic of the ADI automated nitrate monitor. reaction with ozone (model 42C, Thermo Environmental Instruments). The analyzer is used with a molybdenum converter for reducing higher nitrogen oxides to nitric oxide. The instrument alternates between sampling and analysis modes. During sampling, the valve between the cell and the pump is open, while the valves between the cell and the NOx analyzer and between the cell and the nitrogen source are closed. Ambient, particle-laden air is drawn through the preimpactor, denuder, and humidifier and into the ICVC at a flow rate of 1 L/min. The humidifier raises the relative humidity to a value between 70% and 99%, depending upon temperature and inlet humidity. The pressure in the cell is regulated by a downstream valve to one half of atmospheric to provide just-critical flow. The humidified particles are collected by impaction onto a small area at the center of the metal strip mounted inside the cell. After collection, the instrument switches to the analysis mode. The valve between the cell and the pump switches to redirect the sample flow to bypass the cell, and the valves to the nitrogen source and to the NOx analyzer are opened. Nitrogen enters the cell through the pumping port at a flow rate slightly less than the 0.6 L/min required by the gas analyzer. The balance enters through the sample orifice above which the nitrogen is supplied in slight excess. After a preset flushing time (typically 20 s), the particle deposit is vaporized in place by rapid resistive heating of the metal strip, and the evolved vapors are transported by the nitrogen carrier gas to the NOx analyzer. The gases pass through the molybdenum converter to reduce NO2 and other nitrogen oxides to NO prior to detection. The resulting peak in NOx concentration is integrated to yield the mass of nitrate collected. To ensure complete removal, the strip is heated a second time. Peak integration time is 20 s; the entire analysis takes 90 s. After analysis, the system returns to sampling mode, and the cycle is reinitiated. Calibration with Aqueous Standards. The analysis step is routinely calibrated using standards applied directly to the collection substrate. The standards are aqueous solutions of sodium nitrate, ammonium nitrate, or ammonium sulfate. The cell is opened, and a sub-microliter drop of the standard is applied directly to the center of the collection strip using a graduated, 1-µL syringe. The droplet is allowed to air-dry, 908

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 5, 2000

as determined by visual inspection (usually less than 1 min); the cell is closed; and the sample is analyzed using exactly the same purging and flashing conditions as in the actual operation. The response is the baseline-corrected integral, in ppb s, of the resulting peak in NOx concentration as measured by the analyzer. The response over a range of concentrations is obtained by varying the amount of standard applied. Laboratory Aerosol Tests. Ammonium nitrate particles were generated by nebulization, and a monomobility size fraction was selected from the nebulized aerosol using a highflow differential mobility analyzer (HF-DMA, 25), as shown in Figure 2. The particle concentration upstream of the ICVC was monitored with a condensation nucleus counter (CNC, model 3760, TSI Inc.) and an optical counter (LAS-X, Particle Measuring Systems). The particle concentrations downstream of the cell were monitored with a second optical particle counter operated at low pressure (LP-OPC). The LAS-X counter allowed us to correct for doublets, i.e., larger doubly charged particles in the test aerosol, while the CNC provided redundancy for evaluating counting efficiency and for evaluating the response for particles below the optical counter detection limit. Particle collection efficiency was measured from the ratio between the LP-OPC counts during collection to that when sample flowed through the bypass line, as during sample analysis or standby modes. The bypass line has a 1 L/min critical orifice so that the absolute pressure in the optical counter is the same for both measurements. Care was taken to minimize the excess nitrogen flow introduced above the orifice during analysis so that the dilution of the test aerosol through the bypass was less than 5%. Particle collection efficiency was measured as a function of particle size for ammonium nitrate and ammonium sulfate test aerosols. Particle penetration through the humidifier was measured with monodisperse particles from the HF-DMA. Concentrations above and below the humidifier were counted using the condensation nucleus counter. Experiments were done with monodisperse fractions of ambient laboratory air particles and with monodisperse nebulized ammonium nitrate particles. The overall instrument response to monodisperse ammonium nitrate particles was evaluated by comparison of the measured nitrate from the flash vaporization to that calculated from the HF-DMA size, upstream particle counting instruments, and sample duration. The nitrate from flash vaporization was calculated on the basis of the aqueous ammonium nitrate calibration standards. The sampled particle volume was calculated using the total number concentration given by the upstream CNC and the particle size obtained from the flow rate and applied voltage in the HF-DMA. This particle volume is corrected to account for

the presence of larger, doubly charged particles based on the size distributions from the LAS-X. Vaporization losses for ammonium nitrate were evaluated by exposing a deposited sample to particle-free air. The particle-free air exposure was obtained by stopping the sampling, adjusting the DMA voltage to zero, and then restarting the sampling once the particle counts dropped to zero. The expected response was calculated from the LAS-X and CNC counts during particle collection. The measured response from the flash vaporization analysis was compared to that expected as a function of the exposure time to particlefree air. Laboratory Interference Tests. Possible interference from other compounds in the flash vaporization analysis of the nitrate was tested by measuring the response to aqueous standards of ammonium sulfate and oxalic acid. This was done following the procedure for the calibration with standards, described above. Standards of ammonium sulfate were run routinely to test that the ammonium ion was not detected as nitrate. Interference from nitric acid was tested in the laboratory by using a permeation source. Nitrogen gas was passed through a particle filter and then through a glass cell containing the permeation source (VCI Metronics). The nitric acid concentration was calculated from the temperature and permeation tube emission rate and checked using the response of a chemiluminescent analyzer in NOx mode. The particle-free nitric acid source was sampled by using the automated nitrate system, with and without a denuder in place. Field Performance Tests. Particle collection efficiency was measured in the field using the LP-OPC to detect particles penetrating through the cell during sample collection. As with the laboratory experiments, the efficiency was calculated from the ratio between particle counts during collection to that measured during analysis mode, when the LP-OPC sampled ambient air immediately downstream of the humidifier. Unlike the laboratory tests, these measurements were not made as a function of particle size. Evaporative losses were evaluated under field conditions by alternating between 4- and 8-min sample collection during a period of relatively high nitrate concentrations. The measured result for longer sample duration was compared to the average of that for shorter-term samples collected just prior to and just after the test sample. Interferences from gaseous species were checked in the field by placing a Teflon filter in a Teflon filter housing at the inlet of the sampling line. Sampling continued as during normal operation, and the indicated nitrate values were compared to those obtained by passing nitrogen gas or clean laboratory air through the cell. Field Deployment. The first version of the instrument was operated in the Northern Front Range Air Quality Study (NFRAQS) over two 10-day periods in January and February 1997 in Denver. The instrument was modified and then operated in the 1997 Southern California Oxidant Study (SCOS’97) and the 1999 Bakersfield Study. For SCOS’97, the system was operated continuously from August 16 to September 3 in Riverside, CA, and from September 4 to September 30 at Mira Loma, CA (10 mi west of Riverside). Both locations are downwind of Los Angeles. The Mira Loma site lies immediately to the east of cattle feed lots. For the Bakersfield study, measurements were made from January 7 to January 31, 1999, at the California Air Resources Board monitoring station in Bakersfield, a city located in California’s central valley. For NFRAQS and Riverside, a KOH-coated glass multitube denuder was used. For Mira Loma and Bakersfield, an activated carbon-impregnated ceramic honeycomb denuder was employed. All studies used the same sampling and analysis cell.

TABLE 1. Particle Penetration through Humidifier particle type

diameter (µm)

penetration (%)

ambient ambient ambient

0.10 0.20 0.40

98 ( 1 97 ( 1 102 ( 6

NH3NO3 NH3NO3 NH3NO3

0.40 0.60 0.80

98 ( 1 97 ( 1 98 ( 1

TABLE 2. Particle Collection Efficiency particle diameter (µm)

date

Denver.

0.1 0.2 0.3 0.4 0.6 0.8

Ammonium Sulfate Particles 99.7 ( 0.2 99.8 ( 0.2 99.7 ( 0.2 99.7 ( 0.4 98 ( 2 95 ( 6

0.2 0.3 0.5 0.8

Ammonium Nitrate Particles 98.6 ( 0.2 99.8 ( 0.2 99 ( 1 97 ( 2 Ambient Aerosols ambient collection efficiency, nitrate (µg/m3) 0.2-1 µm (%)

1/17a 8/22b 8/23b 9/30c a

collection efficiency (%)

6 3 17 20 b

97 ( 2 89 ( 5 98 99 ( 0.4

Riverside. c Mira Loma.

Field calibrations were done with aqueous standards, exactly as in the laboratory as described above. Field calibrations were done in duplicate at each of three levels of ammonium nitrate, three levels of sodium nitrate, and one level of ammonium sulfate. Standards are routinely run in duplicate for 46, 86, and 173 ng of nitrate for the ammonium nitrate standard, in duplicate for the equivalent levels of sodium nitrate, and at one level (65 ng of sulfate) of ammonium sulfate. Additionally, for the NFRAQS and SCOS’97, the particle collection efficiency for ambient particles was measured using the LP-OPC, as described above. The data from each field program were reduced using a least-squares fit to the ammonium nitrate standards of the form y ) axm, where y is the blank-corrected response for standards and x is the mass of applied nitrate. This form accounts for nonlinearity introduced by the slightly lower response at higher concentrations. The data were also corrected for the field blank, determined by sampling filtered ambient air using a Teflon filter mounted inside a Teflon filter holder at the sample inlet. Uncertainties were calculated from one standard deviation of the filtered air blank.

Results and Discussion The individual steps in the operation of the automated nitrate monitor are the transmission of particles through the humidifier, their collection within the ICVC, and their subsequent analysis for nitrate. These were tested independently. Interferences and evaporative losses were specifically examined. The complete system was tested with laboratory generated aerosol and was compared in the field against more traditional, time-integrated measurements. Particle Transmission through the Humidifier. Particle transmission through the Nafion humidifier, located upVOL. 34, NO. 5, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

909

TABLE 3. Summary of Vapor Interference Measurements

a

sample

denuder

exposure time

equiv NO3- concna (µg/m3)

nitrogen gas stream filtered lab air HNO3 in N2, 150-320 ppb HNO3 in N2, 150-320 ppb filtered ambient air Riverside, Aug 17-30, 1997 Mira Loma, Sep 4-30, 1997 Bakersfield, Jan 4-30, 1999

none none none KOH-coated glass

700 s 300-2200 s 600-700 s 200-700 s

0.0 ( 0.0 0.4 ( 0.2 1.4 ( 1.0 0.3 ( 0.1

KOH-coated glass activated carbon activated carbon

480 s 480 s 480 s

1.3 ( 0.8 0.7 ( 0.4 0.4 ( 0.3

Calculated for 8-min sample collection period.

FIGURE 3. Response of flash vaporization analysis to aqueous standards of sodium nitrate, ammonium nitrate, and ammonium sulfate, plotted as a function of the mass of nitrate (or sulfate) applied. stream of the ICVC, was measured in the laboratory using monodisperse fractions of ammonium nitrate and ambient laboratory aerosol. Results are shown in Table 1. Indicated precisions are derived from replicate measurements. The particle losses through the humidifier are less than 2%. Particle Collection Efficiency. The collection efficiency in the ICVC was measured in the laboratory for ammonium sulfate and ammonium nitrate particles of known size. Results are shown in Table 2. Uncertainties are the 2σ deviation from replicate measurements. Particle collection efficiency for humidified ammonium sulfate and ammonium nitrate are above 95% for particle diameters ranging from 0.1 to 0.8 µm. Measurements were not done below 0.1 µm because smaller particles are not detected by the LP-OPC used to measure particle penetration. The errors are greater for the larger particles due to lower particle concentrations. Particle collection efficiencies were measured for ambient particles by comparing the counts at the LP-OPC during sample collection to those during analysis when the LP-OPC was sampling ambient air immediately downstream of the humidifier. Results from measurements in three different cities are given in Table 2. Measured collection efficiency for ambient particles ranged from 89% to 99%. Somewhat lower collection efficiency values were observed during a period of lower nitrate concentration. This may be due to the presence of a higher percentage of non-hygroscopic particles when nitrate was low. The results shown in Table 2 are for the final system configuration that employs a jet-to-plate spacing of 1.9 mm (approximately 5 jet diameters). Initial tests with ammonium sulfate at jet-to-plate spacings of approximately 1 jet diameter 910

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 5, 2000

showed many particles with diameters smaller than that of the test particles reaching the LP-OPC detector, indicating that these wet particles tended to “splatter” upon impact at these small spacings. In the system design, the jet-to-plate spacing was selected to maximize the particle collection efficiency. Response to Aqueous Nitrate Standards. The analysis step of the nitrate monitor was calibrated using liquid standards applied directly to the collection substrate. Calibrations are routinely done at three concentrations and with both sodium and ammonium nitrate. An example calibration response is shown in Figure 3. The response is approximately linear, with a slope of 0.037 ppm s ng-1 nitrate for the ammonium salt. As was generally observed, the response to sodium nitrate is slightly (5-10%) higher, especially at the highest standard levels. The measured response to known nitrate standards can be compared to that expected on the basis of the NOx calibration of the chemiluminescent analyzer and its sample flow rate. The integral of the NOx concentration, C, is related to the mass of the deposited nitrate, S, by the mass balance equation:

S ) Qc

MW ν

∫C dt

(1)

where S is the mass of particle nitrate in nanograms, Qc is the carrier gas flow rate into the gas-phase analyzer in liters per second, MW is the molecular weight of nitrate in grams per mole, v is the molar volume of carrier gas in liters, C is the gas-phase concentration of NOx in parts per billion, and t is the integration time in seconds. This equation assumes that the entire particle nitrate is converted to NOx upon vaporization and efficiently transported to and quantified by the analyzer. For an analyzer gas flow rate of 0.6 L/min, the theoretical response for nitrate corresponds to 0.04 ppm s ng-1 as compared to a measured value of 0.037 ppm s ng-1 in Figure 3. Response to Non-nitrate Compounds. In the design of the system, the flash vaporization conditions were selected to minimize the response to ammonium ion, as indicated by the analysis of ammonium sulfate standards, while providing approximately equal response to sodium nitrate and ammonium nitrate. The possible interference from ammonium ion was tested repeatedly using aqueous ammonium sulfate standards. This was done routinely as a part of each calibration procedure, as shown in Figure 3. For a mass of applied ammonium sulfate equivalent to 65 ng of sulfate, the response was less than 0.1 ppm s or less than 4% of the response to nitrate. Similarly, tests with an aqueous standard of oxalic acid showed less than 0.1 ppm s for 300 ng of oxalic acid. Interferences from Vapor Compounds. Vapor interferences were tested in the laboratory with nitric acid vapor and in both laboratory and field using filtered air. Results are summarized in Table 3. All tests were done with the humidifier

TABLE 4. Summary of Measured Evaporative Losses Laboratory Ammonium Nitrate Particles particle diameter (date of experiment)

sample collection time (s)

time exposed to particle-free air prior to analysis (s)

0.6 µm (12/31/96) 0.6 µm (12/31/96) 0.4 µm (8/10/97) 0.2 µm (8/12/97) 0.4 µm (12/4/97) 0.6 µm (12/4/97) 0.5 µm (11/8/98) 0.4 µm (2/26/99) 0.4 µm (2/26/99) 0.4 µm (2/26/99)

130 130 300 300 480 280 180 60 60 60

481 479 300 300 484 280 180 60 240 480

fraction recovered 1.00 1.02 0.99 0.91 0.98 0.93 1.01 0.97 1.00 1.00 0.98 ( 0.04

average Ambient Particles date and time 9/30/97 11:08 9/30/97 11:14 9/30/97 11:24 9/30/97 11:30 9/30/97 11:43 9/30/97 11:49 9/30/97 11:59 average

FIGURE 4. Response of nitrate monitor to laboratory ammonium nitrate particles: (a) primary sizes of 0.55, 0.61, and 0.87 µm; (b) primary sizes of 0.41 and 0.32 µm. in place. Some were done without a denuder, as indicated. For exposure of the sampling substrate to a stream of nitrogen, no measurable blank is observed. Sampling 150-320 ppb of nitric acid in a nitrogen carrier gas without a denuder gave 11 ( 8 ng of nitrate (equivalent to 1.4 µg/m3 for 8-min sampling). Use of a KOH-coated glass denuder ahead of the humidifier reduced this signal to 2.4 ( 0.8 ng (equivalent to 0.3 µg/m3). For measurements of filtered ambient air using the same KOH-coated denuder in Riverside, CA, during August 1997, the blank values were high, averaging 10 ( 6 ng or 1.3 µg/m3. The blank values tended to be higher during periods of elevated NOx concentrations (R 2 ) 0.65). One infers this interference is from some vapor compound other than nitric acid. For subsequent measurements in Mira Loma and Bakersfield, an activated carbon-impregnated ceramic honeycomb was used in place of the KOH-coated denuder. Field blanks for this denuder were lower by a factor of 2 on average than for measurements with the KOH denuder. The ambient levels of NOx were 30% higher, and particulate nitrate values were a factor of 3 higher when filtered air samples were collected with the carbon denuder, as compared to the periods of filtered air with the KOH denuders. While a direct comparison between the two types of denuders was not made, the indication is that the activated carbon denuder is more effective at removing interfering vapor compounds and

collection time (s)

measured nitrate (µg/m3)

240 480 240 480 240 480 240

35.49 37.46 39.34 44.48 44.32 44.76 47.22

ratio of long to short sample 1.00 1.06 0.98 1.01 ( 0.04

reducing blank values. The variability in this field blank, equivalent to 0.4 µg/m3, is the factor that governs the method detection limit. Evaluation of Vaporization Losses. For ammonium nitrate, sampling loss due to evaporation is always a concern. For filter sampling, losses as high as 80% have been reported (26). Much smaller losses have been observed for nitrate measurements made with the Berner impactor (27, 28). For the automated nitrate monitor, the short sampling times (8-12 min) and the high relative humidity during collection will tend to minimize losses. On the other hand, the use of denuders upstream of the collection point will tend to increase losses. This issue required experimental investigation. In the laboratory, evaporative losses were evaluated by collecting monodisperse ammonium nitrate with parallel measurements by the optical and condensation nucleus counters. Some samples were analyzed immediately at the end of the collection period to provide a value for the expected nitrate response. Other samples were collected for the same sampling time but then exposed to particle-free air prior to analysis. Recovery is calculated as the ratio of the nitrate measured after exposure to particle-free air to that measured immediately after collection, normalized to the particle number concentrations measured during the collection periods. Correction was also made for the blank value corresponding to the sampling of particle-free air. Results are shown in the upper portion of Table 4. Particle collection periods were varied from 1 to 8 min. Exposure time is the duration of the subsequent sampling of particlefree air, which was a factor of 1, 4, or 8 times longer than the initial sample collection period. These experiments mimicked the situation in which all of the nitrate was collected at the beginning of a collection period and then was exposed to the air stream throughout the remaining period. Measured recovery ranged from 91% to 102% independent of exposure time, with a mean value from 10 experiments of 98 ( 4%. In the field, evaporative losses were evaluated through variation of the sample duration. The lower portion of Table VOL. 34, NO. 5, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

911

FIGURE 5. Daily averaged nitrate measured by the automated monitor as compared to 24-h nitrate from Teflon filters (T1 and T2) and from the sum of Teflon and nylon filters from the FRM sampler at Mira Loma and Riverside, CA: (a) time series, (b) scatterplot. (FRM data courtesy of South Coast Air Quality Management District.)

FIGURE 6. Daily averaged nitrate concentration measured by the automated nitrate monitor and the Harvard University HEADS system at Riverside (HEADS data courtesy of P. Koutrakis, ref 27).

FIGURE 7. Daily averaged nitrate concentration measured by the automated nitrate monitor and the Met One SASS speciation monitor at Bakersfield (SASS data courtesy of T. Merrifield, ref 28).

4 lists nitrate concentrations indicated for alternating 4- and 8-min collection periods during a time of high particle nitrate concentrations in Mira Loma. The longer duration samples are compared to the average of the concentrations indicated from the short-duration samples immediately preceding and immediately following the longer sample. The mean ratio of long duration to short duration samples was 1.01 ( 0.04. If nitrate were volatilized during collection, then the samples of longer duration would yield lower values. The consistency of these data indicate that particle nitrate was not lost during sampling. Response to Ammonium Nitrate Test Aerosols. The automated nitrate system was tested in the laboratory with monomobility nitrate aerosols generated using the HF-DMA. The nitrate concentration measured by flash vaporization is compared to the volume concentration for the test aerosol indicated from the measurement of number concentration during sample collection and from the particle size indicated by the flow and voltage settings of the mobility analyzer. Calibration for the flash vaporization is derived from the liquid standards, as described above. Calculation of the total aerosol volume took into account the presence of larger, doubly charged particles with the same mobility diameter. The number concentration of the doubly charged particles was obtained from the LAS-X optical counter. The doublets accounted for 3%, 20%, 20%, 50%, and 44% of the aerosol volume at 0.87, 0.61, 0.55, 0.41, and 0.32 µm, respectively. Results are presented in Figure 4a for primary particle sizes of 0.55, 0.61, and 0.87 µm, and in Figure 4b for 0.32 and

0.41 µm. The nitrate measured by flash vaporization is shown as a function of the sampled aerosol volume derived from the particle number concentrations, mobility diameters, and sample duration. Larger particle deposits were obtained by increasing the sampling times at approximately the same particle concentration. The correlation between the two measurements is excellent, with R 2 > 0.95. A small change in the slope is observed for measurements at different particle sizes. This can be accounted for by systematic differences between the two experiments, such as in the magnitude of the correction for doubly charged particles in the test aerosol. Theoretically, the slope of the regression line is the product of the fraction of the particle mass that is nitrate and the particle density, in units of g/cm3. For perfectly dry ammonium nitrate particles, the expected slope is (62/80) × 1.7 ) 1.3 g/cm3. Our measured values are consistently lower, in the range 0.83-0.97 g/cm3. This lower response could be due to residual water associated with the particles or impurities in the particles. The experiments with laboratory ammonium nitrate show consistency between the measured nitrate and the calculated volume of sampled aerosol. The nitrate response increases systematically with the mass of nitrate sampled. Measurements at different particle sizes and concentrations yield similar results. Field Comparison with Filter-Based Measurements. On days selected during the SCOS’97 study, the South Coast Air Quality Management District made collocated measurements with two Federal Reference Method PM2.5 (FRM) samplers.

912

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 5, 2000

TABLE 5. Comparison of Automated Nitrate (y) against Filter-Based Methodsa filter sampler (x) FRM with tandem Teflon-nylon filters HEADS with denuded Teflon and carbonate filtersb SASS with denuded nylon filterc SASS with Teflon filterc c

site and dates

N

regression eq

R2

Riverside and Mira Loma, Aug-Sep 1997 Riverside, Aug 1997 Bakersfield, Jan 1999 Bakersfield, Jan 1999

9

y ) (0.96 ( 0.02)x - (1.2 ( 0.3)

0.996

y ) (1.06 ( 0.05)x y ) (1.02 ( 0.04)x y ) (1.01 ( 0.03)x

0.94 0.95 0.97

12 21 21

a N, number of data points; R 2, square of correlation coefficient; y, automated nitrate; x, filter sampler. SASS data from Dutcher et al. (30).

One of these was operated with a tandem Teflon-nylon filter pair, the other had a single Teflon filter. Samples were collected for 24 h commencing at midnight PST. No explicit denuder was used, although the cleaned aluminum surfaces in the inlet are expected to provide a fairly effective denuding of nitric acid vapor (27). The nylon filter in this sampler will collect particulate nitrate vaporized during sampling plus any vapor nitrate species not removed by the inlet. As such, the sum of nitrate from the Teflon and nylon filters in the FRM is expected to provide an upper limit for PM2.5 nitrate. Comparison to the Teflon-nylon filter sum from the FRM sampler is displayed in Figure 5. This data set includes 4 days of measurements in Riverside in August and 5 days at Mira Loma in September. Figure 5a also shows nitrate concentrations derived from both Teflon filters. The automated nitrate values follow the values from the FRM nylonTeflon filter sum, though offset slightly lower. The leastsquares fit of the automated nitrate data against the Teflonnylon filter sum yields a slope of 0.96 ( 0.02, an intercept of -1.2 ( 0.3, and a correlation coefficient of R 2 ) 0.996. On many days they are higher than for the Teflon filter alone, suggestive of evaporative losses from the Teflon filter. At Riverside, the nitrate monitor was collocated with a Harvard EPA annular denuder sampler (HEADS) operated by Harvard University (29). The HEADS used a carbonatecoated glass denuder to remove nitric acid, followed by a Teflon filter and two sodium carbonate filters to collect particle nitrate and to correct for positive artifacts from nitrogen dioxide. Samples were collected for 23.5 h commencing at 10:30 am PDT. Data from both the HEADS and the nitrate monitor were obtained on 12 of the 15 sampling days, as shown in Figure 6. The values from the automated nitrate system tended to be slightly lower than from the HEADS at the beginning and slightly higher than the HEADS during the latter half of the study period. Regression of the automated nitrate values against the HEADS gives a correlation coefficient of R 2 ) 0.94 and a slope of 1.06 ( 0.05. In Bakersfield, daily 24-h nitrate measurements were made with a SASS speciation sampler (Met One Instruments). This multilegged sampler employed a MgO-denuded nylon filter, and a parallel, undenuded Teflon filter. The flow rate through each leg of the sampler was approximately 7 L/min. Data for the 3-week study are shown in the time series of Figure 7. On most days the data from all three measurements (Teflon filter, denuded nylon, and automated nitrate) follow each other closely. Regression of the automated nitrate values against those from the denuded nylon filter yields a slope of 1.02 ( 0.04 and a correlation coefficient of R 2 ) 0.95. Regression statistics for all three field comparisons are summarized in Table 5. Diurnal Nitrate Profiles. Example data from the summertime measurements at Riverside and Mira Loma are shown in Figure 8. Both Riverside and Mira Loma showed considerable time variation in the nitrate concentrations throughout the day. Often a factor of 10 difference between the daily minimum and maximum 10-min averaged concentrations was observed. Both of these sites tended to show a maximum concentration at midday.

b

HEADS data from van Loy et al. (29).

FIGURE 8. Nitrate concentrations measured in Riverside (top) and Mira Loma (bottom), CA, during the summer of 1997. At Riverside, many of the study days showed two daily maxima in the nitrate concentration. The first peak occurred in the midmorning, with a sharp drop in concentration near noon, followed by a second peak in the afternoon. The relative magnitude of the morning and afternoon concentration peaks varied from day to day. The morning nitrate maximum occurred after the input of NO from the morning rush hour but prior to the increase in ozone. On several days, it was coincident with a slight increase in the indicated NO2. The afternoon increase in nitrate generally coincided with the rise in ozone concentration, with a nitrate maximum occurring near the time of the maximum daily temperature. On most days, the elevated ozone levels persisted longer than the nitrate. For the period August 28-31, the magnitude of the morning nitrate concentration peak relative to the afternoon peak increased with each successive day. The highest nitrate concentration period was observed on the morning of August 31, with 10-min concentrations in excess of 30 µg/m3 between the hours of 10 and 11 in the morning. On this day, the afternoon nitrate concentration peak is negligible by comparison. Nitrate concentration profiles for the winter measurements in Bakersfield are shown in Figure 9. The most marked feature of this data set is the period of elevated nitrate concentrations over the 7-day period January 10-16, 1999, followed by a period of relatively low concentrations. Variability within the high-concentration days was a factor VOL. 34, NO. 5, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

913

FIGURE 9. Nitrate concentrations measured in Bakersfield, CA, in January 1999. of 3-6. Unlike the summertime measurements at Riverside, the maximum concentrations did not fall at a consistent time of day. Rather, the time series shows episodes of several days duration. Summary of Field Performance. During the 6-week deployment in Riverside and Mira Loma and the 3 weeks in Bakersfield, the operation of the automated nitrate system was nearly uninterrupted, apart from routine calibrations. The system gathered data around the clock, with measurements of nitrate concentrations every 10 min. Data recovery was 97%, excluding down time for power interruptions and calibrations. This provided nearly uninterrupted diurnal patterns in nitrate concentration. The diurnal profiles observed during the different field campaigns were quite distinct. Yet comparisons in each city show that the 24-h averaged nitrate values obtained from the automated nitrate monitor are comparable to values obtained by denuder filter sampling.

Acknowledgments The authors are most grateful to Petros Koutrakis of Harvard University, Solomon Teffara and Mel Zeldin of the South Coast Air Quality Management District, and Tom Merrifield of Met One Instruments, who gave permission to publish the comparison of our results with their filter nitrate data. We thank Peter Mueller of EPRI, who provided the initial opportunity to develop this method. We thank our sponsors: EPRI under Contracts WO513501 and WO915203; the Coordinating Research Council and the National Renewable Energy Laboratory under Contract AP-22; and the San Joaquin Valleywide Air Pollution Study Agency under 98-2PM. We thank Profs. Kimberly Prather and Glen Cass, the members of their research groups at the University of California, Riverside, and California Institute of Technology, and Dabrina Dutcher of the University of California, Davis; all of whom checked our instrument in the field in our absence.

Literature Cited (1) Malm, W. C.; Sisler, J. F.; Huffman, D.; Eldred, R. A.; Cahill, T. A. J. Geophys. Res. Atmos. 1994, 99 (D1), 1347-1370.

914

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 5, 2000

(2) Charlson, R. J.; Langner, J.; Rodhe, H.; Leovy, C. B.; Warren, S. G. Tellus 1991, 43 (A), 152-163. (3) Air Quality Criteria for Particulate Matter; U.S. Environmental Protection Agency: Washington, DC, April 1996; EPA/600/P95. (4) National ambient air quality standards for particulate matter; final rule. Fed. Regist. 1997, July 18; 40CFR, Part 50; Vol. 62, No. 138. (5) Coburn, J.; Husar, R. Atmos. Environ. 1978, 12, 89-98. (6) Tanner, R. Atmos. Environ. 1980, 14, 121-127. (7) Huntzicker, J. J. Anal. Chem. 1986, 58, 654-656. (8) Allen, G. In Methods of Air Sampling and Analysis; Lodge, J. P., Ed.; Air and Waste Management Association: Pittsburgh, 1988; Section 7-13. (9) Turpin, B. J.; Cary, R. A.; Huntzicker, J. J. Aerosol Sci. Technol. 1990, 12, 161-171. (10) Patashnick, H.; Rupprecht, G. U.S. Patent No. 5,196,170, 1993. (11) Hansen, A. D. A.; Rosen, H.; Novakov, T. Appl. Opt. 1982, 21, 3060-3062. (12) Hansen, A. D. A.; Rosen, H.; Novakov, T. Sci. Total Environ. 1984, 36, 191-196. (13) Simon, P. K.; Dasgupta, P. K. Environ. Sci. Technol. 1995, 29, 1534-1541. (14) Simon, P. K.; Dasgupta, P. K. Anal. Chem. 1995, 67, 71-78. (15) Khlystov, A.; Wyers, G. P.; Slanina, J. Atmos. Environ. 1995, 29, 2229-2234. (16) Buhr, S. M.; Buhr, M. P.; Fehsenfeld, F. C.; Holloway, J. S.; Karst, U.; Norton, R. B.; Parrish, D. D.; Sievers, R. E. Atmos. Environ. 1995, 29, 2609-2624. (17) Karlsson, A.; Irgum, K.; Hansson, H. J. Aerosol Sci. 1997, 28, 1539-1551. (18) Zellweger, C.; Ammann, M.; Hofer, P.; Baltensperger, U. Atmos. Environ. 1999, 33, 1131-1140. (19) Hering, S. V.; Stolzenburg, M. R. U.S. Patent 5,983,732. (20) Roberts, P. T.; Friedlander, S. K. Atmos. Environ. 1976, 10, 403408. (21) Hering, S. V.; Friedlander, S. K. Atmos. Environ. 1982, 16, 26472656. (22) Winkler, P. J. Aerosol Sci. 1974, 5, 235-240. (23) Stein, S. W.; Turpin, B. J.; Cai, X.-P.; Huang, P.-F.; McMurry, P. H. Atmos. Environ. 1994, 28, 1739-1746. (24) Yamamoto, M.; Kosaka, H. Anal. Chem. 1994, 66, 362-367. (25) Stolzenburg, M. R.; Kreisberg, N. M.; Hering, S. V. Aerosol Sci. Technol. 1998, 29, 402-418. (26) Hering, S. V.; Cass, G. J. Air Waste Manage. Assoc. 1999, 49, 174-185. (27) John, W.; Wall, S. M.; Ondo, J. L. Atmos. Environ. 1988, 22, 15951600. (28) Zhang, X. Q.; McMurry, P. H. Atmos. Environ. 1992, 26A, 33053312. (29) Van Loy, M.; Saxena, P.; Allan, M. A. Characterization of PM2.5 and sampling method intercomparison of fine particle composition at six urban sites; EPRI report; Electric Power Research Institute: Palo Alto, CA, in press. (30) Dutcher, D. D.; Perry, K. D.; Miller, A.; Cahill, T. A.; Change, D. P. Report for the California Air Resources Board on the Instrument Comparison Study, Bakersfield, CA, 1998-1999; Final Report Contract 97-536; Submitted by University of California, Davis to the California Air Resources Board, 1999.

Received for review August 16, 1999. Revised manuscript received November 30, 1999. Accepted December 6, 1999. ES990956D