Analysis of Beta Attenuation Monitor Filter Rolls for Particulate Matter Speciation

Instruments such as Beta Attenuation Monitors (BAMs) provide real-time PM2.5 mass concentration data and are now widely used in compliance monitoring networks. These devices collect the PM on a filter tape in 1-hour samples. In this work, the use of BAM filter spots for chemical speciation was investigated. Filter tapes from several sites in California were analyzed for a series of chemical species including elements, ions, organic and elemental carbon, and molecular markers. A major issue was the blank values in the filter tape. Based on blank spot analyses, it was determined that measurement of organic and elemental carbon (OC/EC) and elements were infeasible. A total of 22 BAM samples (each comprising 24 1-hour spots) from 12 sites were analyzed for ions and black carbon (BC). Additionally, 336 1-hour spots were composited to analyze for molecular markers (MM). Measurements of ions and BC at each site appear to have been underestimated likely due to volatilization losses. MM measurements in these 336 BAM filter spots suggest that organic speciation of BAM filters could be a viable method for measuring useful marker species. Statistical analysis was conducted by grouping samples into classes with mass concentrations greater and less than 80 μg m. Measurement of Delta-C concentrations (i.e., Delta C = BC370 nm – BC880 nm) in these two groups of samples revealed that most of the high PM2.5 concentration days PM2.5 greater than 80 μg m) were likely not substantially impacted by biomass combustion particles. A major portion of the samples analyzed in this work with high concentrations of PM2.5 were inorganic species such as ammonium sulfate, ammonium nitrate and other ions. Results suggested that some useful information may be derived from BAM tape analyses. However, different filter substrates and sample handling and storage practice might make a wider range of analyses possible.


INTRODUCTION
Fine particles with diameter less than 2.5 microns (PM 2.5 ), contains microscopic solids or liquid droplets that cause serious health problems and thus, ambient concentrations are regulated to protect the public health.To ensure compliance with the air quality standards, ambient PM 2.5 concentrations need to be monitored.Recent changes in the San Joaquin Valleywide Air Pollution Study Agency's air monitoring network design, and changes in instrument selection for particulate matter monitors have led to a decrease in the number of filter-based samplers for Particulate Matter (PM) speciation.Several monitoring stations have switched to non-traditional PM samplers primarily because of the need for real-time PM 2.5 data, and continuous monitors require far fewer technician trips over the course of a year, and, thereby, reducing costs.Continuous PM monitors such as a beta attenuation monitor provide a wealth of real-time PM 2.5 mass concentration data.However, they provide no information on the chemical components that makeup the PM 2.5 mass.Traditional speciation samplers are typically needed to understand the chemical composition of ambient particles by collecting multiple samples on different filters that can then subjected to chemical analysis.
The continuous monitors used in the compliance networks in California are MetOne Model 1020 (Grants Pass, OR) Beta Attenuation Monitors (BAMs).These units measure the beta transmission through a blank piece of a filter tape, move the tape into the collection position, collect PM for an hour, move the tape back to the C-14 source and measure the transmission.The difference in beta transmission between the clean and exposed filter is related to the collected PM mass.There are 24 spots collected per day under ambient conditions unless there is high relative humidity.Then a "smart heater" warms the air to reduce the amount of water associated with the particulate matter.Good agreement has been shown with Federal Reference Method samples (Quok and MacDougall, 2006) and have been approved by the U.S. Environmental Protection Agency as Federal Equivalent Method samplers.
The BAM filter tapes are made of glass fiber filter and each filter spot is exposed for PM collection during one hour.Each spot is approximately 0.5 inches in diameter with a volumetric air flow rate through each filter spot of 16.7 L min -1 .In this report, we present the first reported analytical experiments conducted to explore the feasibility of using existing PM 2.5 samples collected on BAM filter rolls for determining the chemical composition of the collected PM 2.5 .

EXPERIMENTAL METHODS
Table 1 provides the sites, sampling dates, and the 24hour average PM 2.5 concentration measured by the MetOne 1020 BAMs.This system operates with an inlet heating system such that when the relative humidity in the inlet stream exceeds 55%, the heater turns on and raises the inlet stream temperature to 3 to 5°C above the ambient temperature.In addition to this, the 95% confidence level of the average value of twenty-four (24) 1-hour measurements are also provided in this table.Archived BAM samples were collected routinely at the nine sites including sites located at Manteca, Tracy, Turlock, Merced (at Coffee Road), Madera City, Clovis, Huron, Hanford, Porterville and Lebec.At each of these sites, samples were collected on a 24-hour basis providing 24 spots per day.These samples were collected by California Air Resources Board (CARB) and provided by San Joaquin Valley Air Pollution Control District (SJVAPCD) for use in this research project.In addition, samples were collected at the San Joaquin Valley Air Pollution Control District's Tranquillity monitoring site using a High-Volume air sampler (Tisch Environmental, Model TE-1000) equipped with 90 millimeters (mm) quartz filter and Poly-Urethane Foam (PUF) plug from November 7, 2011 to March 1, 2012.However, there was no BAM sampler deployed at this site.Fig. 1 provides the approximate locations of the sites where the samples studied in this work were collected.
There are two aspects to be determined to assess the utility of BAM tapes for chemical analysis.First, it is necessary to determine the blank values in unexposed BAM filter roll.Glass fibers typically have a relatively high level of contamination and thus, there was concern that the background values would make the determination of the deposited PM composition difficult.In order to conduct the blank analyses, filter spots (area = 0.528 cm 2 ) were punched from an unexposed BAM filter roll (Met One Instruments, Incorporated) of the same type used in the BAM units.Based on the results from these blank analyses, analyses of actual BAM filter samples were performed.
The second issue is the amount of collected PM available for analysis.Since each spot represents only a 1hour collection at 16.7 L min -1 so there is relatively little collected mass to analyze.However, with improved analytical methods, there may be sufficient sensitivity to allow useful analyses to be performed.Blank and sample filter spots were punched from the tapes and stored frozen in labeled petri dishes prior to analysis.Fig. 2 provides a schematic outline of the analytical methods used for the BAM filter spots.Blank filter spots were first measured for Black Carbon (BC), Organic and Elemental Carbon (OC/EC), ions and molecular markers.Based on the results from blank analysis, BAM samples were then analyzed for selected chemical species.In the following section, we discuss the analytical methodology for measuring each chemical species.

Analytical Methodology Black Carbon Analysis
The filter punches were placed in a Magee Scientific, Dual-Wavelength Optical Transmissometer (Model OT-21) to measure Black Carbon (BC).Five replicate measurements were made for each sample filter spot.Each measurement provides the ultraviolet (UV) and infra-red (IR) attenuation by the particles present on the filter spot.During a sample measurement, blank spots cut from a used BAM filter roll were used to determine blank attenuation.The sample filter spot (with particle loading) was then used to determine the light attenuation.Attenuation at 370 and 880 nm were measured.
Based on the attenuations, area of the filter exposed for particle collection (in square centimeter -cm 2 ) and the volume of air sampled (in cubic meters -m 3 ), the concentration of black carbon (880 nm) and UVBC (370 nm) are calculated in micrograms per cubic meter (µg m -3 ).In this work, we recommend using the classical "BC" relationship between the mass of the particles and the associated UV and IR attenuation.Delta-C was calculated as the difference between UVBC and BC (Allen et al., 2004, Wang et al., 2011).

Truncated OC/EC Analysis
Glass fiber filter punches (approximately 0.5-0.8cm 2 area) of BAM filter tape spots was measured for OC/EC using an Atmoslytics DRI Model 2001 Carbon Analyzer and a "modified" IMPROVE_A temperature profile (Chow et al., 2007).Typically, the OC/EC analyses are conducted on quartz filters.In the present work, a truncated OC/EC analysis method was employed where the upper temperature limit was set at a lower level (than the temperature levels typically used for quartz filters) to prevent melting of glass fibers, while the absorbance is being measured.We acknowledge that this "truncated" method is expected to provide only an approximate estimate of the OC/EC species.
A truncated IMPROVE temperature protocol were used to analyze blank filter punches in this work with pure helium with an upper temperature limit of 520°C for the glassfiber filter (Chow et al., 2005).The temperature profile was 120°C, 250°C, 450°C, and 520°C to measure the corresponding thermal carbon fractions derived at these temperatures, and were denoted OC1, OC2, OC3, and OC4, respectively.The temperature was ramped to the next step when the FID response returned to baseline or when it remained at a constant value for more than 20 to 30 seconds.For the EC, analysis atmosphere was switched to 2% oxygen and 98% helium.Temperature plateaus in this atmosphere were 520°C, 650°C, and 800°C for EC1, EC2, and EC3, respectively.

Elemental Analysis by ICP-MS
BAM filter spots allocated for elemental analysis were extracted using microwave assisted acid digestion (US EPA Method IO-3.1, 1999).The extracts were analyzed for elements using a Thermo X Series II ICP-MS (Inductively Coupled Plasma -Mass Spectrometry).Selected filter spots, including samples and blank spots were extracted and measured exclusively for elements (Laing et al., 2014).

Ion Chromatographic Analysis
BAM filter spots were extracted using 5 mL deionized water for 30-45 minutes in an ultrasonic bath.Aliquots of 100 µL were used for anions and cations measurement using ion chromatography.Anion analyses were performed on a Dionex DX-500 ion chromatography system.Isocratic separation was performed using an Ionpac AS4A-SC column and 2.7 mM Na 2 CO 3 /0.3mM NaHCO 3 eluent (RTI, 2003a).Cation analyses were performed on a Dionex DX-500 ion chromatography system.Gradient separation is performed using an Ionpac CS12A column and 22 mM H 2 SO 4 eluent (RTI, 2003b).

Molecular Markers (Hydrocarbons)
The organic molecular markers such as polycyclic aromatic hydrocarbons and other organic compounds were measured using Gas Chromatography/Mass Spectrometry (GC/MS).Molecular markers in BAM filter spots were measured by compositing several filter spots together.For example, Zheng et al., (2002) composited about thirtythree (33) 24-hour Southeastern Aerosol Research and Characterization (SEARCH) air monitoring network samples (47 mm quartz filters collected at 16.7 LPM) to measure solvent-extractable organic compounds.The measured organic compounds were then used as tracers to conduct source apportionment.Based on the work by Zheng et al. (2002), we estimate that a minimum of one-week of BAM filter spots are required to obtain detectable levels of molecular markers.
The extraction approach used an Accelerated Solvent Extractor (ASE 300, Dionex Co.).The ASE300 is an innovative sample preparation system that combines elevated temperatures and pressures with liquid solvents to achieve fast and efficient extraction.Prior to extraction, internal standards were spiked into the samples.These internal standards were used to monitor extraction efficiency and losses during the sample extraction and concentration steps.The samples were extracted with solvent mixture of methylene chloride and methanol in the ratio of 4:1.The extraction step was conducted over 20 min (Wang et al., 2012).
After extraction, the combined extract was filtered and the samples were concentrated.High purity nitrogen was used to evaporate the extract from 40 mL to 1 mL at room temperature in a Turbovap LV (Zymark Corporation).This concentration process normally takes approximately 2 hours.The concentrated extract was split into two portions.One portion is analyzed directly for non-polar organic analysis.The other fraction is analyzed after derivatization for polar organic analysis.
Analyses of the samples were performed with a TRACE GC coupled by Polaris Q MS system (Thermo Finnigan, San Jose, CA, USA).An Rtx-5-MS capillary column of 30 m length and 0.25 mm internal diameter with a 0.25 µm film thickness was used.The stationary phase of the column is 5% dephenyl-95% dimethyl polysiloxane (Restec Corp., PA, USA).The temperature program applied in the GC/MS analysis was initially held at 60°C for 2 minutes.It was then increased to a final temperature of 300°C with a ramp of 8 °C min -1 , and then held for 10 minutes.Total analysis time for one sample was about 45 minutes.For polar compounds, derivatization was used to convert them into more volatile derivatives that can be more easily eluted from the GC column.Briefly, derivatization is the process of chemically modifying a compound to produce a new compound that has properties suitable for analysis by GC, usually by increasing the volatility of the compounds created by elimination of polar OH, NH and SH groups.

BC Measurements
Table 2 provides sample results obtained using blank filter spots punched from an unused BAM filter roll.Each BAM filter spot is expected to sample about 1.00 m 3 of ambient air in one hour.Therefore, a nominal volume of 1.00 m 3 was assumed for each blank filter spot to calculate the BC and UVPM particle concentration in a blank filter spot.The results show that the BC concentration to be below instrument detection limit, and indicates the prospect of this methodology to measure black carbon particle concentration on a BAM filter spots with ambient particle loading.In addition, BC measurements are non-destructive such that the filter spots can be used for further chemical analysis.
Based on the diurnal profiles of the PM 2.5 mass concentration, (between midnight starting the sampling day to the midnight ending the sampling day), two spots were selected and used for BC measurements.Table 3 provides the BC, UVBC, and ΔC concentrations for selected samples collected in Clovis, CA.The table also provides the corresponding PM 2.5 mass concentrations.The average BC concentration in the two spots collected on November 17, 2012 was 37 nanograms per cubic meter (ng m -3 ) while the corresponding average PM 2.5 mass concentration was 72.5 µg m -3 .On the 11 th and 29 th of December 2011, the average PM 2.5 mass concentration was 98.3 µg m -3 , and the BC concentration in these four spots were 192.1 ng m -3 .The ΔC concentrations were positive on November 17, 2011 and on January 31, 2012 suggesting the presence of biomass combustion particles (Wang et al., 2011) BAM samples collected on the December 11, 2011 at Tranquillity were analyzed for BC, UVBC, and ions.The concentration of BC measured on the 10:00 AM spot was 44.4 ng m -3 , and 23.7 ng m -3 in the 12:00 pm spot.The UVBC concentrations were 59.4 and 44.2 ng m -3 at 10:00 AM and 12:00 PM (noon), respectively.Both spots showed positive ΔC concentrations of 15 and 20.5 ng m -3 .

OC/EC Measurements
Table 4 provides the OC and EC fraction results obtained using the truncated temperature plateaus for blank BAM filter spots made of glass-fiber.Table 4 shows that the truncated method does not work very well for the BAM filter spots (blank) because the glass-fiber filter melted at temperatures lower than 520°C producing high OC and EC values in blank filter samples.Some glass fiber filters use a binder to improve the strength of the filter and that may be the case with the BAM filter tape and would explain the high blank values.Therefore, it was concluded that measurement of OC/EC has limited value with these glass-fiber BAM filters.

Elemental Analysis
In order to measure the concentration of elements that are present in a blank filter spot, concentration of metals were measured in 1, 6, and 24 blank filter spots obtained from an unexposed blank filter roll.Fig. 3 shows the elemental concentrations in blank BAM filter spots.Results show the presence of high mass loadings of several elements present in an unused BAM filter (blank sample).Only a few elements, such as Li, V, Mn Co, Ni, Cu, As, Se, Rb, Cd, Sn, Sb, Pb and U, are below the 100 nanograms (ng) dashed line (Fig. 3) in a single BAM filter spot.In combined filter-spot analysis using 6 and 24 spots, the concentrations of elements were significantly higher than the single spot analysis, indicating aggregation of elements present in the multiple filter spots.We note that the mass present in each spot was similar, and the mass increase was approximately linear as the number of filter spots combined were increased (with an average correlation R 2 ~0.88).Results from blank analysis using ICP/MS indicate very high levels of several elements of interest in blank filter spots.Therefore, we concluded that measurement of elements in existing BAM filter rolls may not provide useful results.

Ion Chromatography Measurements
Table 5 provides the results of ion chromatographic analyses conducted on blank BAM filter spots.Results in this table indicate a high blank concentration level for potassium, chloride, and sulfate than other ionic species.We also note that the blank values have a linear increase in the concentration level as the number of spots was increased (1, 6 and 24 spots) in the analysis.This linear increase in concentration indicates consistent concentration of ionic species present in each filter spot that was analyzed (with average R 2 ~0.9).Based on the blank analysis results in Table 5, combining a maximum of twenty-four sample spots is feasible for the analysis of actual BAM samples.This upper limit on the maximum number of spots combined will help keep the blank levels of these ions lower than the typical sample amounts (particulate matter loading).If the diurnal variations of the ionic species are important, then a smaller number of filter spots may be combined.
PM samples collected using the BAM instrument were measured for anions and cations.Table 6 provides the estimated mass concentration of selected major ionic particles.Estimated ammonium sulfate was 0.03 µg m -3 and negligible ammonium nitrate concentrations were found in the BAM filter samples.The estimated mass of ammonium sulfate and ammonium nitrate in the December 11 th sample collected using a collocated speciation network sampler (RAAS 2.5-400) were 1.28 and 3.74 µg m -3 .The much lower mass concentrations of these species in the BAM filter spots suggested significant losses of these species.The loss of ammonium nitrate is to be expected (Hering and Cass (1999).However, sulfate should not volatilize and thus the low values are the result of some other loss mechanism or incomplete extraction of the material from the filter.This extraction procedure has been tested and found to be effective (Yli-Tuomi et al., 2003;Laing et al., 2013).
Since the integrated filter samples from these sites had large sulfate and nitrate concentrations but negligible masses of ammonium sulfate and ammonium nitrate measured in the BAM filters, there are clearly problems with determining these ionic species in exposed BAM filter tape.

Molecular Markers Measurements
Three-hundred, thirty-six (336) 1-hour spots from each of Hanford and Madera City were composited to analyze for organic molecular markers.The Hanford samples were from the 9 th and 10 th , and from 12 th to the 23 rd December 2011.The Madera City samples were from 10 th to the 23rd December 2011.The three-hundred, thirty-six (336) 1-hour spots from each of the two sites were composited, extracted and analyzed.The Tranquillity samples were analyzed using the same approach as Wang et al. (2012).Fig. 4 shows the chromatograms obtained from experiments on a blank BAM filter spot.The analysis for polar and noncompounds in a blank filter spot are shown together with the chromatograms for polar and non-polar compound standards.The chromatograms show that extract of the blank filter spot produced no peaks with the retention time similar to the organic compounds present the standards.However, a peak (at about 14.71 minutes) is present in the chromatogram for polar compounds.This peak represents an extraneous unidentified contamination that has mass fragments at 73, 147, and 281 dalton (Da), but do not interfere with other compounds that are of interest.
Multiple weeks of BAM samples from Hanford and Madera City were composited and extracted for molecular marker measurements using GC/MS.The mass concentrations of organic acids in the Hanford samples were lower (420.33ng m -3 ) than the mass concentrations measured in Madera  City samples.Concentration of PAHs such as phenanthrene and pyrene were higher (0.49 and 0.89 ng m -3 ) than the concentrations measured in Madera City (0.40 ng m -3 of phenanthrene and 0.59 ng m -3 of pyrene), and at the Tranquillity site (0.33 ng m -3 of phenanthrene and 0.53 ng m -3 of pyrene).The concentration of succinic acid at Tranquillity was 478 ng m -3 , nearly two times the mass concentration of succinic acid (268.23 ng m -3 ) measured in Madera City samples.Fig. 5 shows the concentrations of molecular markers measured in BAM filter samples along with the Tranquillity high volume samples.Organic species measured using high-volume air samplers and those measured using BAM filter spots show comparable mass concentration of organic acids and levoglucosan.However, mass concentration of alkanes and PAHs are typically lower in BAM samples than in High Volume air samples again suggesting the potential for loss of semivolatile species.
The mass concentration of organic acids in the Hanford samples were lower (420 ng m -3 ) than the mass concentration measured in Madera City samples.Concentration of PAHs such as phenanthrene and pyrene were higher in Hanford samples (0.49 and 0.89 ng m -3 , respectively) than in Madera City samples (0.40 ng m -3 of phenanthrene and 0.59 ng m -3 of pyrene), and in Tranquillity (0.33 ng m -3 of phenanthrene and 0.53 ng m -3 of pyrene).Concentration of succinic acid was 477.77 ng m -3 in Madera City samples, 268.23 ng m -3 in Hanford samples, and 24.37 ng m -3 (average) in the Tranquillity samples.In the Madera City BAM samples, concentration of 1-methylnaphthalene was 4.25 ng m -3 which was lower than the average mass concentration measured in the ten samples from Tranquility (10.01 ng m -3 ).

Statistical Analysis of Measured Species and PM 2.5 Mass Concentration
Analyses of the BAM sample measurements indicate that a major portion of the inorganic particle mass were very low in comparison to the measurements from a colocated standard speciation sampler (Andersen RAAS2.5-400).The concentration of the inorganic species such as ammonium sulfate and nitrate are lower.An attempt has been made to determine the major component of the PM 2.5 mass based on the BC, UVBC, and ion chromatography measurements coupled with the PM 2.5 mass concentration reported by the BAM instrument.Delta-C concentration measured in each of the spots were grouped as spots with PM 2.5 mass concentration greater than 80 µg m -3 and into spots with PM 2.5 mass concentration lesser than 80 µg m -3 .Analysis of Variance (ANOVA) suggests no influence of the inorganic species mass on the measured PM 2.5 mass.This result is likely due to the loss of semivolatile species from the filter spots as suggested by Hering and Cass (1999).However, strong correlations existed between Delta-C and PM 2.5 mass concentration.Fig. 6 shows the Delta-C concentration in the two groups of BAM filter spots (one below 80 µg m -3 and the other above 80 µg m -3 ) in a boxand-whisker plot.ANOVA indicated a statistically significant difference between the two groups at the 95% confidence level (p = 0.0001).The statistically significant difference suggests that a major portion of the PM 2.5 mass in the group with PM 2.5 mass concentration greater than 80 µg m -3 did not have large contributions from biomass combustion and were likely affected primarily by secondary inorganic particles.

CONCLUSIONS
Results from this work indicate some potential to use BAM filter samples for PM speciation.It appears that BC as well as some organic constituents can be measured.We cannot be sure that there have not been losses of some of the MM compounds.Thus, it would be best if the tapes were refrigerated or frozen after they are recovered from the sampling site.However, the filter substrate was not particularly suitable for measurement of a number of chemical species because of high blank values (elements by ICP-MS) and the apparent loss of material from the filter spots (ions as measured by IC).Therefore, there is a need to assess the potential for other filter materials and their effectiveness in the BAM instrument.There is now an in-situ XRF system that uses a moving Teflon tape that has low blank values and might also be useful in the BAM such that the resulting spots could be subsequently analyzed off-line.
In addition, sample handling and storage may have played a role in the quality of results derived from field samples.The tapes were stored at room temperatures and we are not aware that they were sealed in any way such as wrapping in aluminum foil.A review of possible changes to the existing BAM instrument and its operating procedures is recommended, so that it minimizes the loss of PM collected on the BAM filter spots.

Fig. 1 .
Fig. 1.Approximate location of the sampling sites studied in this work (Note: Green marker indicates temporary site in Tranquillity, CA and red markers indicate exclusive BAM sample sites).

Fig. 2 .
Fig. 2. Schematic diagram of analytical approach used for PM collected on BAM filter spots.

LiFig. 3 .
Fig. 3. Blank Filter Roll Analysis for Elements using Thermo X Series II ICPMS.Dashed Black Line Indicates 100 ng Mass of Each Element Measured.

Fig. 4 .
Fig. 4. Chromatograms of Blank BAM Filter Spot Analysis for Polar and Non-Polar Organic Compounds shown together with the Polar and Non-Polar Organic Compound Standards.

Fig. 5 .
Fig. 5. Comparison of Organic Molecular Markers Measured in BAM and High-Volume Air Samples.Sizes of the pie are proportional to the measured total MM concentrations.

Table 1 .
Average PM 2.5 concentrations measured by the BAM sampler for samples selected for analysis.

Table 2 .
Sample Results of blank BAM filter spot analysis for BC and UVPM particles.

Table 3 .
Concentration of BC, UVBC and delta-C in selected BAM samples from Clovis CA along with PM 2.5 concentration measured by BAM instrument.

Table 4 .
OC and EC Fractions in Blank BAM Filter Spot using Truncated IMPROVE Temperature Plateaus.Units in the data reported in this table are in µg cm -2 .

Table 5 .
Ion chromatographic analysis of blank BAM filter roll.Note: Concentrations are reported in mg L -1 .Columns highlighted gray are ions that have higher blank values.

Table 6 .
Estimated inorganic particle mass concentrations from ion chromatopgraphy analysis.