Characterization of Fine Particle Material in Ambient Air and Personal Samples from an Underground Mine

ABSTRACT

Personal samplers representing 4 job classifications and sta­tionary samplers at 2 locations in an underground mine were de­ ployed to measure fine particle carbon (organic/elemental), ions (sulfate plus nitrate), elements (metals and others), and speciated organic compounds including polycyclic aromatic hydrocarbons (PAH), oxygenated PAH, and hopanes/steranes. Chemically segre­gated size distribution was investigated after collection with a mul­tistage impactor placed at 1 sampling site.


All samples exceeded the currently proposed mine air standard of 160 „g/m3 total carbon, and most exceeded the interim standard of 400 ug/m3. Carbon ac­ counted for about 70% of the fine particle mass (described as a reconstructed mass of all measured chemical species); sulfate and ore/waste rock-derived metals constituted most of the remainder. Most of the personal samples were more concentrated than the am­bient samples; 1 sample exceeded 2.5 mg/m3 total mass. 

The PAH consisted mostly of gas-phase/semivolatile compounds and minor amounts of the particle-phase species, which is consistent with the composition of diesel exhaust, the major source of fine particle material in the mine. Size-segregated chemistry showed that the majority of the material below 1 „m of aerodynamic diameter was carbon, with the largest amount at approximately 0.2 „m. Metals, derived primarily from resuspended ore/waste rock, comprised the majority of the material above 1 „m. Results are placed in context of current mine-monitoring techniques that aim to regulate diesel particulate material.

MATERIALS AND METHODS

Study Site

The study site was an underground gold mine located in the eastern part of Nevada (in the U.S.). Activities at this mine in­volved processing 1,300 tons of ore per day using heavy-duty diesel equipment and drilling devices. The ore was processed during two 12 h shifts that required 80 miners, 20 mobile equipment mechanics, 4 fixed-maintenance mechanics, and 4 electri­cians. Samples were collected at the site in June of 1999.

The equipment used in this mine is specialized for mining operations. These include transport vehicles and vehicles used for drilling and other jobs in the mine. Most of these vehicles operate on off-road diesel fuel (1300 ppm sulfur content) with a range of commercially available diesel engines ranging from 1.5 to 12.5 L engine displacement. The equipment used in this mine did not have emissions control devices, as they are currently not required for metal and nonmetal mines.

An integral component of the mine environment is the ven­tilation system. This mine consisted of 2 main systems of ven­tilation, termed through flow and auxiliary. The through flow ventilation of the mine consists of approximately 142 m3/s of airflow that is circulated by a large surface fan blowing fresh air into the mine. 

The fresh air enters the mine via an access drift from the surface, and control doors are used in the mine drifts to direct the air to where it is needed most, typically to the working areas. The entrance/exit portal to the main decline is the main exhaust for the air after it has circulated through all parts of the mine. Auxiliary ventilation consists of smaller portable fans that are placed in areas of the mine where additional ventilation is needed.

Sampling Strategy

Personal and ambient samples were collected after size- selective sampling through cyclones with a 2.5 //m cut point. A summary of the samples that were collected during this study is presented in Table 1, and details regarding the sampling meth­ods are described in the following section. Samples were col­lected for 9- 10 h to represent the duration of a shift in the mine. Two stationary air sampling sites, or area samples, were moni­tored: an “upper” (U) site located near the entrance/exit of the mine where the ventilation air exits the mine, and a “lower” (L) site that was closer to the active mining areas. 

Based on the flow through the mine, the material collected at the U site was approximately 10 min of air transit time from the L site, so it was slightly aged. Samples were collected over the duration of 4 shifts at the U site and 3 shifts at the L site . Due to changes in the mining activity, the L site was moved closer to the ac­tive mining areas following the first sample day. Thus data from the first period (L#1) were obtained from a different location than data for the 2 later periods. 

For personal exposure samples, miners were selected at random to carry the sampling devices. Personal samplers were placed immediately next to the miners throughout the course of their shifts. Therefore personal sam­plers were not taken directly in the breathing zone of the miner, but rather immediately adjacent to the miner as they worked and moved throughout the mine. This approach does not measure exposure directly, but is a better assessment than area samplers for occupations that are not stationary or work in confined spaces not measured by the area samples. 

Personal samples were col­lected for truck drivers, bolters, jammers, and a shifter. Bolters operated diesel vehicles that drilled and reinforced surfaces in the mine. Jammers drove heavy-duty diesel vehicles that packed processed waste rock back into cavities in the mine. 

A shifter supervised operations inside the mine. Fifteen personal sam­ ples were collected, representing shifts for 2 bolters, 4 jammers, 2 shifters, and 7 truck drivers. In addition, 1 sample was col­lected to include 2 side-by-side personal samplers along with 1 of the stationary samplers to compare the 2 methods. Three personal samples were also collected in the offices of some mine employees located above ground (see Table 1).


Sampling Methods
Ambient and personal samples were collected on Teflon (Pall-Gelman, 2.0 //m pore size, 47 mm diameter) filters to determine gravimetric mass and elements, quartz filters (Pall­ Gelman, QAO, 47 mm diameter, heat treated) for carbon and ion analysis, and a Teflon-impregnated glass fiber (TIGF; Pallflex, 110 mm) or quartz fiber filter (for personal samples; Pall-Gelmann, QAO, 47 mm, heat treated) backed by a polyurethane foam (PUF)/XAD-4/PUF sandwich cartridge for speciated
particle-phase and semivolatile OC. Ambient samplers were op­ erated at a flow rate of 10-20 Lpm for the Teflon/quartz filter samples and 113 Lpm for the TIGF filter. 

During this study per­sonal samplers were constructed with a 2.5 //m cyclone (BGI, Incorporated, Waltham, MA) that was described by Koistinen et al. (1999) and Kenny and Gussman (1997). These samplers were operated at 5 Lpm, with flow split between Teflon and quartz fiber filters. Quartz filters were backed by the PUF/XAD- 4/PUF cartridges (Zielinska et al. 2001).

In addition to the ambient and personal samples, micro-orifice uniform deposit impactor samplers (MOUDI; MSP Corp., Minneapolis, MN) were implemented at the U site to provide size-resolved chemical characterization. Two MOUDI samplers were used at this site, one fitted with Teflon media for elemental analysis and the other with aluminum media for carbon and ion analyses. 

The top MOUDI stages (which would nominally be 5.6 ¹m up to 10 ¹m) were greased with high vacuum grease to collect any larger particles that could have bounced and created positive artifacts at lower stages. The stages at 3.16, 1.78, 1.00, 0.54, 0.37, 0.148, 0.105, and 0.054 ¹m (50% cutoff) were used without grease with an after  filter to collect particles smaller than 0.054 ¹m. These lower collection stages were ungreased because grease will interfere with the chemical measurements. This approach has been used in numerous investigations of compositional size distribution using the MOUDI (e.g., Kleeman et al., 2000). Particle bounce can still be a concern at these lower sizes, but as shown by Jones et al. (1983), it is greatly decreased for particles <3 ¹m. Each MOUDI operated at a flow rate of 30 Lpm, which was checked before and after each sample with a calibrated rotameter. Between the pump and the MOUDI sampler was a 20 L air tank that served as a ballast to limit the vacuum impact on the MOUDI by allowing the vacuum to increase slowly and minimize problems with pressure surges in the sampler that dislodge sample substrates.


Analytical Methods

The Desert Research Institute (DRI) Environmental Analysis Facility has documented the methodology used for the analysis of particulate mass by gravimetry, carbon by thermal/optical reflectance, ions (sulfate, nitrate) by ion chromatography, and elements (all period 3, 4, and 5 elements except Ar, Sc, Ge, Nb, Te, Ru, and Rh; other quantified elements included Ba, La, Au, Hg, Tl, Pb, and Ur) by x-ray fluorescence (Chow et al. 1993; Watson et al. 1999), so the methods will not be discussed here.

The x-ray fluorescence (XRF) measured elements include the alkali earth metals and transition metals as well as several met­alloids and nonmetals, and these analytes are often grouped to­gether for discussion as XRF elements. The analytical methods used by the DRI Organic Analytical Laboratory to character­ize the speciated semivolatile and particle-bound organic com­ pounds are briefly described below.

Before shipment to the field, sampling media (TIGF filters, PUF, XAD-4) were thoroughly cleaned as described elsewhere (Zielinska et al. 1998). One set of cleaned media from each batch was analyzed for the same compounds as in the samples. Any chemical species quantified on the clean sample media was subtracted from results of the collected samples.

For sample preparation, the  filters and XAD-4 were subjected to microwave extraction (CEM 1000) with dichloromethane and combined with the PUF material that was soxhlet extracted for 6 h in hexane/ether (90:10). Prior to extraction, samples were spiked with a suite of deuterated internal standards that closely resembled the chemical structures and behavior of the analytes of interest. These internal standards served to mimic the behavior of the target analytes throughout the extraction process. The internal standards used included 12 PAHs ranging in volatility from naphthalene-d8 to coronene-d12 (for quantification of PAHs) and tetrocosane-d50 (for hopanes/steranes). After extraction, solvents were concentrated by rotary evaporation and filtered through a 0.2 ¹m acrodisc filter (Waters, Inc.). Extracts were then evaporated to »100 ¹l under a gentle stream of nirogen and brought to 0.1 ml (for personal samples) or 1 ml (for
ambient samples) with acetonitrile prior to analysis. 

For the hopanes and steranes, samples were fractionated before analysis using a solid-phase extraction (SPE) technique adapted from Wang et al. (1994a,b). The samples were cleaned on 6 ml Supelco SPE cartridges packed with 0.5 g of SiOH. Hopanes, steranes, and the tetracosane-d50 internal standard were eluted with 1 ml of hexane.

Samples were analyzed on a Varian Star 3400CX gas chromatograph equipped with an 8200CX automatic sampler interfaced to a Saturn 2000 ion trap operated in electron impact ionization and selective ion storage analysis mode. Splitless injections (1 ¹l) were made onto a phenylmethylsilicone fused-
silica capillary column (30 m, 0.25 mm £ 0.25 Mm; DB-5 ms, J&W Scienti c). The operating conditions were as follows: 65±C for 2 min; 12±C/min to 180±C; 8±C/min to 320±C; hold at 320±C for 10 min

Samples were quantified by comparing the response of the deuterated internal standards to the analyte of interest. Ana­lyte response was referenced to calibration curves created from standard solutions made with authentic PAHs purchased from Aldrich, Inc., and the National Institute of Standards and Tech­nology standard reference material (SRM 1647). The following authentic standards were used for quantifying hopanes and ster­anes: C27 20R -5®,14®,17®-cholestane (purchased from Aldrich), and 17^(H), 21^ (H)-hopane (purchased from Chiron AS, Norway). The remaining hopanes and steranes were iden­tified based on comparison of their mass spectra and retention time to data available in the literature (Wang et al. 1994a,b). The hopanes and steranes for which authentic standards were not available were quantified based on the response factor of standards most closely matched in volatility and retention char­acteristics.


RESULTS
Comparison of Personal/Stationary Samplers

Samples were collected side-by-side during 1 sampling pe­riod to compare results from the stationary samplers at the upper site (U#4) and 2 personal samplers. Table 2 shows results of the comparison for an abbreviated list of compounds , includ­ing the most abundant species measured and some species that were close to the analytical detection limits. 

Percent differences are presented in Table 2 for comparisons between the area sam­ple (U#4) and the average/range between 2 personal samplers. Agreement between personal and stationary samplers were bet­ter than 20% for carbon and the sum of measured species, 24% for mass, and 93% for sulfate. Sulfate agreement was very poor between the 2 sample types, but good between the 2 personal samplers (range D 1.1).

In contrast, the total sulfur concentra­tion agreed within 21% between the 2 samplers. Total sulfur was measured with a different technique (XRF versus ion chromatography) and on a different filter (Teflon). Excess sulfate measured on the higher volume quartz filter may have arisen as an artifact from sorption and oxidation of sulfur dioxide onto the filters, as described by Coutant (1977). The better agreement for sulfur measured on the Teflon filters compared to quartz mea­ surements of sulfate can be explained by the fact that Teflon is more inert to sulfur dioxide gas sorption than quartz (Coutant 1977). 

The difference between the personal and area sampler can be explained by the much higher volume of air that passed through the area sampler, giving more contact with the sulfur dioxide gas. Without more samples for comparison it is difficult to conclude if this artifact is present in all of the area samples.

Most of the XRF elements and the more abundant speciated organics were within 30% of each other. Many of the speciated organics in these samples had agreement worse than 30%. Poor agreement for many of these compound s can most likely be at­ tributed to the fact that lower volumes of material (~40 times) were collected in the personal samples so there was less material available for accurate analysis. 

For the more abundant speciated organics it is difficult to speculate why larger differences were observed, especially with the low amount of samples used in this comparison. Further work needs to be conducted to thoroughly quantify differences between these measurement tech­niques with more samples and in a controlled environment. For this study only these few samples were collected side-by-side .


PM2.5 Concentration and Composition

The bulk chemical composition of the reconstructed particle mass (RCM) for ambient and personal samples is shown in Figures 1 and 2. The reconstructed material consisted of EC, OC, sulfate, and summed elements. Some individual compo­nents were modified to reflect their true mass in the environment. OC was multiplied by 1.2 to compensate for oxygen, hy­drogen, and other mass not accounted for by the thermal/optical reflectance carbon analysis (Pierson and Brachaczek 1983). 

The OC presented in Figure s 1 and 2 includes both the speciated or­ganic compounds measured for the present study and any other organic material in the mine air. Some of the XRF elements were corrected by multiplying aluminum by 1.89, silicon by 2.14, and iron by 1.43 to include unaccounted oxygen associated with the oxides of these elements.

A comparison of the RCM to the gravimetrically determined mass for the ambient and personal samples is shown in Figure 3. The RCM is approximately 22% lower than the gravi­ metric mass. Part of this difference may be due to the uncertainty in the correction factor that was applied to the OC measured by thermal/optical reflectance. 

The 1.2 factor was used because it is based on previous corrections to motor vehicle exhaust (Pierson and Brachaczek 1983), likely to be the predominant source of OC in the mine. Other researchers have used larger numbers (1.4) for ambient air pollution measurements, but it was not possible to know the correct factor to use without know - ing the exact elemental composition of the OC in these sam­ples. Another potential source of the difference between RCM and gravimetric mass may be due to the presence of a chem ical species collected on the filter but not accounted for by the chemical analyses. 

A likely candidate for this unidentified species is particle-bound water. The filters were equilibrated at 30% relative humidity before and after sampling. However, this may not be adequate to remove water that is adsorbed to the filters during sampling prior to weighing. The presence of particle-bound water on filters has been identified and discussed in the literature (Van Loy et al. 2000). Van Loy et al. (2000) concluded that water may contribute significantly to the gap between chemical species and gravimetric mass. For the re­mainder of this paper, RCM will be used instead of gravimetric mass.

Figure 1 shows that, with the exception of sample L#1, the samples collected at the L site (closer to mining operations) were about 40% more concentrated than at the U site (located toward the mine exit). As mentioned earlier, the L site was moved after the first sample. The differences are clear in the total amount de­tected at the first location versus the second—there was nearly 2.5 times as much mass at the second location. However, the composition between L#1 and the other samples remained rela­tively constant.
The average concentration of RCM for personal samples, as shown in Figure 2, was about 1.1 mg/m3 with a standard devia­ tion of 0.58 mg/m3. The single highest value was over 2.5 mg/m3 for a jammer. These results are comparable to an underground mine exposure study by Cantrell et al. (1993), who observed an average of 0.89 mg/m3 mass with a standard deviation of 0.44 mg/m3 for samples collected with personal-sampling de­ vices that had 0.8 //m cut point impactors. The size cut point during the current study was 2.5 gm, so more material would have been collected using the samplers implemented here.

Both the stationary ambient samples and the majority of per­sonal samples shown in Figures 1 and 2 consisted of 20- 30% OC and 50- 70% EC. Exceptions were observed for the first bolter, the second shifter, and the second and the fourth truck drivers. For all of these samples, the OC/EC distributions were close to even. These samples were enriched in OC, which could have arisen from exposure to drilling oils or cigarette smoke or the diesel vehicles operating in a mode that produce d a more evenly distributed OC/EC composition relative to what was most com­ mon in the mine. 

The OC/EC distribution in diesel emissions will shift depending on the operating conditions and load on the vehicle. A vehicle operating under heavy load will have an OC/EC distribution consistent with what was observed in the mine air (20/70), whereas a vehicle at idle or low load can have much more OC present (Shi et al. 2000). Sulfate represented an average of 5- 10% of the reconstructed mass in both personal and ambient samples, whereas the XRF elements composed an average of 12- 14% of the reconstructed mass. The bulk of the XRF elements were composed of species such as silicon, calcium, and iron that would be derived primarily from resuspended ore/waste rock in the mine.

Size-Segregated Chemistry and Mass of Mine Aerosols


The size distributions of OC, EC, sulfate, and the most abun­dant elements are shown in Figure 4 for a M OUDI sample from the U#3 sampling period. OC, EC, and sulfate were collected on the MOUDI operating with aluminum collection substrates, while a simultaneous sampler was operating to collect material on Teflon substrates for XRF elemental analysis. For the sam­pler deployed with the Teflon substrates, the substrate holder for the seventh stage (0.105 im) was improperly loaded in the lab prior to shipment to the field, so that when the substrate holder was installed the MOUDI would not seal properly. 

Since it was not possible to repair the filter holder without contaminating the sample, a spare substrate holder was installed in its place. This holder’s surface was greased to collect particles at this stage while not affecting the other stages. Thus no data were obtained at this size for elements; however, this did not impact any other data obtained for this sampler. The impact of this on the overall presentation of the data is small, as the elements measured on the Teflon substrate are primarily of geologic origin and are mostly collected on the higher size ranges.

The compositional size distribution is portrayed in Figure 4. Data are presented as the fractional mass amount (dm/m total) for each incremental dlogdp. Since the MOUDI stages are 50% cut points, and there are mass on each stage that overlap from both the higher and lower cut points, aerodynamic diameters are represented as the geometric mean between impactor cut points. As shown, a considerable amount of material is present in the larger size fraction (~3 ^m). This fraction is comprised of mostly inorganic species originating from resuspension of ore/waste rock in the mine. 

The 3.16 im collection stage was the highest MOUDI stage used in this study. Previous studies (Marple et al.1986; Cantrell and Rubow 1991) have shown that a large amount of mass is present in size fractions between 2 and 10 gm. Since the current study focused on the fine-particle fraction (<2.5 im), it is likely that a large portion of the coarse particle mass was not accounted for in this mine. The size data presented here suggest that this coarse fraction was composed primarily of ore/waste rock-derive d material.

The smaller size fractions (<1 gm) showed small amounts of the geologically-derived material and an enrichment of carbonaceous material. This was especially pronounced in the “ac­ cumulation mode,” which was observed at about 150- 200 nm aerodynamic diameter. The accumulation mode is a size range where the mass of carbonaceous material tends to accumulate after smaller particles have coagulated and condensed onto larger particles. This is very common for emissions from diesel vehicles, which make up the majority of the material in this size range in the mine. Figure 4 shows that the ratio of EC to OC changed throughout the distribution. In the largest and small­ est size fractions shown, OC was enriched or present in nearly equal abundance compared to EC. In the size range s around the accumulation mode, EC was enriched by at least a factor of 3:1 compared to OC.

Other species were measured for their size distributions but are not shown in Figure 4. With the exception of sulfate, the ma­jority of these species were at or below the analytical-detection limits during this study. Sulfate was evenly distributed among the higher stages at approximately 2 / g/m3 per stage, absent in the accumulation mode stages, and enriched in the lowest size fraction at about 7 /g/m3. The sum of sulfate collected on the MOUDI stages was about 1 of the measured sulfate collected on the stationary sampler during this period. In addition, sul­fate quantified in the stationary samplers was collected under different conditions with a different filter type (quartz versus aluminum) and pressure drop (MOUDI samplers run under a larger vacuum than a single-stage filter).

Speciated Composition of Mine Air and Personal Samples

Table 3 shows an abbreviated list of the bulk material, elements, and speciated organic compounds measured in this study. Data are shown as averages representing the L site, U site, 4 jobs in the mine, and office samples taken with the personal samplers as a control. Measurement uncertainties presented for these data are the larger of the one-sigma standard deviations between tests or the root mean square of the analytical measurement uncertainties (SQRT((replicate precision ¤ analyte concentration)2 C (analyte detection limit)2 ). Most of the uncertainties reflect the standard deviations between averaged samples, as the between sample variability was greater than the analytical uncertainty in most cases. As shown in Figures 1 and 2, carbon, sulfate, and the soil-derived XRF elements were the major chemical species identified for all samples in the mine.

Table 3 shows that the gas-phase PAHs (methylated naphthalenes and biphenyls) were the most abundant organic com pounds in all samples collected in the mine, followed by the semivolatile PAHs (phenanthrene, methyl-fluorene, and methyl-and dimethyl-phenanthrene). The heavier particle-phase PAHs (benz(a)anthracene-coronene in Table 3) were in much lower concentrations. Both the hopanes and steranes were present in mine air and in personal samples, with the sum of hopanes approximately 10 times more concentrated than the sum of steranes at both stationary sampling sites. Hopanes and steranes were not quantified in all personal samples collected. For the personal sample from which the hopanes and steranes were quantified, the analytical uncertainty was high because of the small amounts of material collected and low compound concentrations.
