Due to the abundance of silicon dioxide mineral content in the earth’s crust, crystalline silica exposure associated with anthropogenic activities is prevalent in many industries, including mining, construction, and manufacturing. There are several polymorphs of crystalline silica, with quartz, cristobalite, and tridymite being those of particular concern. Of these, α-quartz, a thermodynamically stable form of crystalline silica, is most commonly encountered in industry.^[1] An analysis of respirable dust samples collected in the United States mining industry from 1997–2011 revealed a mean percent silica concentration in the metal/nonmetal mining industry and sand and gravel mines of 8.2% and 9.8%, respectively,^[2] with silica content variability increasing inversely with sample mass. Drilling, cutting, and grinding of source media with even limited crystalline silica concentrations have the potential to disperse substantial amounts of respirable sized particles. The respirable dust dose along with exposure duration and crystalline silica content are all important factors in the development of occupational lung disease.^[3]

Silicosis, a form of pulmonary fibrosis, is caused by the inhalation and subsequent deposition of respirable crystalline silica (RCS).^[4] While the association of RCS exposure and silicosis has been known for over a century, the U.S. Occupational Safety and Health Administration (OSHA) estimates that 2.3 million U.S. workers remain at risk.^[5] A recent review of silicosis deaths in the U.S. revealed a statistically significant decline from 164 (death rate = 0.74 per 1 million population) in 2001 to 101 (0.39 per 1 million) in 2010 (p = 0.002). While these statistics appear promising, new occupations and tasks (hydraulic fracturing of gas and oil wells and fabrication and installation of engineered stone countertops) were identified in addition to traditional occupations that place workers at risk for silicosis.^[6] Of the mining industry silicosis deaths from 1990–1999, 86 (42%) occurred in metal mining.^[7] In addition to silicosis, RCS exposure increases the risk of emphysema, bronchitis, tuberculosis, and renal disease.^[8,9] Crystalline silica dust (quartz and cristobalite) is classified as a Class I human carcinogen by the International Agency for Research on Cancer.^[10]

Historically, in the United States regulatory occupational exposure limits for respirable dust containing silica have been dynamic values with reductions based on the quantity of crystalline silica in the sample.^[11–14] These exposure limits consider the three polymorphs of crystalline silica (quartz, cristobalite, and tridymite). Both OSHA (historically) and MSHA (metal/nonmetal sector only, e.g., mining operations other than coal) define a further reduction (1/2 that calculated for quartz) in the allowable exposure limit for respirable samples containing crystobalite and/or tridymite.^[11,14] The American Conference of Governmental Industrial Hygienists (ACGIH^®) recommends a threshold limit value (TLV) for RCS (quartz and cristobalite) of 0.025 milligrams per cubic meter (µg/m³).^[15] Recent regulatory efforts in the U.S. have resulted in an OSHA final revised 8-hr time-weighted average (TWA) RCS (quartz and cristobalite) standard with a single-value permissible exposure limit of 50 micrograms per cubic meter (μg/m³), coupled with an action level of 25 μg/m³. This final rule on silica became effective June 23, 2016, with initially 1 and 2 years allotted for construction and general industry/maritime compliance, respectively;^[5] OSHA extended this enforcement date in the construction industry to September 23, 2017, citing the need for additional outreach and enforcement personnel training.^[16] MSHA indicated in its regulatory agenda that it will consider a similar revision to the rule for crystalline silica exposure limits for metal/ nonmetal mines, but for now the strategy for a respirable dust exposure limit remains a dynamic value based on the quantity of crystalline silica.

Silicosis is a preventable disease. Regular evaluation of worker breathing zone RCS exposures along with applicable control measures are critical components to an effective disease prevention strategy. Current air sampling techniques for RCS include the use of a cyclone for the size separation and collection of respirable particles (4-μm 50% cut-point) on polyvinyl chloride (PVC) filter media.^[17] The gravimetric res-pirable dust concentration, along with the crystalline silica concentration, is then typically quantified by an independent laboratory. The two principal analytical techniques used to quantify RCS are X-ray diffraction (XRD) and mid-infrared spectrometry (IR). Of these, XRD is specified for metal/nonmetal mining (NMAM 7500)^[18] and OSHA (ID-142)^[19] compliance, while IR is used for coal dust RCS analyses.^[20,21] While direct- on-filter RCS analysis employing IR and XRD is specified in some international methods,^[22] in the U.S., IR and XRD laboratory analyses are currently indirect methods which require multiple laboratory preparation steps to analyze the RCS concentration. These steps include pre- and post-filter weighing, muffle furnace, plasma ashing, or solvent dissolution to remove the filter and organic interferences, and redeposition of the sample onto a separate filter or into a potassium bromide pellet, followed by analysis. These preparation steps are time consuming, and, as a result, RCS laboratory results are typically not available for 1–2 weeks, potentially delaying industrial hygiene control strategies and worker protection. This slow turnaround of data is especially impactful in occupational environments with rapidly changing exposure conditions.

Field-based analytical techniques for RCS are currently not available commercially. The Pittsburgh Mining Research Division, a division of the National Institute for Occupational Safety and Health (NIOSH), initiated a program to evaluate the application of a field-portable, end-of-shift, direct-on-filter method for RCS measurement. Direct-on-filter quantification of RCS via IR works without ashing and sample redeposition laboratory preparation and has been demonstrated as a viable method for quantifying RCS.^[23,24] This technology will potentially enable an on-site industrial hygienist to quantify RCS at the conclusion of the sampling period. While XRD is advantageous in that it can distinguish between the polymorphs of crystalline silica, IR technology was selected due to its widespread availability, cost, field portability, and ease of use. Recognized disadvantages of IR technology include limitations resulting from absorption peak interferences.^[22,25] When infrared absorption spectra are graphed, the absorbance peak of crystalline silica is at or near a wave number of 800 cm^-1. Additional minerals in the sample, such as clay- and mica-based minerals that have absorption peaks coinciding with this crystalline silica analytical peak, may produce erroneous results, and spectrum corrections for these confounding minerals typically must be applied.^[21,25]

Several applications of IR technology exist. Fourier transform IR (FT-IR) spectrometry has been most commonly used in field-portable, end-of-shift, direct- on-filter assessments.^[26–31] FT-IR uses an IR emitting source to illuminate the sample with numerous wave lengths over a wide spectral range. The raw data from the returning signal must then be analyzed mathematically by a Fourier transform algorithm.^[32] FT-IR may be operated in several different modes including transmission, diffuse reflection, and attenuated total reflection. FT-IR performed in the transmission mode has been demonstrated as an effective method for quantifying a-quartz directly on PVC filter media obtained from laboratory-generated coal dust^[26,27,29] as well as PVC filter media from active coal mines.^[31] The FT-IR technique yielded strong correlations with the IR MSHA P7 integrated laboratory method when a correction for kaolin clay, a major confounding mineral associated with coal dust, was applied.^[26] The correlation of silica mass detected with FT-IR vs. the IR MSHA P7 method was strengthened with a partial least squares regression of the transmission spectra approach.^[31] Weakely et al. expanded this research into non-coal mines with analysis of respirable dust from field surveys collected from silver, granite, and limestone mines, and demonstrated that partial least squares regression of the transmission spectra most accurately predicted silica mass relative to laboratory XRD analysis.^[28] To date, no comparisons between metal/nonmetal RCS concentrations analyzed using FT-IR and XRD have been reported in the scientific literature. Such a comparison is needed before the FT- IR method can be employed by industrial hygienists to evaluate occupational RCS exposures in the metal/ nonmetal mining industry.

The goal of this study was to demonstrate the application of RCS concentration analysis via FT-IR in comparison with results obtained via laboratory XRD analyses using field samples collected from three metal/nonmetal mines. The goal will be addressed through three research objectives. First, the strength of the associations between RCS concentrations at each mine using the FT-IR and XRD methods will be assessed. Second, a prediction equation will be developed to predict XRD concentrations given FT-IR concentrations for each mine. Finally, the level of agreement between the FT-IR RCS concentrations, the predicted XRD concentrations, and the actual XRD concentrations will be evaluated.

Area air sampling for RCS was conducted in three metal/nonmetal mines in the northwestern United States, including an underground gold mine (Mine A), a surface copper mine (Mine B), and an underground silver/lead/zinc mine (Mine C).

In Mines A and B, which were actively producing mines, the area sampling was conducted in the same locations (near crushing and conveying processes) during each sample day. These areas were selected with assistance from mine management and were designated as the areas and processes in the facilities that were anticipated to generate the highest concentrations of particulate matter. Mine C was a non-producing, university campus mine training lab housed in an historic underground silver/lead/zinc mine abandoned in the mid-1950s. No crushing and conveying processes occur at Mine C, unlike at Mine A and Mine B. Mine C is used to train students in underground mining practices and does not produce a saleable product. Student mining occurs in the host rock and activities are for instructional purposes. Since ore was not processed from this facility, area sampling was instead performed near tasks anticipated as most likely to generate the largest fraction of particulate matter. Dust-generating tasks included jackleg drilling, blasting, and mucking.

Respirable dust area air sampling was performed at each mine site using GK 2.69 cyclones paired with 37 mm three-piece cassettes, low ash polyvinyl chloride (PVC) filters with a pore size of 5 μm, and SKC Airchek pumps. The GK 2.69 cyclones were pre- and post-calibrated using a high-flow Bios Defender Drycal primary flow meter at 4.2 L/min. Sample durations ranged from 2–5 hr, and sample volumes were within the recommended limits according to the NIOSH methods for silica determination— NIOSH7500, NIOSH7602, and NIOSH7603.^[18,20,33]

Gravimetric analysis of dust samples was performed via the NIOSH Manual of Analytical Methods (NMAM) 0600, particles not otherwise regulated, respirable.^[17] Samples were then analyzed by FT-IR with a Bruker Optics model Alpha FT-IR spectrometer operated in the transmission mode at 4 cm⁻¹ resolution to determine the RCS concentration in μg/m³. Because the samples are unaltered during the FT-IR analysis, it was possible to perform subsequent analysis on the same samples. All the samples were submitted for further analyses via XRD by an independent American Industrial Hygiene Association (AIHA^®) accredited lab via NIOSH 7500, crystalline silica by XRD (filter redeposition).^[18]

The limit of quantification (LOQ) for respirable dust analysis was 5 μg. The limit of detection (LOD) and LOQ for RCS analysis via the FT-IR method were 5 and 15 μg, respectively, while the LOQ for RCS via XRD through the accredited lab was 17 μg. Only samples meeting or exceeding the RCS LOQ values were considered for data analysis.

In addition to the respirable dust air sampling, a minimum of three bulk ore samples were collected at each mine to characterize the mineral content of the ore. 500 g of bulk ore fine particulate were collected within close proximity of the air sampling sites and placed in labeled, sealed, plastic bags. The bulk samples were later analyzed via XRD with the Rietvield quantification method.^[34]

Summary data for the air sampling conducted in the three mines are provided in Table 1. Of the 40 samples collected from Mine A over 9 days, 10 samples did not meet the 15 μg LOQ criteria. Twenty-five of the 27 samples collected during 9 sampling days from Mine B met the LOQ criteria. Of the 29 samples collected over 11 days from Mine C, 18 samples did not meet the 15-μg sample silica criteria. It is postulated that the limited activities in Mine C, limited sample durations, and effective dust control measures were factors. In order to ensure an adequate number of samples, bulk material from Mine C was dispersed via air flowing into a glass vessel containing the dust, which fed aerosolized material into a calm air dust chamber, and nine additional respirable dust samples were collected employing the same sampling methods previously described. All of these additional samples met the silica LOQ requirements and were included in the XRD and FT-IR analysis for Mine C, for a total of 20 viable samples, as reported in Table 1.

The mean percent silica for samples meeting the LOQ criteria as determined via FT-IR in each mine was 20.3% for Mine A, 13.4% for Mine B, and 13.5% for Mine C. The geometric mean RCS concentration obtained via FT-IR and XRD were 96 and 105 μg/m³, respectively, for Mine A, 125 and 161 μg/m³ for Mine B, and 300 and 398 μg/m³ for Mine C. The mean percent error between the RCS concentrations as determined using FT-IR relative to the concentrations determined via XRD was 12%, 22%, and 28% at Mine A, B, and C, respectively. The RCS concentrations statistics for each mine analyzed using both methods are presented in Table 2. A strong, positive association was observed between FT-IR and XRD RCS concentrations. The Spearman correlation coefficient for the FT-IR and XRD RCS concentrations was 0.97 for Mine A, 0.86 for Mine B, and 0.96 for Mine C. Spearman coefficients for each mine are shown in Figure 2.

The RCS concentrations measured at each mine were log-transformed, and the matched FT-IR and XRD pairs were assigned random numbers. Half of the pairs were randomly selected to construct linear regression models for each mine. The linear model proved to be a good fit for the data, with R-squared values of 98.7%, 96.8%, and 87.8% for Mines A, B, and C, respectively. The regression coefficients for Mines A, B, and C were 1.07, 1.15, and 0.99, respectively. Fitted-line plots and regression model equations for each mine are shown in Figure 3.

The mean differences between the original log-transformed FT-IR RCS concentrations and the original log- transformed XRD concentrations for the unselected half of the data were —0.072 for Mine A, —0.120 for Mine B, and —0.134 for Mine C. The mean differences between the predicted log-transformed XRD concentrations and the original log-transformed XRD concentrations at each mine were —0.049, —0.027, and —0.021, respectively. There was a statistically significant reduction in the mean differences in log concentrations for Mine B (p = 0.002) and Mine C (p = 0.044) when using the predicted XRD concentrations as compared to the original FT-IR data. The improved level of agreement for Mine A was not statistically significant (p = 0.215). The quantitative levels of agreement analysis results are summarized in Table 3.

A follow-up analysis was performed to determine if the lack of a statistically significant improvement in the level of agreement for Mine A following correction was due in part to the high level of agreement in the original data. A Wilcoxon rank sum test was performed to see if there was a difference between the mean log-transformed FT-IR RCS concentrations and the mean XRD RSC concentrations at Mine A without applying a correction factor. Indeed, there was no statistically significant difference between the two analytical methods without the use of a calibration correction (p = 0.534).

The qualitative Bland-Altman plots using untransformed data for each mine are provided in Figure 4. The method mean differences (bias) and 95% limits of agreement are provided in the plots. Generally, most of the paired difference fell within the limits of agreement, and the variability of agreement increased as the average of the paired measurements increased.

In addition to quartz, other minerals were identified in the bulk samples collected from each mine. Quartz, albite, and muscovite were common in all three mines. In addition to these minerals, Mine A contained orthoclase, pyrite, and jarosite; Mine B contained calcite, microcline, kaolinite, and montmorillonite; and Mine C contained calcite, kaolinite, montmorillonite, orthoclase, basanite, bloedite, hornblende, and apatite. The mean percentage by weight of crystalline silica (quartz) present in the bulk samples was 22.7% for Mine A, 25.7% for Mine B, and 16% for Mine C.

The variability of the RCS concentrations as determined using the two methods were fairly comparable, as evidenced by similar standard errors and standard deviations (Table 2). The observed level of precision of FT-IR in this study was consistently lower than XRD with higher coefficients of variation observed at each mine, although FT-IR was within 10% of XRD coefficient in each case.

Three research objectives were proposed to demonstrate the application of RCS concentration analysis via FT-IR in comparison with results obtained via laboratory XRD analyses using field samples collected from three metal/nonmetal mines. The initial objective was to assess the strength of association between RCS concentrations at each mine using FT-IR and XRD methods. The results of this study demonstrate that direct-on-filter analysis of RCS concentrations analyzed via FT-IR were highly correlated with results obtained via laboratory XRD analysis. Strong, positive correlations between FT-IR and XRD RCS concentrations in this study ranging from 0.86–0.97 are comparable to those reported by Miller et al.^[26] (0.90–0.97)^[26] in an assessment of crystalline silica mass in coal dust via FT-IR vs. the MSHA P7 analytical method.

In the second objective of this study, we demonstrated that an inverse linear regression model can be used to predict log-transformed XRD RCS concentrations given FT-IR RCS concentrations and to generate a site-specific calibration correction factor. The simple linear model proved to be a good fit for the data. The high R² values confirm that virtually all of the variability observed in FT-IR-derived RCS concentrations can be explained by the variability observed in the XRD concentrations. Further, the regression coefficients (slopes) observed in each mine’s linear model using RCS concentration in the log-scale were close to 1 (range 0.99–1.15, as shown in Figure 3).

The mine-specific calibration factor derived from the inverse linear regression model was applied to the second half of the paired data set for each mine. Considering this calibration factor, the level of agreement between the FT-IR RCS concentrations, the predicted XRD concentrations, and the actual XRD concentrations was then evaluated to fulfill objective three. Of the three mines, prior to any statistical modeling, the strongest positive association between FT-IR and XRD RCS concentrations was observed with Mine A. When the calibration factor was applied to the second half of paired data for Mine A, an improved level of agreement was observed but the improvement was not statistically significant. As verified by the follow-up analysis, this insignificant improvement was likely attributable to the already very high level of agreement between both methods at Mine A. The correction factor did result in a statistically improved level of agreement for Mines B and C. These results suggest that while strong, positive correlations were initially observed in RCS data obtained via FT-IR and XRD analysis, the application of a correction factor, specific to each mine, improves the prediction of XRD RCS concentrations using FT-IR.

The qualitative Bland-Altman plot analysis revealed important information related to the untransformed results from each mine. First, the bias for each mine was negative, indicating that the FT-IR tended to underestimate RCS concentrations as compared to XRD. Second, the limits of agreement, like the data themselves, varied considerably for each mine, from a relatively narrow range of differences of approximately 70 ug/m³ for Mines A and B, to a range of over 500 ug/m³ for Mine C. As no target limits of agreement were established for this study a priori, it is inappropriate to comment on the adequacy of the ranges. However, for comparison, the Bland-Altman limits of agreement for two inhalable dust samplers was 480 ug/m³ (upper limit: a mean difference of 0.27 mg/m³, lower limit: a mean difference of −0.21 mg/m³).^[35] In addition, the increasing variability of the differences observed in this study as the average between the paired measurements increased suggest that the use of log-transformation of RCS concentration differences is indicated.

Development of a correction factor specific to each mine through an inverse linear regression approach indirectly addresses the presence of mineral interferences in a specific dust assuming that their content is relatively consistent. While analysis of mineral interferences was not the primary scope of this research, additional minerals, identified through analysis of bulk mineral samples collected from each mine, were present. In addition to kaolinite, minerals characterized through IR absorption bands that may coincide or overlap with the crystalline silica absorption band (800 cm⁻¹) include muscovite (799 cm⁻¹), montmorillonite (797 cm⁻¹), and albite and orthoclase (788 cm⁻¹).^[23] Several of these minerals were observed in bulk sample analysis, with the highest fraction in Mines B and C. Because the calibration significantly improved the level of agreement in these mines, this finding may suggest that higher fractions of mineral interferences in the respirable dust sample increase the need for and effect of a site-specific correction factor.

The application of FT-IR technology to field samples collected in three Northwestern U.S. metal/nonmetal mines was assessed in comparison with results obtained through traditional XRD analysis. The FT-IR method resulted in mean RCS concentrations that were lower than the XRD method. The variability of the two methods was similar, and the observed level of precision was somewhat lower for the FT-IR method. Strong, positive correlations were observed between FT-IR and XRD RCS concentrations for each mine. The application of a mine site-specific calibration factor obtained through inverse linear regression significantly improved the level of agreement between the two analytical methods in two out of the three mines. A high level of agreement between both methods without any correction was noted in the mine that was not significantly improved by the correction factor. These results suggest that the FT-IR method is capable of delivering more timely RCS exposure concentrations that are comparable with those obtained through the slower, traditional XRD laboratory method.

This article demonstrates a simplified approach for adjusting FT-IR data to reduce the difference with standard XRD analysis via a mine specific correction factor. The application of this end-of-shift technology, if implemented by non-FT-IR experts such as industrial hygienists, could substantially decrease analytical delays associated with current integrated crystalline silica analyses, and thus improve the application of appropriate industrial hygiene control strategies and worker protection measures. This end-of-shift feedback of sampling results is especially impactful in mining environments where geological conditions and subsequent RCS exposures may change rapidly, however, the approach is also applicable to non-mining environments.