Method for the Analysis of Respirable Airborne Particulates on Filter Using the Mineral Liberation Analyser

Abstract

In recent years, the Mineral Liberation Analyser (MLA) has played a pivotal role in analysing respirable and inhalable ambient air samples collected on filters from both underground coal and metalliferous mines. Leveraging backscattered electron (BSE) image analysis and X-ray mineral identification, the MLA offers automated quantitative mineral characterization. The escalating prevalence and severity of mine dust lung diseases, particularly among young miners, have reignited interest in comprehensively understanding the dust’s characterization, encompassing mineralogy, particle size, and shape. Merely measuring total respirable dust exposure and its duration based on gravimetrically determined weight is no longer deemed sufficient in addressing the evolving landscape of occupational health challenges in mining environments. Since the publication of previous studies, efforts have been dedicated to refining the Mineral Liberation Analyser (MLA) methodology for respirable dust sampling. This refinement, discussed in detail in this paper, encompasses various enhancements, such as the implementation of data checks to identify carbon contamination, backscattered electron (BSE) drift, and the misclassification of X-ray spectra. Additionally, an examination of sampling efficiency led to the exploration of using smaller samples as an alternative to the time-intensive analysis of entire filters. Furthermore, this paper presents a reanalysis of paired filter sample sets previously reported using the Sarver Group Methodology. These samples are subjected to analysis using the Mineral Liberation Analyser, providing a more detailed illustration of the outputs derived from the updated methodology and compared to previously published MLA data.

1. Introduction

Since the re-identification of coal workers’ pneumoconiosis (CWP) in Queensland in 2015, the health surveillance system has been strengthened and over 421 cases of various mine dust lung diseases (MDLDs) have been diagnosed in Queensland [1,2]. The United States has also seen an increase in the prevalence and severity of the disease over the past two decades, including regional variations in prevalence rates [3,4]. These regional variations were also noted historically in the United Kingdom by the British Pneumoconiosis Field Research Unit in the Interim Standard Study that included 24 test collieries [5]. The history of the setting of the exposure standards for the UK, USA and Australia and the data this was based on were discussed in a previous paper [6].

The increase in the prevalence and severity of the disease, particularly in young miners, has placed a renewed interest in understanding the characterization of the dust, including the mineralogy, size and shape of the particles. It is no longer adequate to simply measure respirable dust exposure based on a gravimetrically determined weight. The particle size distribution of the dust present in underground coal mines is known to have become finer over time, and much of this finer dust, below 2.5 microns, can penetrate beyond the terminal bronchioles and into the gas-exchange region of the lungs [7,8].

Technologies for imaging micron-sized dust particles have become increasingly better and more cost effective over time. While characterisation studies were attempted in the 1970s and 1980s, the step change in technology now allows a more meaningful level of detail [9,10]. These previous studies were only able to calculate an overall particle size distribution of the dust and detect the presence of certain elements. Now, modern Scanning Electron Microscopes (SEMs) have very stable backscatter electron (BSE) signals and can generate high resolution particle images down to the submicron size [11]. This step change in technology allows for images of the individual particles to be taken and correlated for the detection of the numerous mineralogies present and the calculation of the particle size distributions for the various mineralogies. Several groups have undertaken characterisation work in recent years in the US, Australia and other parts of the world using various characterization methodologies [12,13,14,15,16,17,18,19,20,21,22,23,24,25]. Many of these studies utilized SEMs with energy dispersive X-ray spectroscopy (EDX or EDS) [12,14,18,20,21,22].

In the last several years, a portion of the respirable coal mine dust characterization work has moved to characterizing ambient air samples on filters due to the many dust sources present in underground mines [13,14,21,22,26,27]. For coal mines, these sources may include coal cutting methods such as the continuous miner or longwall, drilling or roof bolting activities, stone dusting (rock dusting), and dust generated by vehicle traffic, including the re-entrainment of roadway dust and diesel particulate emissions, as well as other dust-generating activities in the mine, such as shovelling. For metalliferous mines, the sources may include drilling, blasting, crushing, and vehicle traffic, among others.

The Mineral Liberation Analyser (MLA) uses a combination of BSE image analysis and X-ray mineral identification to provide automated quantitative mineral characterization [11]. The MLA was designed for liberation analysis for mineral processing to analyse samples of crushed ore encased in resin pucks. Creelman and Ward previously analysed the mineral matter in New South Wales coal using crushed coal samples encased in resin pucks. This was performed using the QEM*SEM (Quantitative Evaluation of Materials), which is another example of an SEM-based platform for automated mineralogical characterisation [28,29,30].

The first published use of the MLA for the analysis of airborne particulate matter on a filter was by Elmes for the inhalable sampling from an iron-ore mine in Brazil [31]. These principles have been adapted for respirable dust sampling in this study and applied to underground coal mines. The MLA produces maps of false colour images for all minerals showing the complexity of the particles, rather than simply assigning based on the predominant mineral. The detail of the complexity of the particles is very useful given the number of particles that are microagglomerates of several minerals.

A large study of eight underground coal mines in Australia was undertaken by the authors. Initially, samples from five mines were sent to Virginia Tech to be analysed using their methodology, as is described in previous papers [12,26]. Mine 5′s samples were too overloaded to be analysed, and therefore only the results of Mines 1–4 are reported. The results of the Sarver Group analysis of the first four Australian Mines have been previously described [13]. The analysis of Mines 6–8 using the MLA methodology was also reported in a previous paper [27].

Since those publications, work has been undertaken to refine the MLA methodology for respirable dust sampling, which builds upon the previous work and additional sampling for metalliferous mines that took place. This method was refined in a number of ways that are discussed in this paper, including data checks being implemented to identify carbon contamination, BSE drift, and X-ray spectra misclassification. The efficiency of sampling was also looked at, by taking smaller samples instead of the longer time needed to analyse the whole filter. This paper re-analyses a paired filter from the sample sets previously reported using the Sarver Group Methodology and analyses them using the Mineral Liberation Analyser to provide an example of outputs from the updated methodology [13].

2. Materials and Methods

2.1. Sample Collection Methodology

Samples were collected with Casella (SKU 116000B) or SKC (model 225-69-37) Higgins–Dewell-type plastic respirable cyclone elutriators with Zefon or SKC 37 mm polycarbonate (PC) filters with 0.4-micron pore size, all ordered through AirMet Scientific (Newstead, QLD, Australia). Depending on the location being sampled, either Casella Apex 2 (SKU APEX2IS) (Kepston, Bedford, UK) or SKC AirCheck 3000 (SKU AC3000) (Eighty Four, PA, USA) pumps were used for sample collection and operated at a flow rate of 2.2 L/m. The D50 for the Casella cyclone has been measured at 4.59 microns [32]. Samples using the SKC sampler were collected before the manufacturer confirmed an oversampling condition at 2.2 L min⁻¹, prompting an adjustment to the recommended flow rate to 3.0 L min⁻¹ [33]. After collection, the PC filters were removed from the cyclone cassettes and placed in 37 mm polystyrene cassettes, where they remained for transport and submission to the laboratory for analysis. More detail on the collection methodology can be found in the paper on the initial analysis of these samples [13].

The initial investigation gathered a total of 39 samples on polycarbonate filters, collected in triplicate at 13 locations spanning Mines 1–4 [13]. In the subsequent MLA analysis presented in this study, 10 filters from the specified locations in the table below were subjected to detailed examination, and are shown in

Table 1. Sample collection information.

Sample	Mine No.	Test Day	Sample Duration (min)	Location Code	Location
1CM_02	1	2	297	CM	Continuous Miner Development Heading
1MG_08	1	1	169	MG	Longwall Maingate hung from Shield 3
2CM_15	2	3	243	CM	Continuous Miner in Development Heading
2MF_19	2	4	251	MF	Longwall Midface Shield 76
2MG_24	2	4	270	MG	Longwall Maingate Shield 10
3CM_29	3	5	242	CM	Continuous Miner in Mains Development
3CV_26	3	5	310	CV	Mains Development Conveyor Roadway
4CM_34	4	6	*	CM	Continuous Miner
4MF_32	4	7	*	MF	Longwall Midface
4MG_36	4	7	*	MG	Longwall Maingate

* Sample duration not recorded.

.

Each sample is designated with an ID to denote the mine, sample collection location, and sample number. The first digit of the identifier contains the mine number (1–4 in this case). The two letters refer to the location: CM = Continuous miner, MG = Maingate of the Longwall, MF = Midface of the Longwall, CV = Conveyor roadway. The last two digits after the underscore are the sample number. Samples were often taken in triplicate and only one sample may have been analysed via this method [13]. All of these mines performed two-heading development for their longwall panels with homotropal ventilation.

2.2. MLA Methodology

Standard MLA samples are received as loose particulates, which are then mixed with resin to fix the particles and then sectioned and polished to provide a high-quality flat surface. However, for the airborne particulate samples, the MLA group received 37 mm filter papers from the respirable dust samples housed in polystyrene cassettes. To minimize the handling of the materials collected on the filter papers, they were simply fixed to a standard aluminium SEM stub using carbon tabs, as shown in Figure 1. Two sizes of SEM stub were used: 12.5 mm and 25 mm. For the 12.5 mm stub, a hole punch was used to cut a smaller (9 mm) diameter section out of the filter paper. Once fixed on the stub, the samples were carbon-coated to a thickness of approximately 20–25 nm to conduct excess electrons away from the sample surface.

The MLA system on which the analyses were carried out uses a ThermoFisher (formerly FEI) Quanta 650 scanning electron microscope (Hillsboro, OR, USA) equipped with a Bruker Silicon Drift Detector. The samples were analysed using the XBSE measurement mode with standard instrument set-up (25 kV accelerating voltage, 40 μA emission current, 13 mm working distance). To remove the background (i.e., filter paper), a backscattered electron (BSE) threshold was applied, and magnification was set to achieve a pixel size of 0.28 μm.

The measurement protocol involved using a mineral reference standard developed from existing standards for sedimentary lead- and zinc-containing materials (Pb-Zn) and refined to include reference standards for other minerals/phases including carbon, halites, and stainless steel. The measurements were set to time when either the number of particles measured OR the total measurement time had been reached (both of these are user-defined). This means that, depending on the particle loading on the filter paper, different areas of filter paper may be measured.

2.3. MLA Methodology Checks

The proper measurement of MLA data necessitates various checks to be performed, including assessing carbon contamination, BSE drift, and X-ray misclassification. Several measurement artefacts have been observed from the samples measured to date. Examples of these, how they can be dealt with, and implications for further measurements are outlined below.

There are two potential sources of carbon, including the collection of real carbon particles on the filter and contamination. Real carbon particles are present in the sample from material collected, typically coal dust or diesel particulate. An example of a carbon particle collected on the filter paper together with the collected X-ray spectra is shown in Figure 2.

There is potential for carbon contamination from carbon deposited during the sputter coating of the filter. Data from blank filter papers that have been carbon coated can provide a baseline for this. For instances where there are a reasonable number of particles sampled, the contributions from this will be negligible; however, if the total particle counts are low, this needs to be taken into consideration.

BSE drift was encountered during the analysis of some of the samples. When this occurs during measurement, the filter paper may be sampled and classified as shown in Figure 3. Figure 3a shows the BSE image, while Figure 3b shows the corresponding particle map. The particles, highlighted by the boundary of the red box, clearly show that the filter paper is being sampled.

Misclassification of collected X-ray spectra was also observed. This is due to variations in peak heights and the shadowing effect it creates, and can result in carbon being over-represented. An example of this is shown in Figure 4, where the X-ray spectra are misclassified as carbon when other peaks are clearly visible.

Misclassification of X-ray spectra may occur due to edge effects and creases in the filter paper. Figure 5a depicts the edge effect created around a 9 mm sample, while Figure 5b illustrates the measurement artifact created by a crease in the filter paper. Particles are highlighted by a red boundary box to make them more clearly visible.

Furthermore, the overloading of the particles on the filters can lead to errors in the particle size information being generated, as distinguishing between the various particles can become challenging. For example, Mine 5 filters were not able to be accurately analysed due to overloading. This may be caused by sampling the dust load present in the air for too long. When a filter is overloaded, the MLA cannot distinguish between the discreet particles. Figure 6 illustrates a heavily loaded filter paper.

Over the course of this work, a number of ways of measuring a smaller region on the filter paper were trialled. This is primarily because, due to high magnification, the measurement of the entire area takes many hours, which is not a practical approach in many cases as measurement time will impact analysis costs and turnaround times. As noted previously, two sizes of SEM stubs were used. For the smaller stub, a sub-sample was cut randomly from the filter paper using a 9 mm hole punch. However, the geometry of the hole punch restricts the potential sample zone on the filter paper such that the centre of the filter paper cannot be sampled. Several examples of the imaging results from the smaller stubs are shown in Figure 7. In some cases, due to the high level of particle loading, the entire area of the filter paper sub-sample is not measured.

For the 25 mm SEM stub, if there was sufficient particle loading, several regions were selected across the area of the stub, typically three, but in some cases nine, as shown in Figure 8.

Figure 9 is an example of the variation in sample loading across two filter papers, showing three locations each on filters Figure 9a,b. There is no reference point on the filter papers, and as such they are randomly oriented for measurement.

3. Results

3.1. Resulting Improvements to MLA Methodology

As a result of the analysis undertaken on the MLA to date, a number of modifications have been made to the procedure to more easily and accurately analyse airborne particulates on filters. The classification script has been modified to include an increased threshold for low counts, a BSE range limit has been set for carbon, and a tighter pattern match criterion has been set for X-ray classification.

Checks should be performed on the data post measurement and classification for quality assurance purposes. The classified image should be checked by zooming out so that the entire measured area is viewable, and selecting all particles measured, this check can be used to assess BSE drift, edge effects, creases, and overloading. The modal mineralogy should be checked in conjunction with the classified image, to assess % carbon, % unknown, % stainless steel, or any other materials that seem spurious or inconsistent with what is known about the sample.

3.2. Sample Results

The mineral reference library employed to classify the samples includes 42 classes of mineralogy, including “other” and “unknown”.

Table 2. Mineralogy classes of samples from MLA analysis (weight percentage). CM = Continuous Miner, MG = Maingate of the Longwall, MF = Midface of the Longwall, CV = Conveyor Roadway.

	Mine 1		Mine 2			Mine 3		Mine 4
Mineral	1CM_02	1MG_08	2CM_15	2MF_19	2MG_24	3CM_29	3CV_26	4CM_34	4MF_32	4MG_36
Native Copper	0	0	0	0.03	0	0	0	0	0	0
Chalcopyrite	0.02	0	0	0.01	0	0	0	0	0	0.02
Bornite	0	0	0	0	0	0	0	0	0	0
Enargite	0	0	0	0	0	0	0	0	0.01	0.01
Copper Sulphate	0	0	0	0.01	0	0	0	0	0.03	0.04
Pyrite	0.01	0.01	0	1.34	0	1.84	2.97	0.09	1.76	1.53
Pyrrhotite	0	0	0	0.11	0	0	0	0	0.03	0.03
Arsenopyrite	0	0	0	0	0	0	0	0	0	0
Galena	0	0	0	0.9	0	0	0.02	0	0.05	0.85
Sphalerite	0	0	0	0.57	0	0	0	0.01	0.01	0.57
Cobaltite	0	0	0	0	0	0	0	0	0	0
Chlorite	8.11	7.87	3.5	0.2	3.71	1.01	0.78	0.86	0.59	1.19
Muscovite	8.22	9.35	6.72	7.75	7.08	29.29	15.4	12.46	51.56	25.13
Biotite	0	0	0	0.55	0	0.99	1.48	0.17	0.49	0.69
Amphibole	0.29	0.24	0.15	6.01	0.24	0.94	9.99	2.27	0.67	2.42
Quartz	0.28	0.29	0.16	0.85	0.34	10.85	11.6	20.73	2.7	15.37
Plagioclase	2.6	4.07	2.8	0.28	3.34	3.85	6.31	2.96	0.52	3.5
Orthoclase	0.21	0.6	0.28	1.32	0.25	39.51	24.25	47.51	15.55	21.6
Kaolinite	0.43	0.62	0.55	1.02	0.51	3.22	2.34	4.27	3.58	2.79
Zircon	0	0	0	0	0	0.01	0	0.01	0.01	0.02
Pyroxene	0.04	0.01	0	0.02	0.02	0.01	0	0	0	0
Calcium Silicate	7.19	3.6	7.68	5.56	10.41	0.71	5.61	1.05	0.88	2.61
FeOxide	0	0	0	0.78	0	0.06	0.16	0.09	0.34	0.75
Rutile	0	0	0	0.01	0	0	0.01	0	0	0
Ilmenite	0.28	0.33	0.16	0.13	0.21	0.08	0.12	0.13	0.6	1.34
Stainless Steel	0.76	0.69	0.51	1.6	1.22	0.02	0.03	0.06	0.44	0.72
Magnesite	0	0	0	0	0	0	0	0	0	0
Calcite	10.84	5.19	6.21	5.87	10.15	0.31	8.65	0.46	0.43	4.81
Dolomite	0	0	0	0.51	0	0	0	0	0.03	0.25
Gypsum	0.2	0.06	0.01	0.15	0.07	0.02	0.01	0.62	0.02	1.51
Barite	0	0	0	0.11	0	0.01	0	0	0.02	0.11
Jarosite	0	0	0	1.17	0	1.4	0.74	0.64	3.57	1.79
Sylvite	0	0	0	0	0	0	0	0	0	0
Halite	0.05	0.01	0.03	0.5	0.24	0	0	0	0.02	0
Apatite	0.43	0.38	0.72	0.37	0.86	0.02	0.05	0.01	0.16	0.05
Carbon	59.78	66.61	70.47	61.36	61.18	4.9	8.31	4.28	15.02	8.68
Other	0	0	0	0.22	0	0.24	0.39	0.46	0.19	0.27
Tramp	0	0	0	0	0	0	0	0	0.01	0.25
Unknown	0.14	0.03	0.02	0.69	0.12	0.69	0.78	0.86	0.7	1.08
Low_Counts	0	0	0	0	0	0	0	0	0	0
No_XRay	0	0	0	0	0	0	0	0	0	0
Clinochlore	0.04	0.04	0.01	-	0.04	-	-	-	-	-
MgOxide	0.01	0	0	-	0	-	-	-	-	-
AlOxide	0.06	0	0	-	0	-	-	-	-	-

shows the mineralogy classes of the samples analysed and the weight percentage of each. Figure 10 shows the mineralogies over 2% by weight to illustrate the predominant makeup of the samples and how they differ between mines and locations.

Mines 1 and 2 exhibited high levels of carbon ranging from 60% to 70%, while Mines 3 and 4 contained lower amounts of carbon, ranging from 4% to 15%. The samples from Mines 3 and 4 contained significantly more quartz, with less than 1% identified in Mines 1 and 2 compared to 3%–21% in Mines 3 and 4. Furthermore, a large amount of muscovite (12%–52%) and orthoclase (16%–48%) were found in Mines 3 and 4, both of which are known to contain potassium, a correlate for cell death in lung tissue studies [34].

Although present in the mineral reference library, bornite, arsenopyrite, cobaltite, magnesite, or sylvite were not found in these samples. The “Other” mineralogy type averages 0.10% of samples with a maximum of 0.46% for sample 4CM_34. Moreover, there was an average of 0.32% “Unknown” mineralogy with a range of 0.02 to 1.08% (4MG_36). Finally, small amounts of zircon were found in five of the samples, with a maximum of 0.02%.

The overall particle size distributions of the samples are shown in Figure 11, and were calculated based on equivalent circle diameter. The equivalent circle diameter is the diameter of a hypothetical sphere that shares the same specific area as the particle [35]. Lines closer to the top left are finer particle size distributions (PSDs), while lines near the bottom right are coarser PSDs. Mine 4 has the finest particle size distribution, where 50% of the particles (P50) of the maingate sample (4MG_36) are less than 2.8 microns; contrast that to Mine 2 (2MG_24), where the P50 is closer to 4.8 microns.

The data generated by the MLA can be further split into the PSDs of the various mineralogies. The MLA software v.8 is able to analyse the individual dust particles and determine the PSD of each mineral class. The MLA software v.8 defaults to calculating the PSD of all particles that contain any amount of a given mineral. However, for the purposes of this research, the data have been filtered to include only particles who have >50% by weight of a certain mineral, thereby calculating the PSD for particles predominately composed of that mineral. Nonetheless, it should be noted that some particles will not be included in any PSD apart from the overall PSD if there is no dominant mineral above 50% weight, such as particles containing three minerals of roughly equal weight percentages.

Figure 12 presents the PSDs of the particles containing greater than 50% quartz, where the mine data follow a similar pattern of PSDs compared to the overall samples. Notably, Mine 4 exhibits the finest particle size distribution, where 50% of the particles (P50) of the maingate sample (4MG_36) have an equivalent circle diameter of less than 2.1 microns, in contrast to Mine 2 (2MG_24), where the P50 is closer to 3.5 microns.

Furthermore, the PSDs of the predominant mineralogies of a sample can visualize the range of PSDs. Figure 13 illustrates the PSD of the Mine 2 maingate sample (2MG_24) by mineralogy. As this sample primarily contains carbon, the coarser carbon particles heavily weight the overall PSD. In this instance, the quartz is the finest PSD, with a P50 of 3.5 microns compared to the overall P50 of 4.8. Meanwhile, the carbon has the largest P50 of 5.1 microns.

Moreover, Figure 14 displays the PSDs of the predominant mineralogies of sample 4MG_36, where the overall PSD is in the middle of the range, with a P50 of 2.8. Orthoclase has the lowest P50, of 2.2 microns, while Kaolinite has the highest P50, of 3.5 microns. Although the PSDs of the individual mineral classes exhibit a slight variation, they remain relatively consistent with the overall PSD, indicating that this sample has a narrower range of PSDs.

In this study, mineralogy mapping of individual particles was carried out using false-colour images provided by the MLA software v.8. The colour legend used in the mapping process is shown below.

Two sets of particle images obtained from the Mine 2 maingate sample, 2MG_24, and the Mine 4 sample, 4MG_36, are shown below in Figure 15 and Figure 16. The former set exhibits a significant presence of carbon, indicated by the black colour, while many of the particles are irregularly shaped and/or elongated. In contrast, the latter set comprises smaller and rounder particles, albeit with a few elongated ones. Furthermore, although microagglomerated particles are still present in the Mine 4 sample, a considerable proportion of the particles appear to be of a single mineralogy. See the next figure for a colour legend for the mineralogy of particles.

4. Discussion

In this study, we refine the application of the MLA for the analysis of respirable airborne particulate matter on a filter, as first introduced by Elmes in the context of inhalable sampling [30]. Unlike more traditional methods of SEM analysis [12] that assign a single mineralogy, based on the predominant mineral present, the MLA generates false colour maps that illustrate the complexity of the particles considering all minerals present. This approach proves valuable in capturing the intricate nature of particles composed of multiple mineralogies.

Our research entails a comprehensive investigation of eight underground coal mines in Australia. Initially, samples from Mines 1–4 were sent to Virginia Tech to be analysed by the Sarver Group methodology [13]. Samples from Mines 6–8 were analysed at the University of Queensland, Sustainable Minerals Institute, using the MLA methodology [27].

In this paper, we reexamine paired filter samples from Mines 1–4 [13], now subjecting them to a more comprehensive MLA evaluation. Moreover, during this interim period, we have made efforts to refine the MLA methodology and improve accuracy and repeatability for respirable dust sampling, building upon previous work and incorporating learnings from additional sampling. Given the prevalence of microagglomerates in these particles, the level of detail provided by the MLA methodology proves highly advantageous.

4.1. MLA Methodology

The analysis of airborne particulates on a filter using the MLA method has undergone changes to improve its accuracy and ease of use. The modifications include increasing the threshold for low counts in the classification script, setting a BSE range limit for carbon, and tightening the pattern match criterion for X-ray classification.

4.1.1. Low Counts

Further research is needed to determine the optimal measurement area of a sample. This includes understanding the confidence limits for particles counted and area measured, such as the position on the filter paper where the measurement was taken. Once the implications of this are understood, then a standard approach should be taken so that measurement times can be optimized. This may include the measurement of a standard number of frames on the filter paper in a ‘standard’ area or randomized across the paper. Striking a balance between capturing sufficient data for accurate dust quantification and minimizing the utilization of costly machine time is crucial in this optimization process.

The proposed research for the standard statistical approach can then be used to assign significance to the data, especially for low particle counts. Particle counts will then be an indicator of particle loading on the filter paper using a standardized approach with a standardized measurement area. In the previously employed Sarver methodology, 1400 particles were analysed across the filter paper, which proved adequate for categorizing particles into eight discrete mineralogies [12]. However, the MLA analysis introduces increased complexity, identifying 42 mineralogies and offering the ability to recognize multiple mineralogies within a single particle. This complexity necessitates a more comprehensive study to assess the variability in analysis and establish a robust protocol to ensure accuracy and reliability in the face of the expanded mineralogical scope.

It may also be possible to correlate the collection of samples with real-time monitoring to ensure that over/underloading is avoided in the field as much as possible. The overloading of the samples is a common problem which has occurred in several sampling campaigns to date, including for the Mine 5 samples in this campaign.

4.1.2. Data Checks

To ensure quality assurance, checks should be performed on the data post measurement and classification. The classified image should be examined by zooming out so that the entire measured area is viewable and selecting all particles measured. This check can be used to assess BSE drift, edge effects, creases, and overloading.

The modal mineralogy should also be checked to assess the percentage of carbon, unknown mineralogy, stainless steel, or any other materials that appear inconsistent with what is known about the sample’s composition. Sputter coating of the filter in carbon is performed to prevent electrical charging. The carbon contamination from the carbon coating was found to be negligible if there were enough particles present. If there are processes being sampled that have very low numbers of particles where this may be an issue, other types of coating may need to be considered, such as gold or palladium.

An additional strategy to mitigate creasing and edge effects involves the thoughtful selection of filters. In this study, we utilized filters from both Zefon and SKC. The SKC filters were characterized by a single-sided design, featuring a shiny side intended for particle deposition and a dull side opposite to it. In contrast, the Zefon filters employed were ‘double-sided’, indicating that both surfaces were shiny and suitable for analysis. This inherent characteristic renders Zefon filters thicker and less prone to wrinkling, offering enhanced stability during the analytical process.

4.1.3. Mineral Reference Library

The mineral reference library for the MLA undergoes continuous updates as other bodies of work are completed. During the intervals between filter analysis for the various mines, distinctions for clinochlore, magnesium oxide, and aluminium oxide were integrated into the library. The addition of these additional mineralogies were negligible to the analysis of these coal samples, but illustrate the potential for additional mineralogies. To enhance precision within specific mineralogical groups, the possibility of introducing dedicated reference samples has been considered. Notably, this approach has been recommended for the muscovite group, given the suspicion that a significant portion of this mineralogy may align with the illite group within.

4.2. Mine 1–4 MLA Results

The examination of airborne dust samples from four distinct underground coal mines has unveiled noteworthy disparities in mineralogical content and particle size distribution. Specifically, Mines 1 and 2 exhibited elevated carbon levels, while Mines 3 and 4 displayed lower carbon content and a prevalence of quartz, muscovite, and orthoclase. Geological factors emerged as more influential than spatial location within a mine in shaping particle size distribution. These variations in mineralogical content and particle size have vital implications for understanding the potential health hazards associated with exposure to mine dust.

Discrepancies between the MLA results and the prior Sarver method outcomes are evident [13]. Notably, in the MLA analysis, sample 2MG_24 reveals a carbon content of 61.18%, while its paired counterpart, sample 2MG_23, exhibits a higher carbonaceous material content at 75.7%. Conversely, the calcite content in the samples is quite similar, with the MLA analysis indicating 10.14%, and the carbonates content measured through the Sarver group methodology registering at 10.86%.

Figure 15 offers insights into potential factors contributing to observed discrepancies. The bright blue highlighting calcite indicates its prevalence as either the dominant constituent in the majority of particles or a minute portion, suggesting a closer alignment with the assignment of a single mineralogy. In contrast, the false colour images reveal small yet noteworthy portions of aluminosilicates in particles primarily characterized by carbon. This observation may elucidate why the Sarver group methodology identified only 6.36% aluminosilicates as the predominant mineralogy. When considering the cumulative instances of these occurrences, muscovite alone constitutes 7.08% of this sample, while plagioclase accounts for 3.34% in the MLA analysis. These findings underscore the complexity of mineralogical interpretations and highlight the importance of comprehensive analytical approaches for the accurate characterization of samples.

Recent investigations indicate a correlation between potassium levels and cytotoxicity, particularly in lung tissue exposed to crushed coal samples [33]. To ascertain whether dust exposures akin to those in Mines 3 and 4, characterized by higher potassium levels and finer particles, pose a greater health risk to workers than dust from Mines 1 and 2, which had minimal potassium, further research into lung tissue studies is imperative. Potassium levels in the mineralogies were identified as 9.8% for the muscovite group, 14.04% for orthoclase, and 52.45% for sylvite based on the mineral reference library.

Comparatively, the MLA results for Mines 6–8 previously published revealed similarities between the amount of carbon in the continuous miner sections of Mines 1 and 2 and those in Mines 6 and 7. Notably, Mines 6 and 7 exhibited higher silica levels compared to Mines 2 and 3. Conversely, Mine 8 stood out, with exceptionally high carbon content, distinguishing it from all other mines. Mines 3 and 4 shared similarities, particularly in their substantial amounts of quartz, muscovite, and orthoclase, along with minimal carbon.

Despite accounting for sampling effects from a cyclone elutriator, the particle size distributions of the samples displayed persistent variability. For instance, in sample 2MG_24, particles were coarser, with P50s ranging from 3.5 to 5.1 microns. In contrast, sample 4MG_36 exhibited finer particles, with P50s ranging from 2.2 to 3.5 microns. Although finer particles may represent less mass, they have a higher likelihood of penetrating deeper into the lungs, posing a greater health risk [8]. Additionally, the abundance of silica and aluminosilicates in the dust samples may exacerbate potential health hazards associated with mining dust exposure. Research has shown that for pure silica, finer particles (0.3 microns) induce greater macrophage activation than coarser silica particles (4 microns), and the rise in disease prevalence in the US has been linked to silica and silicates through lung pathology [35,36].

This study underscores the non-uniform nature of coal mine dust, revealing variations not only between different mines but also within individual mines, contingent upon factors such as location and the type of cutting equipment utilized. The findings challenge the suitability of a singular exposure standard for coal mine dust. Certain mines may exhibit elevated concentrations of smaller particles (<2.5 microns), capable of deeper lung penetration despite having minimal mass, while others may have an abundance of larger particles (approaching 10 microns) possessing substantial mass but a limited ability to penetrate the alveoli. Consequently, there is a need to prioritize correlating dust exposures with specific health hazards, advocating for a nuanced approach tailored to the unique characteristics of each mining environment, rather than adopting a uniform exposure limit.

5. Conclusions

The characterization of mine dusts is crucial for understanding the potential health implications of exposure. Variations in mineralogy and particle size distributions among different sources of dust within a mine may result in variable health hazards. Comprehensive characterization studies such as this can aid in establishing the true exposures of workers, and furthering our understanding of the deposition and toxicity of mine dust.

The findings of this study indicate that the gravimetric mass measurement of dust used in compliance sampling may oversimplify the health hazard posed by dust. The mineralogy, size, and shape of particles affect their potential to penetrate deep into the lungs and cause harm. Moreover, the mineralogical content of particles can impact the body’s ability to clear them.

The MLA methodology used in this study provides a detailed analysis of the characteristics of respirable dust, which is particularly beneficial for analysing microagglomerated particles. The modifications made to the method, as presented in this paper, will lead to more consistent analysis. This study demonstrates the potential of the MLA as a tool for understanding the characterization of dust in airborne particulate matter in the respirable fraction, and it could be applied to other industries to gain insights into their exposures.