Comparison of the Concentrations of Heavy Metals in PM2.5 Analyzed in Three Different Global Research Institutions Using X-ray Fluorescence

Abstract

This inter-lab study aimed to evaluate the comparability of heavy metal concentrations in the same samples using three X-ray fluorescence spectrometers (XRFs) in three different global re-search institutions, namely a collaboration lab between Soonchunhyang University (Asan, Korea). and PAN (a branch of Malvern PANalytical, Seoul, Korea), RTI (Research Triangle Institute, NC, U.S.A), and Aerosol laboratory in Harvard University, Boston, U.S.A. Indoor air filter samples were collected from 8 homes using 3 filters in each household (n = 24) of individuals with asthma, and the same filter samples were sequentially analyzed separately in the collaboration lab Soonchunhyang-PAN, Harvard University, and RTI. Results showed the detection rates of most heavy metals (n = 25 metals) across the three institutions to be approximately 90%. Of the 25 metals, 16 showed coefficient of determination (R²) 0.7 or higher (10 components had 0.9 or higher) implying high correlation among institutions. Therefore, this study demonstrated XRF as a useful device, ensuring reproducibility and compatibility in the measurement of heavy metals in PM2.5, collected from indoor air filters of asthmatics’ residents.

1. Introduction

Among air pollutants, fine dust generated from various sources, including vehicles and combustion processes, is a major cause of air quality deterioration and contains various heavy metals and harmful pollutants [1].

Fine particulate matter refers generally to all particles with aerodynamic diameter less than 2.5 μm (PM2.5), and its effect on human health depends on the size and/or composition of the fine particles [2,3].

NAAQS (National Ambient Air Quality Standards) and USEPA (United States Environmental Protection Agency) have set a limit value in ambient air for heavy metals such as lead (0.15 μg/m³). According to Integrated Risk Information System (IRIS), National Center for Environmental Assessment (Washington, DC, USA) has set a accepted limit value at 0.2 ng/m³ for cadmium because of its health effects such as bronchial and pulmonary irritation as well as increased frequency of kidney stone formation [4]. In addition, it has been reported that higher concentration of chromium in ambient air can result in ulcerations of the septum, bronchitis, and decreased pulmonary function [5].

Owing to such harmful effects on human health, Korea and other countries have established and managed air quality standards for PM2.5 since 1990 [6]. Outdoor air quality containing PM2.5 had been declared a Group I human carcinogen by the World Health Organization (WHO) in 2013.

A growing number of studies in epidemiology, toxicology, and other related fields have shown that exposure to PM2.5 is closely related to the incidence of a wide range of human conditions and mortality rate. The pathological toxicity as well as chemical properties of PM2.5 are known to trigger a wide range of respiratory diseases [7]. Inhaled PM2.5 can be deposited in many parts of the respiratory tract, where it interacts with immune cells and epithelial cells which trigger systemic inflammatory responses [8]. In addition, the positive relationship between cardiovascular mortality and the exposure to PM25 has been proved in many large time-series and case-crossover studies [9].

Emerging epidemiological evidences have shown that PM2.5 is associated with dementia [1], various cognitive disorders, Alzheimer’s disease, and Parkinson’s disease [10,11,12]. Moreover, epidemiological studies demonstrated that ambient air pollution may be detrimental on cognitive function especially in aging populations. According to various studies, PM2.5 exposure may play a substantiate role on the onset of neurodegeneration through processes such as systemic inflammation, oxidative stress, and neuro-inflammation among others [13]. In 2015, the dementia mortality rate due to ambient PM2.5 was approximately 15% (11–19%) of the total mortality due to ambient PM2.5 pollution [14]. With a hazard ratio of 1.55 for a 1 μg/m³ increase in PM2.5 (95% confidence interval (CI): 1.00 ± 2.41, p-value 0.05), PM2.5 emissions have been associated with dementia, and air pollution and traffic jam-related pollution may be important independent risk factors for dementia [15].

The type of disease caused by metal compounds depends on the nature of the element, dose, physicochemical form, host factor, and exposure conditions among others. Metallic dusts that are inhaled and deposited in the lung may trigger the pulmonary fibrosis as well as functional impairment, depending on the fibrogenic potential of the agent. In addition, inhalation of iron compounds causes pneumoconiosis and siderosis [16]. Epidemiological studies have demonstrated the relationship between exposure to lead with neurodegeneration in cross-sectional human studies [17]. A study conducted in Korea has shown the statistically significant correlation between cadmium, lead, and mercury concentration in the blood with cognitive degradation [18].

Evidence on the potential effect of environmental exposure to heavy metals in cardiovascular diseases prevalence has increased over the past two decades [19]. Both environmental and occupational exposures to inorganic arsenic have been related to an increased cardiovascular mortality [20], and clinical studies demonstrated arsenic induced hypertension atherosclerosis, stroke, coronary heart disease, and diabetes mellitus in a dose-dependent manner [21]. Additionally, the occupational exposure limit for lead in air set out in the Regulations is 0.15 μg/m³, and a concentration above that level may have been shown to affect blood which may lead to anemia, and affect the nervous system and kidney, as well as infertility in males [22]. Cadmium, being carcinogenic in humans, had an exposure limit measured over a 15 min period and has been set at 0.05 mg/m³ of air [23]. In addition, the exposure to elements such as, Cd, Hg, and Pb is statistically associated with cardiovascular diseases [21].

Nickel exposure limit was set at 0.5 mg/m³ averaged over an 8 h period [24] and its exposure above the accepted level has shown significant effect on heart rate variability [25], and increased concentrations of nickel, zinc, and lead have been reported to be associated with an increase in emergency department (ED) visits for cerebral hemorrhage in China [26].

Several published papers have suggested that PM2.5 is associated with cardiovascular or respiratory diseases, and dementia [27]; still level or component of indoor PM2.5 components affecting health remains insufficient.

Due to such vital importance of PM2.5 components, recently, studies on heavy metals in PM2.5 have been actively conducted to identify their sources. Inductively Coupled Plasma (ICP) Spectrometer and X-Ray Fluorescence Spectrometer (XRF) have been used to analyze the heavy metals [28,29,30]. Although ICP has a lower detection limit which ranges from <1 ppm to >100 ppm [31], and XRF detection limits for most elements are 2–20 ng/cm² for micro sample surface, thin samples, aerosols, and liquids [32], both XRF and ICP devices have comparable element detection capacities for ambient samples, with lower measurement error. Both techniques are capable of giving excellent accuracy and precision [33]. However, ICP requires relatively sophisticated pre-sample treatment techniques using ferric acid with hydrochloric acid [34], and ICP is often a destructive analytical procedure, meaning that the sample cannot be recovered after analysis, whereas XRF is a completely non-destructive method of testing that provides accurate materials analysis, and the same sample can be used for other analyses with a short sample preparation time, thus more advantageous in the analysis [35].

According to Niu et al. [36], in a two-week study that assessed the effect of heavy metals in PM2.5 on human health using ICP-MS and XRF, Fe, Mn, Zn, Pb, and Cu, these were found to show correlation coefficients of 0.7 or more. For 24 h filter samples studied with Inductively Coupled Plasma Mass Spectrometry (ICP-MS), the detection limit was low; therefore, the number of elements that passed the quality criteria was high. By default, direct comparison of the two different techniques, ICP-MS, and Energy Dispersive X-ray Fluorescence (ED-XRF), showed good correlation for both 24-hour and 2-week filter samples [36]. Results of previous studies had shown that ICP is beneficial for the precise analysis of low-concentration samples as low as 1.3 ng/m³ [36], although the sample preparation process was relatively longer. XRF data, on the other hand, can be analyzed in various ways without sample preparation, suggesting the analysis to be relatively faster and reproducible.

XRF can sufficiently measure heavy metals in the atmosphere with samples collected at a low flow rate (0.5 L/min). Most heavy metals can be analyzed as long as the PM mass in the filter is above a certain level because low sample mass affect the reliability of metal analysis for both XRF and ICP-MS, therefore the 0.06 mg minimum cut-off for particle mass ensures reliable analysis results for most elements using ICP-MS [37]. This study aimed at evaluating the comparability of heavy metal concentrations contained in PM2.5 in the same samples using XRF. In this study, three different global analyst institutions, namely the branch of the PANalytical Institute in Seoul collaborated with Soonchunhyang University, (Co-lab of SCHU-PAN), Korea, the Aerosol Institute of Harvard Chan School of Public Health in Massachusetts, USA, and the Ultrafine Dust Measurement and Commercial Analysis Laboratory of RTI in North Carolina, assessed the correlation across the same samples, determined the correlation coefficient across the values obtained, calculated the determination coefficients between institutions, and provided information on the reliability of XRF results across the institutions.

2. Materials and Methods

2.1. Study Participants and Sampling

From November 2017 to February 2018, three PM2.5 filter samples were collected from the indoor area (living room) from 8 homes of individuals with asthma, recruited through the Department of Respiratory Internal Medicine, Soonchunhyang University Bucheon Hospital and Pediatrics Department of Inha University Hospital; heavy metals contained in the PM2.5 samples were analyzed thoroughly. PM2.5 samples were collected at a flow rate of 0.5 L/min using MicroPEM (RTI, Research Triangle Park, NC, USA). The device contained a replaceable impactor stage that provided PM2.5 or PM10 aerodynamic cut-points, consistent with the U.S. Environmental Protection Agency (EPA)-defined particle size system [38]. The MicroPEM collects both real-time PM mass concentration data and gold-standard integrated filter. Real-time PM concentration is measured via light scattering nephelometer over a range from 3 to 15,000 μg/m³. The MicroPEM collects PM on a 25 mm PTFE filter for gravimetric and speciation analyses.

After installing a replaceable PTFE filter for repeated collection of PM2.5 samples, the collected PM2.5 amount on filter was measured in Soonchunhyang University using a 7-digit (0.1 µg) microbalance in a chamber sustaining a temperature of 20 °C and relative humidity of 50%. Then, metals of filter samples were first analyzed using XRFin PAN and Soonchunhyang University collaboration lab (Co-lab SCHU-PAN), and then in Aerosol laboratory in Harvard University, Boston, USA. Finally, the filters were analyzed by RTI. For this consecutive analysis among 3 instistitutions, the same filter samples were directly transported to Harvard University and RTI with 4 blank samples. Blind analysis was performed at Harvard School of Public Health as well as at RTI, a manufacturer of MicroPEM, for component analysis using XRF. The initial and final filter weights for PM2.5 samples and blank filters were further determined again by RTI with a 7-digit (0.1 µg) microbalance.

2.2. Heavy Metal Content Analysis Using XRF

XRF Analysis

Three institutions (Co-lab SCHU-PAN: Institute #1, RTI: Institute #2, Harvard: Institute #3) analyzed the heavy metals using energy dispersive X-ray Fluorescence Spectrometer for PM2.5 collected in a Teflon filter.

We selected a total of 25 elements which were commonly detected in all three institutes and had low detection limit (LOD): aluminum (Al), arsenic (As), barium (Ba), bromine (Br), calcium (Ca), cadmium (Cd), chlorine (Cl), chromium (Cr), cesium (Cs), copper (Cu), iron (Fe), potassium (K), manganese (Mn), nickel (Ni), lead (Pb), rubidium (Rb), sulfur (S), antimony (Sb), selenium (Se), silicon (Si), tin (Sn), strontium (Sr), titanium (Ti), vanadium (V), and zinc (Zn) (Figure 1).

Institute #1 analyzed the elemental components of heavy metals using ED-XRF (Epsilon 4, Malvern PANalytical, Almerow, The Netherlands) among standalone XRFs, and QCQA was conducted with the calibration standard film (MicroMatter Technologies Inc., Vancouver, BC, Canada). Epsilon 4, Malvern PANalytical, which is a new high-performance benchtop analytical tool for determining the chemical composition of any material throughout the analytical range, provides quick findings with low detection limits. It is also calibrated with reference materials that match the routine sample’s composition. In case of non-metallic components depositing on the metal gas filter, compound gases, such as fluoride, sulfide, iodide, and chloride were deposited on the filter, and qualitatively and quantitatively impregnated for each component by Canada MicroMatter (MicroMatter Technologies Inc. In 2017; Available online: https://www.micromatter.com/ (accessed on 23 April 2022)). It was adjusted using the standard filter [39].

The lower limit of detection (LLD) calculation of institute #1 used the EPA IO-3.3 method, which was developed and applied to XRF to analyze ambient and aerosol samples using an energy and wavelength dispersion spectrometer. It was able to sample inhalable atmospheric particulate matter (<10 μm) with a dichotomous sampler, and analyzed specific metals by X-ray fluorescence [40].

Institute #2 XRF analysis was performed on a Thermo ARL Quant’X EDXRF instrument (Waltham, MA, USA) using 6 different excitation conditions under atmospheric vacuum to enhance the response of the light elements. It provides low-volume, high-volume, and trace element quantification for a wide range of samples including bulk solids, thin films, and liquids. It is also ideal for elemental analysis. The output of power ranging in the 6 excitation conditions is 4–50 kV to achieve maximum response for the 33 project-specific elements. A calibration curve was developed using 25 mm element thin film calibration standards. The certified standards were purchased from MicroMatter Inc. (Vancouver, BC, Canada), and are stored away from light in ambient temperatures. With each analytical run, a certified multi-element thin film standard is analyzed to verify the instrument functionality and stability across the 6 excitation conditions. The QC acceptance for the multi-element standard is less than 5% coefficient of variation and a recovery of 90–110%. Additionally, replicate samples were analyzed at a rate of 10% of the total number of samples.

The elemental analysis for institute #3 was conducted using an Epsilon 5 EDXRF spectrometer (PANalytical, The Netherlands) which utilizes secondary excitation from 10 secondary selectable targets. Epsilon 5 ED-XRF spectrometer provides easy sample preparation, rapid sample throughput, and Epsilon 5 EDXRF spectrometer is capable of the required detection limits which is an important factor for the analysis. The spectrometer employs a 600 W dual (scandium/tungsten, Sc/W) anode X-ray tube, a 100 kV generator, and a solid state germanium (Ge) detector. A total of 48 MicroMatter XRF calibration standard polycarbonate films (MicroMatter Co., Vancouver, BC, Canada) were used for calibration of 48 elements ranging in atomic number from 11 (Na) to 82 (Pb). We also used the U.S. National Institute of Standards and Technology (NIST) standard reference material (SRM) 2783, which is to simulate ambient PM2.5 on filter media, for quality control of the analytical procedure.

2.3. Data Analysis

Descriptive analysis was performed to understand the concentration distribution of PM2.5 heavy metals (ng/mg, ppm). The measurement distribution was evaluated for normality using the Shapiro–Wilks test followed by a nonparametric analysis. We could derive the Spearman’s correlation coefficient, and finally, the corresponding slope and coefficient of determination to evaluate the association of concentration distribution across institutions (Institute #1–Institute #2, Institute #1–Institute #3, and Institute #3–Institute #2) in the regression model. R (ver. 4.0.3) was used for statistical analysis.

3. Results

3.1. Descriptive Analysis

The mass of 25 common elements in PM2.5 samples, collected from 8 households, was determined in three institutes, and their detection rate and median (quartile) concentrations (ng/mg, ppm) are shown in

Table 1. Median (IQR) concentration (ng/mg, ppm) of measured elements across 3 institutes.

		Institute #1				Institute #2				Institute #3
	LOD	(n = 24 Filters)				(n = 24 Filters)				(n = 24 Filters)
		Detect (%)	Median	P25	P75	Detect (%)	Median	P25	P75	Detect (%)	Median	P25	P75
Al	8.28	100.0	12,621.0	7701.2	17,176.9	100.0	3467.6	2103.8	5514.3	100.0	2898.4	1826.3	4138.6
As	3.69	100.0	21.6	16.6	39.9	100.0	19.8	10.4	66.5	25.0	ND	ND	2.1
Ba	2.12	100.0	257.2	166.2	346.8	100.0	550.1	379.3	813.8	16.7	ND	ND	ND
Br	0.20	87.5	109.5	47.4	191.7	100.0	241.8	158.2	353.6	100.0	479.5	298.7	570.2
Ca	0.50	100.0	830.9	704.3	1587.8	100.0	2749.9	1849.9	3817.6	100.0	1751.3	1150.9	2336.5
Cd	3.23	95.8	83.5	47.6	113.8	100.0	517.3	299.4	659.1	75.0	ND	ND	3.4
Cl	1.04	100.0	3499.6	2174.5	4167.9	100.0	1448.2	650.8	1921.9	100.0	628.2	351.3	796.6
Cr	0.12	100.0	121.3	73.3	161.7	100.0	107.1	43.3	147.6	100.0	56.2	25.5	92.4
Cs	ND	100.0	37.1	27.5	56.9	100.0	225.0	138.5	337.5	37.5	ND	ND	50.7
Cu	0.28	100.0	112.6	82.5	301.2	100.0	397.0	283.5	768.3	100.0	409.4	219.4	585.8
Fe	0.45	100.0	2297.8	1442.5	4364.2	100.0	6243.0	3207.5	10,202.8	100.0	3685.4	2074.5	6479.4
K	2.07	100.0	4603.6	2997.7	7092.6	100.0	7939.0	4939.6	12,151.4	100.0	4221.2	2741.7	6522.3
Mn	0.32	100.0	181.1	96.5	263.6	100.0	457.4	258.5	769.5	100.0	234.2	113.1	367.1
Ni	0.6	12.5	ND	ND	ND	100.0	157.0	99.4	264.6	100.0	83.0	55.0	154.3
Pb	0.37	100.0	255.9	163.5	406.7	100.0	562.1	360.8	941.3	100.0	521.5	295.0	789.1
Rb	0.39	79.2	3.6	1.1	5.3	100.0	17.8	10.2	24.0	16.7	ND	ND	ND
S	2.33	100.0	39,752.0	24,926.1	59,736.0	100.0	71,063.0	43,553.1	107,350.6	100.0	39,158.0	24,333.6	58,308.0
Sb	3.04	58.3	42.7	ND	125.7	100.0	1478.1	855.3	1883.2	41.7	ND	ND	35.7
Se	0.29	95.8	13.6	8.5	25.0	100.0	39.8	17.5	73.1	75.0	23.5	2.0	46.1
Si	1.68	95.8	6930.0	3979.1	8617.0	100.0	11,840	5741.4	16,071.9	100.0	7744.8	3860.2	10,276.5
Sn	3.01	79.2	112.3	10.8	231.8	100.0	1108.5	641.5	1412.4	100.0	207.5	118.3	384.1
Sr	0.34	100.0	9.5	5.9	15.2	100.0	18.6	11.2	23.7	100.0	37.1	14.7	86.0
Ti	5.54	100.0	1078.1	678.5	1425.0	100.0	249.5	141.2	463.1	100.0	205.4	162.5	372.2
V	0.44	100.0	156.7	115.6	314.7	100.0	332.2	238.5	714.5	100.0	185.5	108.2	369.2
Zn	0.41	100.0	1256	708.9	1563.0	100.0	2497.9	1478.2	3843	100.0	1504.3	920.3	2264.8

ND: Not detected.

. The detection rate was defined by the number of samples showing larger value than LOD divided by the entire samples’ numbers, i.e., 24.

Among the 25 commonly detected metals by 24 filters among 3 institutes, institute #1 showed a high detection rate (87.5% to 100%), except for Ni, 12.0%, Rb, 79.2%, Sb, 58.3%, and Sn: 79.2%. Institute #2 showed 100% detection rate for all elements. Institute #3 also provided 85 to 100% detection rate, except for the following 7 elements: As, 25.0%, Ba, 25.0%, Cd, 25.0%, Cs, 25.0%, Rb, 25.0%, Sb, 25.0%, and Se, 25.0%.

Using PM2.5 gravimetric mass level of each sample provided by Institute #2, we calculated the concentrations of elements in PM2.5 for each institute. The median (quartile) concentration of all 25 heavy metals across the three institutions showed Sr to be the lowest at 9.5 (5.9–15.2) ppm, and Al to be the highest at 12621 (7701.2–17176.9) ppm among outcomes of Institute #1. In case of Institute #2, the same metal (Sr) showed the lowest concentration at 18.6 (11.2–23.7) ppm while the highest one was Si (11840.0 (5741.4–16071.9) ppm. In Institute #3, Se was the lowest at 23.5 (2.0–46.1) ppm and Si was the highest at 7744.8 (3860.2–10276.5) ppm.

3.2. Correlation and Regression Analysis of Heavy Metals between Three Institutes

Spearman’s correlation coefficients for the metals between institute #1 and institute #2, and those between institute #1 and institute #3 are shown in Figure 2. Between institute #1 and institute #2, Fe, S, V, Zn, Mn, Pb, and K showed high correlation coefficients (0.90 or higher; p < 0.01) while Cu, Cl, Ba, Cd, Al, and Ca showed correlation coefficients between 0.80 and 0.89 (p < 0.01). Among the other elements, the correlation coefficients were 0.789 for As, 0.765 for Cr, 0.753 for Cs, 0.737 for Ti, 0.730 for Se, and 0.607 for Sr.

The results between institute #1 and institute #3 showed that correlation coefficients of Cu, Ba, and Cd were 0.757, 0.24, and 0.138, respectively. Other than those, the coefficients were 0.808 for Ti, 0.835 for Cr, and 0.887 or higher for Pb. For Ba, Cd, As, and Cs, the correlation coefficients between institute #1 and institute #2 were high while those between institute #1 and institute #3 were low, due to the large number of <limit of detection (LOD) in these metals in institute #3′s data.

Table 2. Coefficients of determination obtained from a simple regression between institutions.

	Summary of Simple Regression Analysis
	Institute #1−Institute #2			Institute #1−Institute #3
	Intercept	Slope	R2	Intercept	Slope	R2
S	4300	0.40 *	0.999	3800	0.87 *	0.999
Mn	−0.9	0.36	0.998	11.6	0.69	0.996
Pb	52.9	0.37 *	0.997	−2.2	0.55	0.990
Zn	156.3	0.38 *	0.996	170	0.63 *	0.995
K	395.9	0.54 *	0.992	487.7	0.97 *	0.990
Fe	4800	0.34 *	0.987	334.0	0.59 *	0.988
V	19.1	0.42 *	0.985	38.2	0.71 *	0.982
Cu	38.3	0.26 *	0.967	64.7	0.26 *	0.930
Sn	−140.5	0.24 *	0.945	−276.3	1.63 *	0.798
Ba	−235.7	0.87 *	0.932	196.6	3.41	0.682
Cr	−47.3	1.68	0.921	−12.9	2.38	0.971
Cs	−50.0	0.39 *	0.913	29.7	0.18 *	0.974
Sr	3.9	0.34 *	0.881	4.2	0.13 *	0.909
Ti	−540.9	5.99	0.878	−334.6	5.92	0.942
Cd	52.4	0.07 *	0.816	113.9	−0.10 *	0.001
Al	−2747.4	5.04	0.807	−2140.3	6.09	0.841
Ca	533.7	0.19 *	0.767	196.6	0.29	0.737
Se	−15.2	0.79 *	0.753	11.4	0.15 *	0.921
As	4.9	0.64	0.668	35.3	−0.03 *	0.000
Cl	−1136.4	3.82	0.641	565.3	5.40	0.521
Br	45.3 ± SE	0.46	0.543	84.4	0.14	0.642
Sb	45.2 ± SE	0.03	0.443	70.1	1.05	0.215
Ni	3.2 ± SE	0.01	0.024	2.1	0.02	0.003
Rb	3.9 ± SE	−0.01 *	0.009	3.5	0.08 *	0.007
Si	11,000 ± SE	−0.07 *	0.006	11,000	−0.10 *	0.012

*: p < 0.05.

summarizes the results of simple regression analysis of institute #1–institute #2 and institute #1–institute #3. The regression equations used are as follows: Institute#1_Xi = a + Institute#2 ∗ b and Institute#1_Xi = a + Institute#3 ∗ b where “a” and “b” are the intercept and the slope, respectively, and Xi represents each heavy metal analyzed. Between institute #1 and institute #2, the coefficient of determination (R²) ranged from 0.9 to 1.0 for S, Mn, Pb, Zn, K, Fe, V, Cu, Sn, Ba, Cr, and Cs, and from 0.8 to 0.9 for Sr, Ti, Cd, and Al. Ca, Se, As, Cl, Br, Sb, Ni, Rb, and Si had a low coefficient of determination, i.e., 0.8 or less (excluding negative coefficient of determination). Similarly, between institute #1 and institute #3, the elements with coefficient of determination between 0.9 to 1.0 were S, Mn, Pb, Zn, K, Fe, V, Cu, Cr, Cs, Sr, Ti, and Se, whereas that between 0.8 to 0.9 was Al. Finally, elements with the coefficient 0.8 or less were Sn, Ba, Cd, Ca, Cl, Br, Sb, As, Ni, Rb, and Si.

Commonly detected heavy metals for among institutes, the metals showed high coefficient of determination (0.9 to 1.0) were Cr, Cs, Cu, Fe, K, Mn, Pb, S, V, and Zn. Among them, S, Mn, Pb, Zn, and K were shown to be 0.99 or higher.

4. Discussion

Recent advances in X-ray fluorescence instrumentation allow for quick measurements of a range of elements that some time ago were available only through complicated and often destructive means of analytical chemistry (ICP (ICP-OES, ICP-MS)) methods for the analysis of heavy metals. XRF is a simple, fast, and safe method for sample preparation without chemical waste and with a non-destructive analytical technique, whereas ICP requires a long time for sample pretreatment. Further, the original state of samples is destroyed due to the complexity of preprocessing. These are usually used as a precision analysis process to determine confidence of the result from XRF [41]. On the other hand, on top of being easy-to-manage and easy-to-use, XRFs are widely used because of their sufficient reliability [42,43].

This study compared and evaluated the correlation and determination coefficients of the obtained results, using the same filter samples, across three global analytical institutes, using XRF. Since ED-XRF measured all the emitted X-rays simultaneously, based on the principle of multi-channel energy dispersion, both qualitative and quantitative analyses of the composition of heavy metals collected in a PTFE filter were possible within a short time. Moreover, since ED-XRF could, theoretically, infinitely measure, owing to minimization of the damage caused by heat generated by the filter, each institution in this study conducted heavy metal analysis using ED-XRF.

In this study, XRF was used to analyze the elemental composition of the measured 25 metals contained in PM2.5. The comparison of metal concentration in PM2.5 by research institutes are illustrated in Figure 1. The mostly detected element in PM2.5 by research institution #1 and #2 was S followed by Al, Si, and K, respectively, whereas Sr and Rb were the least abundant elements detected. Those results are similar to the findings of a study conducted in Istanbul, Turkey that mostly detected Si and Al as PM2.5 metal components in dense traffic area [44]. In institute #3, Al was the most abundant element followed by S and Si.

Results of the analysis (ng/mg, ppm) between institute #1 and institute #2 showed 16 out of 25 metals to have high (0.73 or larger) correlation between the two institutes. In particular, correlation coefficient of Fe, S, V, Zn, Mn, Pb, and K was 0.97 (p < 0.05) or higher, implying a very high correlation between the measured values from the two institutes.

Compared to the coefficients between institute #1 and institute #2, those between institute #1 and institute #3 were relatively low; we assumed this to be due to the results being affected by low detection rate in the institute #3 analysis. Institute #2 has yielded measurable values for heavy metal analysis of all samples. Using XRF instrumentation with improved sensitivity, PM component analysis studies like this will become more possible in the future using a low-burden PM monitor which can greatly contribute to an accurate health risk assessment. In addition, the coefficient of determination (${\mathrm{R}}^2$) obtained from simple regression analysis for 18 out of 25 heavy metals was 0.76 or more between institute #1 and institute #2.

Further, we compared the mean concentration of heavy metal components contained in our PM2.5 samples with those inside residential homes (n = 10) in the vicinity of the Xinqiao mining area, Tongling, China [45]. Although we could not compare the entire metal proportion profile because the latter reported concentration distributions of only 7 selected metals, the concentration level and ratios of most metals were similar. The mean concentration of Zn of this study was 1256 (ng/mg, ppm) while the one from Yangbing et al.’s study was 979 (ug/g, ppm) [45]. Ratios of mass concentration of Pb, Cr, Cu, Cd, and Ni in this study and Yangbing et al.’s study were 0.20, 0.28; 0.10, 0.06; 0.09, 0.33; 0.06, 0.01; not detected, 0.05, respectively. In case of Fe, the ratios were 1.83 (this study) and 27.7 in Yangbing et al.’s study. As characteristics of study areas between the two studies were different, there is some limitation on further interpretation. Nevertheless, the authors thought that our study results are informative in terms of indicating metal composition of PM2.5 collected from an urban residential asthmatic’s home.

On the contrary, according to previous studies conducted in Nepal’s urban area (outside housing) [46] and U.S. urban area (outside housing) [47], this study’s indoor Ca, Cr, Fe, and K levels were somewhat similar to the results from outside areas of Bhakrapur and Pennsylvania, with exceptions for S, Pb, and V.

Use of the same filter samples for consecutive analyses in three global institutes was a unique study design and the major advantage of this study. PM2.5 samples on filters were transported to institute #3 and institute #2 after the first analysis in Korea. They were safely transported, along with blank samples. Since the first sample analysis results by institute #1 and the last sample analysis results by institute #2 were highly correlated, we assumed that there was no serious systemic error during the process of transport. In this study, we used real-world field samples. Institute #1 has had a calibration curve prepared on a 40 mm filter for their daily routine commercial analysis while institute #2 prepared the curve on a 10 mm filter. Field operators used personal monitoring devices with 10 mm filters. The authors recognized that institute #1 used a modified 10 mm nozzle instead of the original 40 mm nozzle on XRF. The authors thought that because of the modified nozzle (10 mm) which had not been well fitted on the XRF hole (40 mm), that there was a systemic difference as institute #1’s results were about 2 times lower than that of institute #2 while coefficients of determination (R²) were high enough (0.9 or larger). Nevertheless, on the basis of this inter-lab comparison, XRF analysis results of the three institutes were concluded to be comparable. Although we had 25 metals collected from 8 households using 24 filters, which is a relatively small sample size, XRF proved to be a convenient method, providing reproducibility and compatibility of the PM2.5 metal component analysis data. As a result of the analysis of one filter in three institutes, most of the values were found to be at a fairly high level with 90% detection rate. Among the commonly analyzed 25 types of components, 16 were found to have a determination coefficient of 0.7 or more (10 components had 0.9 or more).

5. Conclusions

All three research institutes showed highly consistent results, notwithstanding the different measurement conditions. The reproducibility and compatibility of the PM2.5 metal component among three different research institutes obtained using XRF strongly highlights the suitability of the used sample preparation method. Despite the relatively small sample size, XRF analysis method allows a fast and easy surface analysis, and has a non-destructive property which makes it a suitable method to perform direct analysis of PM2.5 collected on filter membranes such as the Teflon filters used in this study for environmental pollution assessment purposes.