Introduction

Erionite is a fibrous zeolite and has a general formula of K2(Na, Ca0.5)8[Al10Si26O72]·30H2O1. It is classified by the International Agency for Research on Cancer (IARC) as a Group 1 carcinogen, and epidemiological data have linked erionite environmental exposure to a malignant mesothelioma epidemic in Cappadocia, Turkey, in the 1970s2,3,4. Erionite is also referred to as an asbestiform mineral because of its morphological similarities to a group of asbestos fibers—chrysotile, asbestos actinolite, amosite, asbestos anthophyllite, crocidolite, and asbestos tremolite—known human respiratory carcinogens5,6,7. Experimental data have indicated that fibrous erionite exhibits higher tumorigenic potential in rodents than crocidolite and chrysotile asbestos8,9,10.

Erionite is one of the five most abundant naturally occurring zeolites (along with clinoptilolite, phillipsite, chabazite, and mordenite) and is commonly found in near-surface sedimentary deposits11,12. Although it is very similar in morphology and chemistry to zeolites such as mordenite, erionite is the only zeolite species regulated as a carcinogen to date4. Therefore, the accurate characterization of individual zeolite fibers is critical for assessing both environmental and occupational exposure risks associated with exposure to erionite.

Understanding the structural and chemical complexities of erionite is critical to ensure that its fibers can be reliably differentiated from similar fibrous zeolites or other fibrous minerals13,14,15,16. There are three different species: erionite-K, erionite-Ca, and erionite-Na, depending on the most abundant extra-framework (EF) cation1,17 The erionite framework is composed of neighboring tetrahedral rings connected through tilted 4-rings along the c crystallographic axis, following the AABAAC. sequence (Fig. 1)14. The framework is characterized by three cavity types: the cancrinite cage, the double 6-ring cages, and columns of erionite cavities between the “B” or “C” 6-rings (Fig. 1)14. K+ cations are located at the center of the cancrinite cages, whereas Na+, Ca2+, Mg2+, and water molecules are located in various disordered positions inside the erionite cavities. One additional cation site was observed inside the erionite cage by Ballirano et al.18 in erionite (Fig. 1).

Fig. 1
figure 1

A framework sketch of erionite with cancrinite cages occupied by K+ ions (yellow circles), small empty double 6-ring cages, and large erionite cavities occupied by Ca++ (and Na+) (blue circles), Mg++ (small gray circles), and a fraction of K+ ions (K2 label, yellow circles). Eventually, the position labelled as Ca (in blue) also hosts Na+ ions. Atomic positions were based on data from Patel et al.19.

Full size image

To analyze the chemical composition of natural erionite, electron probe microanalysis (EPMA) with wavelength-dispersive spectrometry (WDS) and electron microscopy coupled with energy-dispersive X-ray spectrometry (EDS) are the most commonly used techniques employed in the majority of previous erionite studies13,15,20,21,22. EPMA is an analytical technique that utilizes higher beam energy levels (15–30 kV) and WDS, providing higher spatial resolution, element sensitivity, and more accurate chemical composition data than EDX23. However, it requires larger sample volumes, complex operating conditions, and extensive preparation (embedded in resin with a polished surface), making it unsuitable for analyzing individual erionite fibers in air samples24,25. EDX is a qualitative or quantitative microanalytical technique that detects the characteristic spectra of X-rays emitted by different elements in a sample after excitation by high-energy electrons generated by transmission electron microscope (TEM) or Scanning electron microscope (SEM)25,26. EDX operates in a “point-and-press” manner to rapidly detect major constituent elements, making SEM/TEM coupled with EDS widely used in environmental monitoring to efficiently identify and count the numbers of individual erionite fibers in occupational exposure and ambient air samples20,21,22. Compared to TEM-EDX, which requires switching from TEM to STEM mode for EDX analysis, SEM-EDX is more time-efficient for locating, identifying, and counting individual fibers within a single operational mode. Additionally, SEM can detect particles that are not distinctly elongated but contain elongated fibrils, key indicators of weathering, erosion, and the fragmentation and dispersion of erionite from its geological source27,28. These advantages make SEM-EDX one of the few practical techniques for detecting and characterizing erionite fibers in environmental samples27,28,29.

However, it is challenging to obtain detailed chemistry of natural erionite using both WDS and EDX technique due to its instability under electron beam, which leads to the migration of EF cations such as Na⁺, Ca²⁺, and a portion of K⁺. As shown in Fig. 1, K⁺ ions in cancrinite cages (K1) are more tightly coordinated with oxygen frameworks and thus less mobile, whereas Ca²⁺, Na⁺, and K⁺ ions in the larger erionite cavities (K2) are weakly bound via water molecules and more prone to beam-induced displacement14,30. The high energy of an electron beam can induce localized heating, producing surface conditions on the fiber that are comparable to thermal dehydration. Ballirano and Cametti30 demonstrated that such dehydration in erionite leads to substantial Ca and Na redistribution via internal ion exchange, involving the expansion of cavity bases and 6-membered rings.

To reliably identify erionite, and to filter out uncertainties caused by cation variations under the electron beam, chemical criteria have been proposed based on a 72-oxygen framework: Si + Al ≈ 36, balance error (E%, Eq. 1) in the range of ± 10 (or ± 20), and Mg content < 0.8 [13,14,15,16]. These criteria are commonly assessed using EPMA analysis results15,19.

$$\:E\%=\frac{\left(Al\:+\:{Fe}^{3+}\right)\:-\:Na\:+\:K\:+\:2\:\times\:\:\left(Ca\:+\:Mg\:+\:Sr\:+\:Ba\right)}{Na\:+\:K\:+\:2\:\times\:\:\left(Ca\:+\:Mg\:+\:Sr\:+\:Ba\right)}\times\:100$$
(1)

Adding to the challenge, assessing the risk posed by carcinogenic erionite requires analysis at the level of individual, small-sized fibers. While fibrous erionite poses a minimal risk in its undisturbed geological form, it presents a considerable health hazard when disturbed and aerosolized into individual respirable-sized fibers3,20,31,32,33. Previous studies have reported that individual erionite fibers in ambient air samples are extremely thin, indicating very low mass per volume2,22. This may lead to reduced X-ray intensity and low weight percentages detected by EDS20,22,29.

The chemical composition of erionite fibers in environmental samples is further complicated, as fibers commonly undergo physical breakdown and weathering and are often mixed within matrices containing both inorganic and organic materials. Therefore, sample extraction processes are essential for isolating fine, individual erionite fibers from complex matrices, such as air and soil samples9,28,34,35. For example, the fluidized bed asbestos segregator, originally designed to separate asbestos fibers from soil, was adapted by Farcas et al.28 for erionite fiber isolation from soil. Similarly, Mumpton34 employed deionized water dispersion to separate fine erionite fibers from bulk building materials, and ISO 13794 recommends distilled water dispersion for releasing asbestos fibers from overloaded air filters. In addition, hydrogen peroxide (H₂O₂) has been applied to remove organic materials and concentrate fibrous minerals, as demonstrated by Fan et al.27 for mordenite.

Given the susceptibility of erionite to both physical disaggregation and chemical alteration during environmental and laboratory processing, it is critical to understand the influence of sample preparation methods on the elemental quantification of individual fibers analyzed via SEM-EDX. The interpretation of SEM-EDX data must account for these preparation effects, particularly when comparing environmental erionite fibers to mineralogical reference criteria, typically derived from high-purity, uncontaminated rock samples13,15,16,36.

Although there are known differences in the spatial resolution and beam interaction depth (surface vs. penetrating) between EPMA and EDX, the extent of cation migration and deviation in the calculated quantitative chemical composition of erionite between the bulk samples analyzed using EPMA and those in individual fiber forms analyzed using EDX, and the degree to which factors—such as fiber size, sample preparation method, and instrument settings—influence this variation, remains unclear16,29,37.

To address these issues, bulk erionite samples with high purity, confirmed by X-ray powder diffraction (XRD) and previously analyzed using EPMA19 were ground into fragments (individual fiber forms), prepared under four conditions, and individual fibers were analyzed using two SEM-EDS systems. This investigation aimed to determine the factors affecting the accurate detection of erionite chemistry, including tetrahedral framework elements (Si and Al) and EF cations (Na, K, Mg, and Ca), across varying fiber sizes, equipment settings, and preparation methods (conditions). The EDX results for the weight percentages (wt%) were normalized to a 72-oxygen framework, allowing direct comparison with EPMA data.

Materials and methods

SEM-EDX settings

Two SEM-EDS systems were used (Table 1): SEM.EDS.1 was coupled to a Hitachi SU-70 Schottky SEM, and EDS.2 consisted of an Oxford EDS paired with a TESCAN CLARA SEM. Both machines were equipped with pre-installed ZAF correction procedures to account for the effects of the atomic number (Z), absorption (A), and fluorescence excitation (F). SEM.EDS.1 uses a traditional Si (Li) detector requiring a nitrogen cooling system, whereas SEM.EDS.2 features a newer, advanced silicon drift detector with advantages such as nitrogen-free operation and better energy resolution at higher count rates (https://nano.oxinst.com/)37.

EDXA was performed on both systems using an accelerating voltage of 15 kV and an acquisition time of 60 s. The EDX spectra were quantified as weight percentages (wt%) using the pre-installed microanalysis software in each system (Table 1). Point-scan analysis was used for spectrum acquisition, with each scan positioned at the approximate geometric center of the fiber. The fiber width was measured precisely at the point-scan location.

Table 1 SEM-EDX analysis settings for the two SEM-EDS systems.
Full size table

Erionite bulk samples

Erionite-containing bulk samples were confirmed through a multi-analytical approach, including micro-Raman spectroscopy, thermogravimetric analysis, electron microscopy, and electron microprobe analysis19. This volcanic type erionite was found within vesicles of volcanic rock and is highly pure with only ~ 2 wt% detectable impurities (smectite and illite) based on the results from quantitative X-ray powder diffraction (XRD) analysis. The erionite was suggested to be erionite-K, as K was found to be the most abundant extra-framework cation compared to Ca and Na in the EPMA results, and the following formula was proposed in Patel et al.19:

$$K_{{2.63}} Ca_{{1.27}} Mg_{{0.98}} Na_{{0.37}} Ba_{{0.03}} Sr_{{0.01}} \left[ {Si_{{28.43}} Al_{{7.56}} } \right]_{{35.99}} O_{{72}} \cdot29.07H_{2} O$$

The EPMA elemental composition values for each element in this formula were compared with the weight percentages (wt%) of the SEM-EDX results obtained in this study.

Erionite fiber preparation conditions

The erionite bulk samples were dried, cryo-milled into a powder, and then prepared under four conditions (Table 2). Each condition was replicated for two sets of samples for the two SEM-EDS systems. The erionite fiber sizes were measured using a measuring tool preinstalled in each SEM and recorded manually.

Four sample preparation methods were applied to simulate the approaches used for the separation of fibers from various environmental matrices. These included dry and air dispersions, similar to the fluidized bed asbestos segregator method for separating fibers from soil28 and deionized water dispersion, following Mumpton34 for isolating erionite from building materials and ISO 13794 for dispersing asbestos fibers from overloaded filters38. Additionally, an organic removal process using H₂O₂ was applied, consistent with Fan et al.27 for mordenite and the protocols used in atmospheric microplastic studies39.

First, dry and air-dispersed individual erionite fibers were prepared by spreading dry powdered erionite onto carbon tape mounted on two SEM stubs, thus reducing potential contamination and the likelihood of ion exchange (Table 2). Second, deionized water-dispersed erionite fibers were prepared by dispersing fragmented bulk erionite samples in deionized water, then the suspension was filtered through polycarbonate track-etched (PC) membrane filters (0.2 μm pore size, 47 mm diameter, Sterlitech). After drying at room temperature for 1 week, two sets of one-eighth of the PC filter were cut and mounted onto two SEM stubs (Table 2). Third, the digested (bulk) erionite fibers were placed into a 300 mL beaker, followed by the addition of 30 mL of 30% H₂O₂ (analytical reagent grade, Fisher Scientific). The beaker was kept in a fume hood at room temperature for 48 h, then placed in a 90 °C oven for 8 h (Table 2). Fourth, the digested (bulk + dust) samples followed the same process as the digested (bulk), except that shrub twig samples were added, collected from a roadside location where the bulk erionite samples were collected. This setup was designed to investigate potential changes in the elemental composition of erionite and ion-exchange interactions in the presence of H₂O₂ and mineral road dust from twig samples. After processing, the digested (bulk) and digested (bulk + dust) samples were rinsed in deionized water and filtered individually onto two PC filters, which were then cut into two one-eighth portions and mounted onto two SEM stubs for subsequent SEM-EDX analysis (Table 2).

Quantitative calculation of erionite chemical composition 

Mineral compositions are typically reported as oxides; however, expressing them in atomic proportions enhances the elemental relationship analysis, aiding in mineral identification25,40. The EDX results (wt%) of each element for individual erionite fibers tested under two SEM-EDS settings were calculated as oxides first, then normalized for quantitative chemical analysis of erionite-relevant elements including Si and Al in the tetrahedral framework, as well as its EF cations (Na, K, Mg, and Ca)14,15. The oxygen content was calculated using stoichiometry29,41and the chemical composition of erionite was calculated using a framework of 72 oxygen atoms and 17.69 H₂O19.

Table 2 The four erionite individual fiber sample preparation conditions.
Full size table

Statistical analysis

All statistical analyses were performed using R (version 4.4.2)42. Linear regression analysis was conducted with the linear model lm() function in base R to evaluate the relationship between fiber width and the ratio of SEM-EDX element wt% to EMPA reference value. Regression intercepts and slopes were extracted for comparison across all sample conditions. Data visualization was performed using the ggplot2 package.

Results and discussion

Overview of elemental detection in the two SEM-EDS systems

A total of 325 individual erionite fibers were tested under the two SEM-EDS systems: 173 (width = 0.96 ± 0.83 μm) and 152 (width = 0.75 ± 0.71 μm) were tested using SEM-EDX.1 and SEM-EDX.2, respectively. Although crystallographic data indicate that the bulk erionite fibers are highly pure, with only ~ 2% impurities19, analysis of extra-framework cations showed that K was detected in 270 fibers, Ca in 243 fibers, and Mg in 254 fibers. Only 66 fibers have Na detected. Fe was also detected from majority of the fibers tested and was considered as an impurity and not belong in the structure of natural erionite14. As a result, Fe was excluded in the following erionite quantitative chemical calculations14,15.

The overall distribution of elemental data from 325 tested fibers, which analyzed across two SEM-EDS systems and four sample preparation methods, is summarized in Table 3. In general, both SEM-EDS systems show that coefficients of variation (CV) increase as elemental concentrations (wt%) decrease. Accordingly, silicon (Si) exhibited the most stable measurements, while calcium (Ca) and sodium (Na) showed the highest variability in detected wt% values.

Table 3 Mean elemental compositions (wt%, 72 Oxygen) and coefficients of variation (CV, %) and standard deviation (sd) for individual erionite fibers analyzed by SEM-EDS under four different sample treatments: dry & air-dispersed, DI water-dispersed, digested (bulk), and digested (bulk + dust). Data are presented for two instruments (SEM-EDS.1 and SEM-EDS.2).
Full size table

SEM-EDX data compared to EPMA data

The variation patterns between the mean elemental weight% (wt%) results for erionite from EPMA analysis, as reported in Patel et al.19 and the wt% results from 325 individual fibers analyzed by SEM-EDX in this study are shown in Fig. 2.The Si wt% values were the most comparable to the EPMA result (28.43 wt%), with overestimations of ~ 13% in SEM.EDS.1 and only 2% in SEM.EDS.2. Al exhibited an underestimation of 32% in SEM.EDS.1 and 4% in SEM.EDS.2 compared to the EPMA result of 7.56 wt%. For the EF cations, K and Ca were consistently detected at concentrations lower than their EPMA values (K: 2.63 wt%, Ca: 1.27 wt%), whereas Mg and Na appeared in higher concentrations in the SEM-EDX results (EPMA: Mg: 0.99 wt%, Na: 0.37 wt%). The general trends of Si overestimation and Al underestimation in SEM-EDX wt% results compared to those in EPMA align with previous findings by Cametti et al.41 for erionite and Pacella et al.29 for standard zeolites (leucite, nepheline, scolecite, and natrolite).

Fig. 2
figure 2

Ratio of SEM-EDX to EPMA element weight percentages (wt%) for individual erionite fibers as a function of fiber width, shown for SEM.EDS.1 and SEM.EDS.2. Panels display Si, Al, K, Ca, Mg, and Na, with EPMA reference wt% values in parentheses19. Data points are color-coded by sample preparation method. Linear regression trends indicate relationships between fiber width and the ratio of SEM-EDX to EPMA wt%.

Full size image

The discrepancies between SEM-EDX and EPMA in the elemental quantification of erionite fibers reflect the inherent limitations of SEM-EDX as a semi-quantitative technique, particularly for light elements and EF cations26,43. Si overestimation in SEM-EDX—with deviations of 13% and 2% in SEM.EDS.1 and SEM.EDS.2, respectively—can be attributed to the higher X-ray energy of Si Kα, which reduces self-absorption and enhances detectability26,37,43. Conversely, Al, which has lower X-ray energy, is more susceptible to absorption losses, leading to underestimation (32% in SEM.EDS.1 and 4% in SEM.EDS.2). These trends are consistent with earlier studies that reported similar biases in zeolite characterization using SEM-EDX29,41.

The under-detection of K and Ca, also observed by Pacella et al.29 likely results from the loosely bound nature of charge-compensating cations in zeolites, which are susceptible to migration or loss under vacuum conditions in SEM-EDX26,43. Similarly, the under-detection of K and Ca in wt% was observed by Pacella et al.29 for standard zeolites. Additionally, Na was absent in most fibers, likely because it is a light element and may be unstable under high-energy electron beams16,26. In a previous study, a Na loss of up to 25% was seen from 10s EDX counting time to 60 s counting time for erionite-Na, where Na is the most abundant EF cation (1.97wt.%)29 while in this experiment, the fibers were suggested as erionite-K, in which Na only constitute of low concentration of 0.37 wt%, which may explain the absence of Na in the majority of the samples19,29. Na loss was observed not only in SEM-EDX but also in the analysis of individual erionite fibers using TEM-EDX, potentially owing to the interaction with the high-energy beam13,16,22.

Interestingly, the Mg concentrations were overestimated for some wt% results from both SEM-EDS systems compared to the EPMA results (Fig. 2). Regarding the working principles of EDX and EPMA, EDX has a larger interaction volume and surface sensitivity, whereas EPMA has a smaller, more focused beam spot that penetrates deeper26,44. Patel et al.19 suggest that some Mg-rich minerals that geologically coexist with the erionite bulk samples might attach on its surface and contribute to the Mg content in the samples. EDX, with its broader beam interaction and surface sensitivity, may detect Mg from erionite surface-attached nanoparticles or the surrounding matrix. For example, H₂O₂-treated erionite fibers exhibited lower Mg content compared to untreated fibers (Fig. 2). Similarly, Mg was overestimated in SEM-EDX wt% of woolly erionite-Ca compared to the EPMA wt% results reported in Cametti et al.41.

Although both techniques operate at the same accelerating voltage (15 kV) and thus with comparable beam currents, the compositional shifts in elemental weight percentages between SEM-EDX analysis of individual erionite fibers and EPMA of bulk samples reflect fundamental differences in analytical principles as well as the challenges posed by the structural and chemical complexities of zeolites14,23,26,45. The systematic overestimation of Si and Mg—along with the underestimation of Al, K, and Ca—aligns with previous studies on zeolites and primarily reflects differences in detector sensitivity, X-ray absorption efficiency, and the larger interaction volume inherent to SEM-EDX26,29,37,43. The pronounced under-detection and potential beam-induced loss of Na further highlights the vulnerability of light elements to analytical conditions, particularly in highly voided materials such as erionite13,16. Although these technique-specific analytical factors account for much of the observed variation, additional contributions from differences in fiber size, surface morphology, and sample preparation processes cannot be excluded, particularly given the larger interaction volume and surface sensitivity of SEM-EDX.

Fiber size effect

Figure 2 illustrates the relationship between the fiber width and SEM-EDX to EPMA weight% (wt%) ratios for key framework and EF elements in individual erionite fibers across two SEM-EDS systems. In general, the ratio data points showed less scattering in the linear regression mode as the erionite fiber width increased, except for Na, indicating more stable elemental detection in wider erionite fibers in SEM-EDX. This effect was more apparent for the Si and Al wt% in both SEM.EDS systems, but less apparent in the EF cations (K, Ca, Mg). These trends align with the established size-dependent analytical artifacts documented in previous studies29,46 where—compared to bulk samples—the diameter decrease of the fragmented zeolite particles resulted in a greater element wt% variance detected from SEM-EDX29,46.

Size-dependent deviations can be seen in Fig. 2 as in the detected concentrations of the framework elements Si and Al across individual erionite fibers analyzed by both SEM-EDS systems, particularly for fibers with widths < 1 μm. In SEM.EDS.1, fibers < 0.75 μm consistently exhibit overestimated Si concentrations relative to the EPMA wt% value of 28.43, while fibers analyzed by SEM.EDS.2 display two-directional variation in Si content, particularly for fibers < 0.5 μm. For Al, SEM.EDS.1 showed notable underestimation in fibers < 0.5 μm, whereas in SEM.EDS.2, Al content varied in both directions across fibers < 1 μm. These patterns reflect the impact of fiber size on X-ray generation in SEM-EDX. Smaller fibers produce fewer X-rays because of their lower mass, leading to reduced element detection. Compared to Si, this effect is more pronounced for lower-energy elements such as Al, which experience greater signal loss in fine fibers26,29,46.

The linear regression line data presented in Table 4 further reinforce these observations, demonstrating that for both SEM-EDS systems, the slopes of the linear regression lines for Si and Al detection decreased markedly for fibers > 0.5 μm in width. This indicates that for fibers > 0.5 μm, the detected Si and Al wt% values are more stable and exhibit minimal sensitivity to further increases in fiber width. These trends are consistent with findings reported by Pacella et al.29 who observed similar size-dependent deviations when comparing SEM-EDXA data for smaller fragments (diameters < 1 μm) of standard zeolites to bulk samples, where reduced absorption and surface roughness effects were found to disproportionately affect smaller particles.

Table 4 The linear regressions (for all the fibers and fibers width > 0.5 μm) towards y = 1 for each element for data from both SEM.EDS systems.
Full size table

Supporting the linear patterns observed in Fig. 2 and the data in Table 4; Fig. 3 shows that elemental detection results from both SEM-EDS systems were significantly more homogeneous for individual fibers with widths greater than 0.5 μm across all elements except Na.

As noted in Sect. 3.1, not all erionite fibers analyzed (N = 325) contain detectable EF cations (K, Ca, Mg, and Na), and fiber width appears to play a major role in this variability. Among the subset of narrower fibers (< 0.5 μm, n = 94) analyzed across both SEM-EDS systems, only 54% detected K (n = 51); 40% detected Ca (n = 38); 41% detected Mg (n = 39); and none detected Na. This sharp decline in EF cation detectability in narrower fibers is consistent with the enhanced beam-induced cation migration and volatilization reported in the literature23,29,46.

Regarding individual EF cations (K, Ca, Mg, and Na) in the tested individual erionite fibers, the substantial data scatter in the wt% results obtained from both SEM-EDS systems precludes the identification of clear or consistent correlations between fiber width and elemental concentrations (Fig. 2). The regression slope values for K, Ca, and Mg in SEM.EDS.1 exhibited a notable decrease for fibers > 0.5 μm, indicating a reduction in size-dependent variability in larger fibers (Table 4). However, in SEM.EDS.2, no such reduction was observed for K, and both Ca and Mg displayed increased slope values for fibers > 0.5 μm, indicating that fiber width exerts a more complex influence on these elements in this system. These findings are broadly consistent with those of Pacella et al.29 who reported greater scatter in wt% data for Ca, Na, and K relative to Si and Al, with no discernible correlation between elemental concentration and fragment diameter across four standard zeolites.

Na detection was particularly limited, with Na identified mostly on individual erionite fibers wider than 1 μm across both SEM-EDS systems. This extreme scarcity of detectable Na in smaller fibers aligns with prior observations of Na instability under electron beam exposure, as documented by Pacella et al.29 and Paoletti et al.46 where beam-induced volatilization and cation migration were shown to be particularly severe for Na in zeolitic fibers13,16.

Fig. 3
figure 3

Coefficient of variation (CV%) of elemental composition (Si, Al, K, Ca, Mg, Na) in erionite fibers analyzed by two SEM-EDS systems under four sample preparation conditions: dry & air-dispersed, DI water-dispersed, digested (bulk), and digested (bulk + dust). Results are separated by fiber width: ≤0.5 μm (left) and > 0.5 μm (right).

Full size image

Sample preparation effect

Understanding the potential impact of different sample preparation methods on elemental quantification in individual environmental erionite fibers analyzed via SEM-EDX is critical when interpreting data against criteria established in mineralogical studies13,16. Individual erionite fibers were prepared under four conditions, similar to the sample preparation methods used in previous environmental erionite studies (Fig. 4)27,28,34.

Fig. 4
figure 4

Images of erionite fiber examples prepared under four different conditions (Table 2).

Full size image

Si demonstrated the most stable detection across all four sample preparation conditions for fibers > 0.5 μm in both SEM-EDS systems (Figs. 3 and 5). As expected, the dry and air-dispersed fibers exhibited the greatest stability, with the linear regression slope for Si approaching zero (0.001) and a consistent mild overestimation (linear regression intercept) of 4.5–9.9% in SEM.EDS.1 and 1.9–4.9% in SEM.EDS.2, relative to the EPMA reference value of 28.43 wt% (Fig. 5; Table 5).

Table 5 Linear regression line equations for Si, Al and K in erionite fibers tested (width ≥ 0.5 μm) by the four sample preparation methods (SEM-EDS.1 and SEM-EDS.2).
Full size table
Fig. 5
figure 5

Ratio of SEM-EDX to EPMA elemental weight percentages (wt%) for individual erionite fibers as a function of fiber width, separated by preparation method and instrument (SEM.EDS.1 and SEM.EDS.2). Each data point represents a single fiber, color-coded by the SEM-EDS system used.

Full size image

The sample preparation sequence, from dry and air-dispersed to digested (bulk + dust), was assumed to reflect the increasing levels of physical and chemical manipulation. The degree of scattering in the Al wt% (expressed as SEM-EDX/EPMA ratios) increased progressively with greater manipulation across both SEM-EDS systems (Fig. 5). Although the regression slopes for Al were small (-0.005–0.066), indicating relatively stable detection for fibers > 0.5 μm, the negative intercepts for all preparation methods (ranging from -0.129 to -0.232 in SEM.EDS.1 and  -0.031 to -0.106 in SEM.EDS.2) confirm a consistent underestimation of Al compared to the EPMA reference of 7.56 wt%. The only exception was the digested (bulk + dust) condition analyzed by SEM.EDS.2, where Al was slightly overestimated (3.7%) (Table 5, red rectangle).

For K—the most abundant EF cation in the suggested erionite-K bulk sample [19]—SEM-EDXA detection was relatively stable across the fiber widths compared to other EF cations (Ca, Mg, and Na). However, a systematic underestimation was observed in both the SEM-EDS systems with K wt% ranging from 51.8% to 73.1% of the EPMA reference (2.63 wt%) in SEM.EDS.1 and from 53% to 70% in SEM.EDS.2. Notably, erionite fibers subjected to H₂O₂ digestion (both bulk and bulk + dust) exhibited more pronounced underestimation of K compared to the less processed samples, suggesting that ion exchange or cation leaching may have occurred during the H₂O₂ plus heat digestion process47.

Ca displayed a distinct pattern (Fig. 5), in which the dry and air-dispersed fibers exhibited the lowest EDX/EPMA wt% ratios in both SEM-EDS systems compared to the more manipulated samples. Similar findings were reported by Ballirano and Cametti45 who documented Ca uptake by erionite fibers from aqueous solutions, including artificial lung fluid, even in fibers that did not initially contain Ca. It is plausible that Ca uptake occurred during the deionized water dispersion and H₂O₂ digestion processes in the current study, although further investigation is required to confirm this hypothesis.

Mg detection exhibited substantial scatters across all preparation methods, with no clear linear relationship between fiber width and SEM-EDX/EPMA wt% ratios. However, Mg detection generally decreased with increasing sample manipulation levels (Fig. 5). Dry and air-dispersed fibers showed the highest Mg detection, followed by deionized water-dispersed, while fibers subjected to H₂O₂ treatment (both bulk and bulk + dust) exhibited the lowest Mg detection. This pattern is consistent with the interpretation that surface contamination from Mg-rich phases (e.g., smectite and illite) contributes to the apparent Mg enrichment in minimally processed samples19. These surface-bound particles are likely removed during water dispersion and chemical digestion, resulting in lower Mg detection in the more heavily processed fibers (Fig. 5).

Na was detected almost exclusively in the fibers prepared under dry and air-dispersed conditions in both SEM-EDS systems. Pacella et al.29 reported a similar phenomenon, attributing Na detection in smaller particles to contamination from the carbon tape. Here, background analysis confirmed the absence of detectable Na in the carbon tape.

It is also noteworthy that the deionized water-dispersed fibers exhibited the greatest scatter in wt% rate values for most elements across both SEM-EDS systems (Figs. 3 and 5). For Si, in particular, increased scatter was observed for narrower fibers (width 0.5–1 μm) prepared using deionized water, with the highest linear regression slope (-0.025 in SEM.EDS.1 and  -0.005 in SEM.EDS.2) among the four preparation methods. This may be partially explained by the higher proportion of narrow fibers analyzed in the deionized water-dispersed condition, as smaller fibers are more susceptible to analytical artifacts such as surface charging, absent mass effects, and element migration, all of which can contribute to greater variability in the detected wt%.

The variations in elemental detection across the four sample preparation methods highlight the considerable influence of physical and chemical processing on the SEM-EDXA of erionite fibers. The dry and air-dispersed fibers showed the most stable detection, particularly for Si and Al, with minimal deviations from the EPMA reference values. In contrast, fibers subjected to deionized water dispersion and H₂O₂ digestion exhibited greater data scatter and reduced detection of EF cations, particularly K and Mg, likely due to ion exchange, cation leaching, or surface cleaning during preparation. These findings emphasize the importance of considering preparation history when interpreting SEM-EDX data for environmental erionite fibers. Beyond preparation effects, instrument-specific differences between the two SEM-EDS systems further contribute to elemental detection variability, explored in the following section.

SEM-EDS instrument setting effects

Two SEM-EDS systems were used here to analyze individual erionite fibers. SEM-EDS.1 is an older instrument with a conventional Si(Li) detector, whereas SEM-EDS.2 is a more advanced system equipped with a modern Si drift detector, which is considered more accurate and comparable to WDS in EPMA25. Several factors can potentially influence SEM-EDX elemental detection shifts, including differences in the SEM spatial resolution settings and different analytical algorithms for spectrum-to-wt% conversion25,26,43. Here, these factors were generalized and classified into instrument settings.

For the framework elements Si and Al, as shown in Table 3, SEM-EDS.1 reported higher average Si and lower mean Al values across the four sample preparation conditions compared to SEM-EDS.2. In comparison to EPMA results, SEM.EDS.2 showed better accuracy than the reference EPMA wt% values (28.43 and 7.56). The regression lines for Si and Al in SEM-EDS.2 are closely aligned with the erionite EPMA results reported in the literature, with intercept values (to y = 1) of 0.023 for Si and 0.035 for Al (Figs. 1 and 3; Table 4). In comparison, SEM-EDS.1 exhibited larger deviations, with intercepts of 0.132 for Si and 0.316 for Al, indicating a greater degree of overestimation for Si and underestimation of Al in the older system, particularly in narrower fibers.

However, for the EF cations K, Ca, and Mg, SEM-EDS.1 demonstrated closer agreement with the EPMA results in terms of the regression intercept values than did SEM-EDS.2 (Table 4), although both systems displayed considerable scatter, especially for Mg and Na. The patterns of elemental variation across the four sample preparation methods are broadly comparable between the two systems. For example, in both SEM-EDX systems, H₂O₂-digested (bulk) erionite fibers consistently exhibited higher wt% values of Si and Mg, while Al, K, and Ca were detected at lower levels compared to fibers in the digested (bulk + dust) condition (Fig. 5). Notably, SEM-EDS.2 detected Na in more individual erionite fibers than did SEM-EDS.1, which may reflect differences in low-energy X-ray sensitivity or background subtraction capabilities between the two systems.

These findings highlight the presence of systematic instrument-specific biases in the detection and quantification of both framework and EF elements across different SEM-EDS systems37. Such variability emphasizes the importance of routine instrument calibration and standard testing prior to environmental sample analysis to avoid systematic instrumental variations.

Application of SEM-EDX data to erionite identification criteria

In this study, 325 individual woolly erionite fibers from a bulk sample with ~ 98 wt% purity, as confirmed by quantitative crystallographic structure data by XRD and multiple analytical techniques, were tested using two SEM-EDS systems. Surprisingly, none of the individual erionite fibers analyzed by SEM-EDX met all the criteria recommended in the literature for the quantitative chemical identification of erionite (Table 6)13,15,16,36. Several factors might have contributed to these results.

First, for the framework elements Si and Al, 87% of the fibers met the criteria Si + Al ≈ 36. This was mainly caused by the size effect mentioned in Sect. 3.3, in which smaller fibers showed greater variation in Si and Al detection, whereas in some of them, EF cation detection was lacking. Surprisingly, Tsi (Si/ (Si + Al)) ratios in the erionite reference range (0.68–0.79) were only observed in 13% of the fibers (Table 6)1. The original Tsi ratio for erionite reported by Patel et al.19 was 0.79. Although the overall detection of Si and Al here was relatively stable, several factors—including the instrument setting effect, positive intercept values for Si, and negative intercept values for Al in both SEM-EDS systems—could have exaggerated the ratio of Si/ (Si + Al). If the data is corrected by the intercept values for each SEM-EDS system, the Tsi ratios for 56% of the fibers are within the erionite range1. The remaining differences may have caused by the size effect, with great variation in fiber width < 0.5 μm, as well as the sample preparation effect. The intercept of the regression lines for Si and Al in SEM-EDS.2, fibers in Digested (bulk + dust) condition showed opposite values (negative for Si and positive for Al), in contrast to the other three sample conditions (Fig. 5; Table 6).

Second, the Mg content of erionite bulk sample was higher (EPMA wt% 0.99) than the value limit (< 0.8) proposed in the literature14,15,19 which led to only 29% of the fibers being within the limit of 0.8. Most of this group of fibers were erionite fibers treated with H2O2 and tested using SEM.EDS.1 (Fig. 5). Together with the low detection rates for Na in both SEM-EDXA, this led to no erionite fibers in this experiment matching the criteria of Mg/ (Ca + Na) < 0.15 (Table 6).

Third, due to the elemental detection variations for the EF cations, only 9% (n = 28) and 24% (n = 78) of fibers met the criteria E% ≤ ±10 and E% ≤ ±20, respectively (Fig. 6). Surprisingly, all the fibers that met the criteria E%≤ ± 20 were prepared in dry powder form on carbon tape and dispersed in deionized water (Fig. 6). Here, no fibers matched both standards of Mg < 0.8 and E% ≤ ±20 (Table 6).

Fig. 6
figure 6

Balance error E% results of the individual erionite fibers tested. Data points are color-coded by sample preparation condition.

Full size image

These results would appear to question the validity of the SEM-EDX technique as an analytical technique for identifying erionite, and caution is required in the application of this analytical technique26. However, it is noteworthy that the chemistry of naturally occurring erionite is highly variable36,48. Previous EMPA studies of bulk samples of erionite from the epicenter of malignant mesothelioma in villages in Turkey, and erionite samples from Nevada, US, which considered standards for animal and cell experiments, also did not meet all the quantitative chemical composition criteria listed in Table 6, as reported in Dogan and Dogan36. Further, the quantitative characterization of air erionite samples (width ≈ 0.5 μm) collected from Cappadocia in Turkey (analyzed using TEM-EDX) showed similar patterns of low or absent Na observed in this study, with only 18.3% passing the E% (10%) test and only 5% of the 60 fibers passing both the E% and Mg < 0.8 criteria13,14.

SEM-EDX remains the most effective and efficient technique for analyzing individual erionite fibers in air samples, including dust, ambient air, and human exposure samples, because of its ability to confirm fibrous morphology and handle large sample volumes, especially when fiber concentrations are low27. However, the findings here indicate that care must be taken when comparing results with the quantitative chemical erionite identification criteria suggested in the literature14,15,18. These identification criteria may not be fully applicable when analyzing smaller erionite fibers in mixed environmental samples using SEM-EDX. The findings here are consistent with those of previous studies that reported similar challenges in air sample analysis using TEM-EDX13,21,22.

Table 6 Quantitative chemical erionite identification criteria and the number of fibers met the criteria in this study.
Full size table

Further research has explored potential solutions to some of these issues, such as using a cryo-holder in TEM-EDX to minimize Na loss and improve EF cation detection; however, this approach has not yet been tested on individual environmental erionite fibers, and its applicability to SEM-EDX remains uncertain16. Nonetheless, the results show that if fiber size effects, instrument settings, and sample preparation conditions are accounted for—and given the relatively stable detection of Si and Al—SEM-EDX accuracy can be improved through pre-testing with erionite standards and applying system-specific corrections43. Pre-testing enables a correction index to be applied using the linear regression intercept to enhance individual erionite fiber identification based on framework element-related criteria (e.g., the Tsi ratio) in future studies conducted at the same locality using the same SEM-EDS system. However, this correction index is sample- and instrument-specific.

Limitations

This study provides insights into the quantitative chemical analysis of individual erionite fibers using SEM-EDX; however, several limitations must be considered. First, natural compositional heterogeneity may have affected the results. Although EPMA was conducted on portions of the same fiber bundle, chemical variations may still exist across different sections of the natural erionite sample19,49. Second, because these are natural erionite samples, some variations can be attributed to their heterogeneous surface geometry26. Third, the effect of the sample coating was not independently assessed but was included in instrumental effects. Variations in the coating thickness and material (e.g., carbon and gold) can influence X-ray absorption and element detection, particularly for low-energy elements26,43. Finally, residual impurities (2%) in the erionite samples may have introduced minor compositional deviations in SEM-EDX results.

Conclusions

Erionite is classified as a Group 1 carcinogen and may pose significant health risks when aerosolized into individual elongated respirable-sized fibers. This study evaluated the reliability of SEM-EDX for the quantitative chemical identification of individual erionite fibers. By analyzing 325 erionite fibers using two SEM-EDS systems and comparing the results with EPMA reference data, systematic shifts of elemental detection were observed between the EDX and EPMA wt% results, which consistent with the findings of previous studies. Specifically, overestimation of Si and underestimation of Al, K, and Ca were observed in the SEM-EDX data compared with the EMPA results. Fiber width-dependent variations were observed in the erionite framework element Si and Al in fibers widths < 0.5 μm, but were not apparent in EF cations (K, Ca, Na). Sample preparation methods, particularly deionized water dispersion and H₂O₂ digestion, further contributed to elemental detection discrepancies, likely due to ion exchange and cation mobilization. Additionally, instrument-specific differences were evident, with SEM-EDS.2 demonstrating improved consistency for the framework elements.

None of the analyzed fibers met all the established erionite identification criteria, highlighting the limitations of SEM-EDX data for reliable quantitative chemical analysis due to variations introduced by small fiber size, different instrument settings, and sample preparation methods.However, the relatively stable detection of Si and Al suggests that SEM-EDX remains a viable tool for preliminary erionite identification, particularly with fiber widths > 0.5 μm. For example, the Si: Al ratio and the fraction of total tetrahedral sites occupied by Si, commonly referred to as the Tsi ratio or R value (Si/(Si + Al)), can serve as indicators for preliminary erionite identification and can be compared to reference values reported in literature1. To enhance reliability, pre-calibration with erionite standards from bulk samples collected from the same locality and the application of systematic correction factors are necessary.

These findings highlight the need to establish environmental erionite fiber identification criteria based on SEM-EDX data and to develop advanced techniques for accurately identifying individual erionite fibers < 0.5 μm in air samples. Enhancing detection accuracy is essential for improving erionite environmental monitoring and occupational health assessments.