SEM-Based Approaches for the Identification and Quantification of Anhydrite

Abstract

For investigating and modeling the swelling potential of anhydrite rocks, it is important to define a fast but accurate, reliable, and repeatable procedure for mineral identification and quantification of anhydrite mineral in rock samples. We propose a quantitative evaluation of the applicability of two different SEM-based approaches (namely, image analysis and the use of the O/S atomic ratio) for the identification and quantification of anhydrite in polished slices of rock. We compare the results obtained with the bulk densities of the samples and with the outcomes of thermogravimetric analyses, demonstrating high convergence between the different data. We eventually propose a critical comparison between the proposed approaches and the existing methods, overall providing a practical guide for the selection of the best analytical procedure for the quantification of anhydrite content in rocks and, consequently, for the correct estimation of swelling potential.

1. Introduction

The swelling of the anhydrite is a phenomenon often occurring when sulfate-bearing rocks interact with anthropic infrastructures (e.g., civil and mining tunnels, bridges, and buildings, etc.), causing volume increase and heave of the rock mass, in turn resulting in structural damage, stability concerns, and increased maintenance costs (e.g., [1,2,3,4,5,6,7]). Interest in this topic has increased in recent years due to the frequent stratigraphic association of sulfate layers to big halite volumes that nowadays represent one of the most interesting solutions for underground energy storge.

The swelling process occurs when anhydrite (CaSO₄, the anhydrous phase of calcium sulfate) is dissolved and re-precipitated as gypsum (CaSO₄·2H₂O, the bi-hydrate phase of calcium sulfate) because of its higher solubility in water under environmental conditions. The growth of these new gypsum crystals along discontinuity surfaces in the rock mass causes the bulk increase in volume (swelling) [4,5,8].

Swelling phenomenon in anhydrites has been described and modeled by different research groups in recent years [8,9,10,11]. Despite a comprehensive and generally accepted understanding of the mechanism still missing, the main controlling factors are generally considered to be related to (i) mineralogical, (ii) textural, and (iii) geological/hydrogeological features [7].

Among mineralogical factors, the presence and proportion of anhydrite obviously play a key role in enabling and promoting the swelling. Conversely, the presence and percentage of gypsum have a dual role: on the one hand, the presence of pre-existing gypsum crystals favors the growth of new crystals, leading to a general increase in swelling potential [4]; on the other hand, if pre-existing gypsum forms a regular border along the main fractures’ surfaces (e.g., [12]), it may create a protective coating that prevents the further occurrence of the anhydrite-to-gypsum reaction, overall inhibiting the swelling [13]. In addition, several authors observed that the presence of clay, in a percentage ranging between 0% and 15%, represents a key factor for the development of swelling, despite the reasons still being unknown. Main hypotheses include the following [7]:

-

clay swelling, preceding sulfate swelling, disintegrates the rock and creates water pathways to anhydrite;

-

clay minerals deliver water to anhydrite thanks to osmosis processes and because of their higher porosity with respect to pure anhydrite;

-

adsorption and absorption of water in clay minerals helps to increase the concentration of calcium sulfates in water, respectively, by hindering water circulation and by subtracting water molecules;

-

clay minerals act as chemical catalysts.

With regard to textural factors, it was observed that an increase in the size of anhydrite crystals implies a reduction in swelling potential [14]. In general, indeed, smaller and more dispersed crystals exhibit a higher swelling potential compared to massive anhydrite or larger crystals, due to their greater surface area available for water interaction. In addition, the overall rock texture (e.g., sulfate-bearing claystone vs. massive or layered anhydrite) was observed to significantly influence swelling behavior [2]. This effect is (again) primarily due to differences in specific surface area, which in turn affect the extent to which water can interact with anhydrite crystals. Eventually, the role of the porosity in increasing the swelling potential was also highlighted [15].

In addition to these micro- and meso-scale features, the geological and hydrogeological context also plays a key role, as it influences water circulation patterns and, consequently, the amount of water that can reach swellable rocks [16,17]. Furthermore, the role of temperature in controlling the extent of swelling deformation of sulfate-bearing rock masses has also been highlighted [18,19].

Given this general state of the art, an accurate, reliable, and repeatable procedure for characterizing the mineralogical and textural features described above could represent a key resource for investigating real rock samples and for quantifying their swelling potential.

At this purpose, we quantitatively evaluated two different approaches based on the use of the Scanner Electron Microscope (SEM) associated with Energy Dispersive X-ray Spectroscopy (EDS) and image analysis techniques, respectively, for the identification and quantification of the anhydrite in polished slices of rock. These kinds of approaches offer several advantages over other existing methodologies, namely, X-ray powder diffraction (XRPD), optical microscopy, micro-Raman spectroscopy, and the thermogravimetric method. Indeed, all these other methodologies require sample preparation that may decrease the quality of the results. Grinding, which is required for XRPD and thermogravimetry, eliminates the rock’s original microstructure, making it impossible to obtain textural information (e.g., crystal size, gypsum crystal position, porosity) that can also significantly affect the swelling potential. In contrast, the preparation of thin sections, necessary for optical microscopy and micro-Raman spectroscopy, involves long water polishing, which could lead to anhydrite dissolution, especially when it occurs in small quantities.

The SEM- and SEM-EDS-based procedures that we propose significantly reduce the water exposure time during the sample preparation. They preserve both mineralogical and textural characteristics. In addition, these procedures have the additional advantage of using intact, undisturbed cylindrical samples that may be directly employed for further swelling tests.

We eventually compared the so-obtained SEM mineralogical percentages with the bulk densities of the samples, the mineralogical information obtained by XRPD, and the results of thermogravimetric analyses to validate the reliability of the proposed methodologies.

2. Materials and Methods

2.1. Tested Material

Rock material used in this study was sampled in a former gypsum quarry located north of Signols (Upper Susa Valley, Western Italian Alps). The quarry is excavated in the Gypsum Unit (also known as the Gypsum Nappe or Gypsum Group), a meta-evaporitic rock unit up to 400 m thick, dating to the late Triassic—e.g., [20]. In the Western Alps, this unit is structurally located beneath the Jurassic to Cretaceous Piemonte-Ligurian ophiolitic units, defining major tectonic contacts.

It is predominantly made of evaporites (gypsum/anhydrite), with layers of dolomite and micaschists. More specifically, in the sampling site, the evaporitic masses occur along the tectonic contact between the Piemonte-Ligurian units and the underlying continental Ambin Massif and other units with Briançonnais affinity (Figure 1).

In the quarry site, we distinguished and sampled portions that display different degrees of rehydration of the evaporite succession [12], respectively, consisting of distinct predominant proportions of anhydrite and gypsum (

Table A1. Characteristics of the samples analyzed in the present study: measured sizes, weight, and bulk density and modal content and average crystal size of the anhydrite obtained by image analysis.

Sample Name	Rock Type	Weight (g)	Height (mm)	Diameter (mm)	Volume (mm3)	Bulk Density (g/cm3)	Modal Content of Anhydrite	Average Size of Anhydrite Crystals (μm)
1	anhydrite	3.33	16.12	9.72	1196.4	2.78
2	anhydrite	3.44	15.58	9.91	1202.79	2.86
3	anhydrite	2.56	16.06	9.82	1215.78	2.11
4	anhydrite	3.33	16.06	9.82	1215.78	2.74
5	anhydrite	3.57	16.83	9.89	1292.03	2.76
6	anhydrite	3.41	15.47	9.91	1193.78	2.86
7	anhydrite	3.47	16.35	9.83	1240	2.8
8	anhydrite	3.53	16.2	9.89	1245.35	2.83
9	anhydrite	3.18	14.72	9.91	1135.65	2.8
10	anhydrite	2.98	13.66	9.89	1050.34	2.83
11	anhydrite	3.04	13.81	9.91	1065.66	2.86
12	anhydrite	3.34	15.42	9.87	1180.85	2.82
13	anhydrite	2.94	13.46	9.88	1030.98	2.85	0.73	185
14	anhydrite	2.46	11.5	9.86	877.8	2.8	0.64	222
15	anhydrite	2.7	12.64	9.84	961.88	2.81	0.60	195
16	anhydrite	2.66	12.33	9.86	941.09	2.83	0.50	192
17	anhydrite	3.07	14.51	9.84	1103.32	2.79	0.64	220
18	anhydrite	3.09	14.36	9.88	1100.44	2.81	0.64	233
19	anhydrite	2.91	13.5	9.87	1032.9	2.81	0.63	232
20	anhydrite	2.86	13.34	9.85	1016.18	2.82	0.57	228
21	anhydrite	2.94	13.84	9.82	1049.18	2.8	0.50	230
22	anhydrite	2.7	12.43	9.84	944.3	2.86	0.54	223
23	anhydrite	2.7	12.32	9.85	938.86	2.87	0.67	222
24	anhydrite	2.41	11.75	9.81	888.06	2.71	0.31	204
25	anhydrite	2.32	10.76	9.87	822.45	2.82	0.68	205
26	anhydrite	2.77	13	9.84	988.19	2.81	0.53	233
27	anhydrite	2.9	13.63	9.86	1041.34	2.78	0.54	213
28	anhydrite	2.71	12.47	9.86	952.55	2.84	0.61	227
29	anhydrite	2.44	11.71	9.84	890.86	2.74	0.52	227
30	anhydrite	2.56	11.88	9.84	903.38	2.83	0.59	190
31	anhydrite	2.73	12.73	9.83	965.53	2.82	0.66	211
32	anhydrite	2.23	10.36	9.83	786.51	2.84	0.69	217
33	anhydrite	missing data
34	gypsum	2.22	12.73	9.8	960.29	2.31
35	gypsum	2.45	14.26	9.81	1077.34	2.28
36	gypsum	2.53	14.43	9.84	1097.72	2.31
37	gypsum	2.24	13.12	9.81	991.91	2.26
38	gypsum	2.41	13.63	9.84	1036.87	2.33
39	gypsum	2.3	13.14	9.82	994.27	2.31
40	gypsum	2.14	12.42	9.81	938.68	2.28
41	gypsum	1.84	10.78	9.7	796.59	2.3
42	gypsum	1.47	8.5	9.61	616.75	2.39
43	gypsum	1.58	9.09	9.81	686.61	2.3
44	gypsum	1.35	7.9	9.7	583.19	2.31
45	gypsum	1.51	9.26	9.64	675.39	2.24
46	gypsum			missing data
47	anhydrite	2.97	13.6	9.87	1039.59	2.86	0.52	169
48	anhydrite	2.57	12.24	9.86	934.03	2.75	0.62	169
49	anhydrite	2.75	12.68	9.84	964.92	2.85	0.64	164
50	anhydrite	2.57	11.86	9.84	902.52	2.85	0.67	170
51	anhydrite	2.64	13.37	9.8	1009.09	2.62	0.39	165
52	anhydrite	2.57	12.55	9.86	957.55	2.69	0.53	166
53	anhydrite	2.61	13.7	9.84	1041.59	2.51	0.38	151
54	anhydrite	2.54	13.38	9.83	1016.13	2.5	0.45	165
55	anhydrite	2.77	13.53	9.84	1029.26	2.69	0.53	172
56	anhydrite	2.5	13.4	9.85	1020.75	2.45	0.23	145
57	anhydrite	2.71	13.43	9.83	1018.45	2.66	0.41	165
58	anhydrite	2.55	13.6	9.84	1033.79	2.47	0.26	153
59	anhydrite	2.64	13.6	9.84	1034.14	2.55	0.20	155
60	anhydrite	2.8	12.91	9.85	984.68	2.84	0.70	184
61	anhydrite	2.77	13.37	9.78	1003.79	2.75	0.57	159
62	anhydrite	2.72	13.88	9.7	1025	2.65	0.39	150
63	anhydrite	2.79	13.6	9.85	1035.89	2.7	0.51	155
64	anhydrite	2.4	12.52	9.65	916.08	2.62	0.21	127
65	anhydrite	2.71	12.45	9.84	946.39	2.87	0.65	178
66	anhydrite	2.93	14.4	9.82	1090.51	2.69	0.59	142
67	anhydrite	3.27	15.39	9.84	1171.15	2.79	0.65	173
68	anhydrite	3.02	14.61	9.81	1103.4	2.74	0.57	166
69	anhydrite	2.31	12.75	9.8	961.73	2.41	0.28	144
70	anhydrite	2.86	14.51	9.83	1101.45	2.59	0.61	140
71	anhydrite	2.8	12.7	9.83	963.58	2.91	0.67	168
72	anhydrite	2.03	11.23	9.81	848.77	2.39	0.32	129
73	anhydrite	2.12	11.69	9.81	883.27	2.4	0.19	133
74	anhydrite	2.49	11.54	9.83	876.64	2.84	0.54	152
75	anhydrite	1.99	10.49	9.78	787.23	2.52	0.33	140
76	anhydrite	2.54	11.71	9.85	892.57	2.84	0.58	153
77	anhydrite	2.52	11.66	9.8	879.51	2.86	0.71	155
78	anhydrite	3.15	16.74	9.82	1266.99	2.48	0.22	128
79	anhydrite	2.63	14.51	9.8	1095.48	2.4	0.30	122
80	anhydrite	2.79	13.13	9.8	989.97	2.81	0.69	152
81	anhydrite	2.82	13.26	9.82	1004.62	2.81	0.66	137
82	anhydrite	2.7	12.85	9.82	972.24	2.77	0.66	135
83	anhydrite	2.54	12.02	9.85	916.19	2.77	0.57	138
84	anhydrite	2.69	12.76	9.83	968.31	2.78	0.61	140
85	anhydrite	2.34	10.87	9.81	820.78	2.85	0.69	135
86	anhydrite	2.56	11.88	9.86	906.19	2.83	0.65	133
87	anhydrite	2.8	13.01	9.69	960.09	2.91	0.59	145
88	anhydrite	2.97	14.04	9.81	1060.58	2.8	0.63	148
89	anhydrite	2.55	12.19	9.89	936.45	2.72	0.64	132
90	anhydrite	3.15	14.67	9.79	1104.92	2.85	0.59	142

in Appendix A).

We resampled the collected rock material in the laboratory of the Department of Earth Sciences, University of Turin, by cutting and processing it with a laboratory core driller, obtaining 90 cylindrical samples. We then dry-polished one of the faces of each specimen (the base area of the cylinder) to ensure a flat and smooth surface. The cylindrical samples have mean sizes of 13.10 mm in height and 9.82 mm in diameter and a mean weight of 2.57 g (Figure 2). The data of caliper-measured sizes and high-precision scale masses for each sample are reported in Table A1. From these data, we retrieved the volume and bulk density for each sample, resulting in a mean volume of 993.21 mm³ and a mean bulk density of 2.68 g/cm³ (Table A1).

2.2. Experimental Methodology

2.2.1. Preliminary Mineralogical Characterization

For the purpose of preliminary mineralogical characterization, we performed an XRPD analysis of one sample considered as representative of the dataset (namely, Sample 12). Analysis was carried out in the laboratory of the Department of Earth Sciences at the University of Turin, using a SmartLab XE diffractometer coupled with Rigaku SmartLab Studio-II software and the Crystallography Open Database (COD) for mineralogical phase identification. We integrated the obtained results with SEM and SEM-EDS data to complete the general mineralogical phase identification and textural description of the material.

2.2.2. SEM-Based Image Analysis Procedure for the Quantification of the Anhydrite

We imaged 76 out of the 90 samples using a JEOL JSM IT300LV scanning electron microscope (SEM) at the University of Turin, acquiring a total of 6 to 8 images per sample. Images were acquired at magnifications of 40×, 100×, and 200× in backscattered electron (BSE) mode. All the images are available in the Supplementary Materials. Samples were not covered with conductive metals, as in the regular SEM routine, to guarantee the possibility of further experimentally assessing their swelling potential in water. For this reason, the SEM was operated under a low vacuum pressure of 30 Pa, and images were acquired in backscattered electron mode (BSE) with an accelerating voltage of 15 kV.

We then analyzed the images of anhydrite samples (64 out of the 76 imaged samples) acquired at 40× magnification with the Fiji/ImageJ software [22]. For each image, we first applied a ‘mean’ filter with a 2-pixel radius to reduce the noise/signal ratio. Then, we performed a black and white segmentation to differentiate the areas of the sample based on compositional differences. In detail, the segmentation aimed at distinguishing anhydrite, being the dominant mineralogical phase in most of the samples, from gypsum and other mineralogical phases (silicates and carbonates). The threshold for the segmentation was manually selected for each picture. All the segmented images are available in the Supplementary Materials. Eventually, by using the ‘Histogram’ function of the ImageJ software, we determined the percentage of pixels classified as anhydrite, considering it as a reasonable approximation of the overall percentage of anhydrite in the sample.

To complete the analysis, we also calculated the average size of the anhydrite crystals by averaging the size of at least 10 crystals for each image. Results are available in Table A1 in Appendix A. The availability of high-quality SEM-BSE images additionally allows obtaining not only the mean size of the crystals but also the full grain size distribution curve, as demonstrated in previous studies (e.g., [23]). Indeed, the size of the crystals is an inherently variable parameter that is better characterized by a probabilistic distribution than by a single value. This represents a useful resource since, in recent years, the use of probabilistic approaches to describe geometrical attributes of rocks and rock masses is being acknowledged as one of the best possible approaches in several fields of rock mechanics (e.g., [24,25]).

2.2.3. Use of O/S Atomic Ratio for the Discrimination of Anhydrite from Gypsum

Since EDS analysis provides information about the elemental composition rather than the mineralogical phase, anhydrite (CaSO₄), and gypsum (CaSO₄·2H₂O), which consist of the same elements, cannot be easily distinguished in the EDS spectra. Nevertheless, the difference in the number of water molecules implies that the atomic ratio between oxygen and sulfur (namely, O/S) will be different for the two minerals, having theoretical O/S ratios of 4 and 6, respectively.

To test the applicability of this theoretical concept for mineralogical identification, we performed a total of 122 EDS analyses on the samples and compared the results with the gypsum/anhydrite assignment made independently based on the grayscale and the textural characteristics.

EDS analyses were conducted using an Oxford Instruments EDS system with Aztec software. The analyses were performed under low vacuum conditions (30 Pa) with a beam accelerating voltage of 15 kV and a probe current of 2 nA. Prior to analysis, the system was calibrated using a cobalt standard. The duration of the analyses typically lasted between 20 and 60 s.

2.2.4. Thermogravimetric Analysis

After the detailed SEM investigation, we ground two samples (namely, samples 56 and 86) for thermogravimetric analyses. We first pre-dried the ground samples at 40 °C for 24 h to remove residual water content. Afterward, we performed two sequential steps of weighing and oven heating (105 °C for 48 h and 200 °C for 72 h, respectively). The weight loss during heating corresponds to the percentage of crystallization water released during the dehydration of gypsum and can therefore be used to calculate the gypsum content in the samples.

After the thermal treatments, the samples were dissolved in distilled water. The remaining insoluble fraction was dried and weighed to retrieve the percentage of insoluble minerals and, by subtraction, the percentage of anhydrite.

3. Results

3.1. Preliminary Mineralogical Characterization

The X-ray diffraction analysis performed on Sample 12 (Figure 3) reveals a heterogeneous mineralogical composition characterized by the coexistence of sulfates (anhydrite and gypsum), carbonates (dolomite and magnesite), and silicate (phlogopite) phases. The dominant crystalline phase is anhydrite, which exhibits the most intense diffraction peaks, suggesting that this mineral represents the primary component of the sample. The presence of gypsum, detected in minor amounts, may indicate partial hydration of the anhydrite. The presence of carbonates (dolomite and magnesite) is probably related to the sedimentary prototype of the rock, while the presence of phyllosilicates (phlogopite) is consistent with its metamorphic history. The optimized pattern demonstrates a good match between experimental data and reference diffraction profiles, confirming the reliability of the phase identification. The relative intensities of the diffraction peaks confirm that sulfate and carbonate minerals represent the main constituents of the sample, with accessory phases occurring in smaller proportions (Figure 3).

The data obtained from SEM-EDS analyses confirmed that anhydrite samples (see

Table 1. Comparison of anhydrite content in samples 56 and 86 as determined by thermogravimetric analysis and SEM-based image analysis.

Sample Name	Gypsum (%)	Calcium Sulfate Hemihydrate (%)	Anhydrite (%)	Insoluble Phases (%)	Anhydrite—Estimated from SEM-Based Image Analysis (%)
56	28.21	28.35	24.98	18.47	22.50
86	5.51	4.41	74.56	15.53	64.60

) are predominantly composed of anhydrite, occurring as idiomorphic to subhedral crystals, ranging in size from tens to hundreds of microns. These crystals frequently show sharp grain boundaries and widespread intracrystalline fracturing, creating a network of veins that are filled with secondary gypsum. This indicates a process of hydration of the anhydrite, suggesting that the samples underwent recent (post-metamorphic) alterations due to the circulation of water-rich fluids.

In gypsum samples (see Table 1), the gypsum aggregates have a massive structure, and the crystals are generally anhedral. We noticed a lesser fracturation in these samples, probably due to a more advanced stage of the recrystallization of calcium sulfate.

In addition to anhydrite and gypsum, SEM imaging and SEM-EDS analyses (Figure 4) revealed the presence of dolomite, magnesite, albite, and phyllosilicates, with traces of celestine and pyrite. Carbonate phases, such as dolomite and magnesite, appear as dark gray to black areas in the SEM images and are heterogeneously distributed within the gypsum and along the edges of anhydrite crystals. These non-evaporitic minerals are organized in layers that contribute to creating a bedding stratification in the rock.

The mineralogical and textural features of non-sulfate minerals are similar in anhydrite and gypsum samples, except for the magnesite, which was only identified in anhydrite samples.

The overall texture reflects significant recent reworking, with evidence of dissolution and recrystallization driven by interactions with circulating fluids (meteoric water—Figure 4a,b).

3.2. SEM-Based Image Analysis Procedure for the Quantification of the Anhydrite

Figure 5 shows two examples of the results of the image segmentation performed with Fiji/ImageJ software. As can be seen, the thresholding operation is quite effective in distinguishing between anhydrite crystals (white pixels) and other mineralogical phases, such as gypsum, dolomite, magnesite, and terrigenous minerals, and/or pores, when occurring (black pixels).

The complete list of anhydrite percentages resulting from the analysis of the segmented images is available in Table A1 in Appendix A. In Figure 6, we compare these values of anhydrite percentages with the bulk density of each sample. Despite some scatter, the graph depicts a clear linear correlation between the two variables (R² = 0.72), consistent with the higher specific density of anhydrite (~2.98 g/cm³) compared to gypsum (~2.32 g/cm³).

3.3. Use of O/S Atomic Ratio for the Discrimination of Anhydrite from Gypsum

Figure 7 shows the atomic percentages of sulfur (S) and oxygen (O) obtained from EDS measurements and used here as reference parameters to distinguish between gypsum and anhydrite. Based on the observed morphology and grayscale contrast, we associated each data point with the expected mineralogical phase (gypsum or anhydrite) and represented them as red and yellow dots, respectively.

The figure clearly highlights two distinct data clusters corresponding to gypsum and anhydrite. Gypsum data shows a higher oxygen/sulfur ratio, consistent with the presence of structural water in its crystal lattice, while the anhydrite points exhibit a lower O/S ratio, consistent with their dehydrated nature. The two clusters can be clearly distinguished by lines corresponding to O/S ratios of 6 and 5, respectively (dashed black lines in Figure 7). The area within the two lines includes, on the other hand, a mix of gypsum and anhydrite, suggesting that results in this range of O/S ratio cannot be reliably resolved by EDS analysis.

3.4. Thermogravimetry

The results of the thermogravimetric analyses are presented in Table 1. We obtained anhydrite contents of 24.98% and 74.56% for samples 56 and 86, respectively. For comparison, Table 1 also includes the anhydrite percentages previously estimated through SEM-based image analysis, being 22.50% and 64.60% for samples 56 and 86, respectively.

The results for sample 56 demonstrate strong consistency between the two methods, with a discrepancy of less than 2%. For sample 86, the difference between the two methods is greater (around 10%), but still indicates a reasonable level of agreement, supporting the reliability of the proposed approach.

4. Discussion

4.1. Interpretation of the Results of SEM-Based Image Analysis Procedure

To enable a more thorough interpretation of the anhydrite percentages obtained through image analysis, let us assume that the non-anhydrite mineral fraction consists of a fixed percentage of non-evaporitic minerals and a variable percentage of gypsum, or, in other words, that the gypsum content increases as the anhydrite content decreases. This assumption is consistent with the observed geological context, with non-evaporitic minerals that create a pervasive compositional layering consistently present in all the sites, while the quantity of gypsum depends on the quantity of dissolved and re-precipitated anhydrite. For the sake of simplicity, we also assumed that the porosity was zero, consistent with the SEM observations that returned a low to very low presence of pores.

Figure 8 reports, alongside the experimental dataset shown in Figure 6, three linear trends that relate the anhydrite content with the bulk density. These trends assume three different percentages of non-evaporitic minerals (40%, 20%, and 5%, respectively) and define the calculated bulk density, provided that the described assumption is accepted. As shown, these three lines effectively represent the upper boundary, the best fit, and the lower boundary of the entire dataset. This suggests that the variability of the dataset may be related to different percentages of non-evaporitic minerals in the samples, which, in turn, influence the bulk density.

The good agreement between these simulated lines and the experimental data overall supports the general reliability of the results and of the proposed methodological approach in estimating the quantity of anhydrite in real rock samples.

4.2. Interpretation of the Results of the Use of O/S Atomic Ratio

The use of the O/S ratio for discriminating EDS spectrograms of anhydrite and gypsum illustrated in Figure 7 returned reliable results. Nevertheless, there is a deviation between the theoretical O/S ratios of anhydrite and gypsum of 4 and 6, respectively, and the empirical values that have been observed to delimit the fields of anhydrite and gypsum (5 and 6, respectively). This upward shifting of the O/S ratio can be attributed to the possibility that the EDS sensor detects not only the gypsum/anhydrite mineral but also some other mineralogical phases located beneath or adjacent to the targeted mineral. Since these other mineralogical phases are likely to contain oxygen atoms, it is reasonable to assume that the O/S ratio tends to be higher than expected.

Similar considerations can also be applied to explain the region between O/S ratios 5 and 6, where gypsum and anhydrite data overlap. Indeed, the noise created by other revealed mineralogical phases to the O/S ratio may create an over-position of gypsum and anhydrite data, making it impossible to discriminate the spectrograms in those conditions.

4.3. Critical Evaluation of the Proposed Methodologies

We proposed two methodological approaches, respectively, dedicated to the mineralogical identification of anhydrite (O/S ratio) and to the quantification of its modal content (image analysis). As discussed in the introduction, these approaches aim to overcome the limitations of existing methods, mainly related to the necessity of preparing the samples through grinding or thin section, which, respectively, imply the loss of the rock’s original texture and structure or long water polishing that could lead to anhydrite dissolution.

Our results demonstrated good performance of the proposed methodologies. However, we also identified some disadvantages that need to be listed and discussed to enable the reader to select the best analytical approach for each specific context and objective.

In detail, the image analysis approach, despite having the advantage of a simple and open-source workflow, is quite time-consuming, since the thresholds need to be manually selected for each image. This also implies a certain degree of subjectivity in the analysis, resulting in a lowering of the data repeatability. In addition, in case of the presence of high tenors of minerals with higher brightness with respect to the anhydrite, as, for example, celestine in the considered material, a second threshold should be introduced. A possible solution to these issues may be the use of fixed thresholds for the entire dataset. However, this would require maintaining the same contrast and brightness parameters for the acquisition of all the images, which can be challenging. Future development of this research could include the implementation of machine learning approaches that would overcome these limitations, reducing the processing time, increasing the repeatability of the data, and allowing for the identification, in the images, of several mineralogical phases.

As previously discussed, the proposed use of the O/S ratio for discriminating between gypsum and anhydrite has the main disadvantage of not always resolving the ambiguity (i.e., data in the range with an O/S ratio between 5 and 6). In the remaining part of the dataset, we obtained a high degree of confidence in the results, with few data points that fall outside the expected O/S threshold ranges (i.e., 3 data points with O/S ratios exceeding 6 and 9 with O/S ratios below 5, representing approximately 10% of the total dataset). Compared to other methods for mineralogical identification (e.g., micro-Raman spectroscopy, XRPD, thermogravimetry), the error is still higher, but the O/S method has the advantage of the contextual acquisition of BSE-SEM images.

The presented results, however, only represent a preliminary validation of the proposed methodological approaches, since we considered a single case study. A complete validation would require the application also on different kinds of gypsum/anhydrite rocks (e.g., with different rock textures and structures different relative contents of anhydrite, gypsum, and other minerals) to verify the boundaries of practical applicability of the proposed approaches.

5. Conclusions

We proposed a quantitative evaluation of two SEM-based approaches (namely, image analysis and the use of O/S atomic ratio) for the identification and quantification of anhydrite in rock samples. The use of SEM has some advantages with respect to other possible approaches for mineralogical identification, since it guarantees the imaging of the microstructure and minimizes, at the same time, the water-sample time contact and the related possible dissolution of anhydrite during sample preparation.

Anhydrite percentages obtained by image analysis showed linear proportionality with the bulk densities of the sample. The variability in the trend can be due to minor changes in the content of non-evaporitic minerals. Results were additionally compared with anhydrite percentages obtained by thermogravimetric analysis, generally showing good agreement.

The mineral identification performed by using the ratio between O and S atomic percentages as a parameter for the discrimination between gypsum and anhydrite showed good reliability, provided that the ratio is lower than 5 or higher than 6. Indeed, in the range between O/S = 5 and O/S = 6, we observed data overlapping that prevents the mineral identification with this approach.

Overall, the results described good reliability of the proposed approaches for the identification and quantification of anhydrite in rock samples. The methodologies described in this paper can therefore represent a useful instrument for the investigation of swelling phenomena in real sites.