Assessment of the Swelling Potential of the Brebi, Mera, and Moigrad Formations from the Transylvanian Basin Through the Integration of Direct and Indirect Geotechnical and Mineralogical Analysis Methods

Abstract

This study evaluates the swelling potential in clayey soils of the Paleogene Brebi, Mera, and Moigrad formations in the Transylvanian Basin (Romania) by integrating direct free-swelling tests (FS; STAS 1913/12-88) with indirect index-property diagrams and semi-quantitative X-ray diffraction (XRD; RIR method). The indirect analysis combines three swelling-susceptibility classification charts—Seed et al. (AI–clay), Van der Merwe (PI–clay), and Dakshanamurthy and Raman (LL–PI)—with mineralogical trends from the Casagrande plasticity chart, complemented by Holtz and Kovacs’s clay-mineral reference fields and Skempton’s activity concept (AI = PI/% < 2 µm). The geotechnical dataset comprises 88 Brebi, 46 Mera, and 263 Moigrad specimens (with parameter counts varying by test), an XRD was performed on a representative subset. The free swell (FS) results indicate that Brebi soils range from low to active behavior (50–135%) without reaching the very active class; most Brebi specimens fall in the medium-activity range. Moigrad spans the full FS spectrum (20–190%) but is predominantly in the medium-to-active range. In contrast, Mera soils exhibit predominantly active behavior, covering the full range of activity classes (30–170%). The empirical classification charts diverge systematically: clay-sensitive schemes tend to assign higher swell susceptibility than the LL–PI approach, especially in carbonate-influenced soils. XRD results corroborate these patterns: Brebi is calcite-rich (mean ≈ 53.5 wt% CaCO₃) with minor expandable minerals (mean ≈ 3.1 wt%); Mera is feldspathic (orthoclase mean ≈ 55.3 wt%) with variable expandable phases; and Moigrad has a higher clay-mineral content (mean ≈ 38.8 wt%). Overall, swelling is controlled by the combined effects of clay-fraction reactivity, clay volume continuity, and carbonate-related microstructural constraints.

1. Introduction

Expansive soils present a major challenge in geotechnical engineering due to their tendency for substantial volumetric changes in response to moisture variations. This swelling potential is governed primarily by the clay fraction’s mineralogical composition and the abundance of clay-size particles in the soil matrix. Among clay minerals, smectite dominates because of its layered structure, high cation-exchange capacity, and strong affinity for water, all of which promote interlayer expansion upon wetting. In contrast, minerals such as illite and kaolinite have much lower swelling capacities due to their more stable lattice structures and limited interlayer hydration.

Clay content also exerts a significant influence: soils with higher clay content can absorb more water and develop greater swelling pressure, whereas soils with a low clay fraction may exhibit limited expansion even if swelling minerals are present. Thus, expansive behavior arises from the combined effects of mineralogical composition, fine-fraction content, and soil microstructure. Water interacts with clay minerals in complex ways: water adsorbed on clay surfaces influences plasticity, while water entering the clay interlayers drives swelling (as in smectite).

Several studies have shown that clay mineralogy can predict geotechnical index properties. For example, Schmitz et al. [1] used an equivalent basal-spacing concept to correlate clay assemblages with Atterberg limits. Other studies combining XRD (and occasionally SEM) with geotechnical tests have linked smectite content and smectite/kaolinite ratios to soil consistency, compressibility, and swelling [2]. Bo et al. [3] highlighted in marine clays an inverse correlation between kaolinite content and liquid limit or activity. In Romania, Marat et al. [4] reported formation-scale mineralogical and physico-mechanical correlations for the Moigrad Formation near Cluj-Napoca, emphasizing smectite and mixed-layer illite–smectite as key factors controlling slope reactivation.

In the study area, the Romanian design code NP 126:2010— “Normative regarding the foundation of structures on soils with high swelling and shrinkage potential” [5] —recognizes predominantly moderate to high swelling and shrinkage potential in analyzed soils. This underscores the need for detailed geotechnical investigations to accurately characterize the expansive behavior of these soils.

In this study, we first estimated the mineralogical composition indirectly using the Casagrande plasticity chart [6], Holtz and Kovacs’s clay-mineral reference fields [7], and Skempton’s activity chart [8]. We then performed semi-quantitative XRD analyses on representative samples to validate these estimates and to quantify the proportions of minerals (smectite, illite, kaolinite, chlorite, carbonate) that significantly influence expansiveness. Comparing the XRD results with the indirect geotechnical indicators confirmed the mineralogical trends at the formation scale.

The swelling potential was assessed using both indirect empirical systems (Seed et al. (1962) [9], Van der Merwe (1964) [10], and Dakshanamurthy and Raman (1973) [11] and direct measurements (free swell index, FS, according to STAS 1913/12-88 [12]). This integrated approach—combining geotechnical testing, mineralogical characterization, and empirical classification—provides a robust framework for evaluating expansive behavior and informing geotechnical risk management in the heterogeneous Paleogene clay deposits of the Transylvanian Basin.

2. Geological Overview

The Transylvanian Basin (Figure 1) is bordered by the Meseş Mts. in the north-west, Preluca-Țicău crystalline massif to the north, the Eastern and Southern Carpathians to the east and south, respectively, and the Apuseni Mountains to the west [13]. The progressive development of the Transylvanian Basin started in the Late Cretaceous as a result of the principal deformations within the Carpathian belt [13,14] and continued for over 60 million years until the Pliocene [15,16,17].

The evolution of the Transylvanian Basin has been marked by regional tectonic events, which influenced the sedimentation processes. The accumulated stratigraphic succession (locally over 5 km in thickness) presents major regional unconformities dividing the succession into four main tectonostratigraphic groups or megasequences, as follows: (i) Late Cretaceous, rift-related and gravitational collapse; (ii) Paleogene, sag-type basin evolution; (iii) Lower Miocene, flexural basin due to the final emplacement of the Pienides nappes; and (iv) Middle–Late Miocene, controlled by gravitational tectonics in a backarc setting [16].

This study focuses on lithostratigraphic units within the Paleogene megasequence, namely from the deposits of the second sag-type basin phase. During this period, a transgressive and regressive cycle governed the deposition mechanisms within Priabonian and Lower Rupelian. The deposited sequence is named the Turea group and follows a continental unit known as Valea Nadășului Formation. It is made up of the following main lithological assemblages: evaporites (Jebuc Formation), shallow-marine carbonates (Cluj Formation), outer shelf marls (Brebi Formation), shallow-marine sands (Mera Formation), ending with continental fluvial deposits (Moigrad Formation) (Figure 2) [16,22].

3. Materials and Methods

3.1. Samples

Soil samples were collected during geotechnical borehole investigations following ISO standards. The sampling area spans from the southwestern Cluj-Napoca municipality, through northern Florești commune, to Sânpaul commune (Cluj County) (Figure 1). Geologically, this area lies in the northwestern Transylvanian Basin and includes Late Priabonian–Rupelian Paleogene sediments. Based on lithostratigraphic interpretations and published data [22,23,24], samples were assigned to three well-documented formations: (i) the Brebi Formation [25], (ii) the Mera Formation [26], and (iii) the Moigrad Formation [27]. All these formations are predominantly fine-grained and clayey, which is reflected in the analyzed samples. A total of 88 Brebi, 46 Mera, and 263 Moigrad specimens were examined, with the number of determinations per parameter varying according to specimen quality and test applicability.

3.2. Methods

3.2.1. Geotechnical Parameters

The geotechnical characterization of the collected soil samples involved the determination of primary parameters through laboratory testing conducted in accordance with EN ISO 17892 standards. In addition, an activity index (AI) was calculated as the ratio between the plasticity index and the clay fraction (<2 µm), i.e., AI = PI/(% clay fraction < 2 µm), following Skempton (1953) [8].

The geotechnical parameters—including plasticity characteristics, grain-size distribution, and carbonate content—together with the testing methodology and associated laboratory procedures are described in detail by Crișan et al. (2026) [28], which focused on the geotechnical characterization of the investigated formations. In the present study, only the results of these determinations are reported, as they are used as input data for: (i) the direct assessment based on free swell (FS) and (ii) empirical classification schemes for swelling potential derived from index properties (Atterberg limits, clay fraction, and activity). These derived empirical indicators are interpreted jointly with the free swell test results and the mineralogical composition of the soils. Specifically, the clay fraction used in the activity index calculation corresponds to the <2 µm particle-size fraction, derived from the grain-size distribution dataset reported in [28].

Besides the parameters determined according to EN ISO standards, an important parameter used for the characterization of the swelling potential of the soils, the free swell, was determined following Romanian standard STAS 1913/12-88 [12] to evaluate the soil’s capacity to increase in volume as a result of water absorption. The test involves introducing a dry soil sample, previously passed through a 0.2 mm sieve and weighing 12 g (corresponding to an initial specimen volume of 10 cm³), into a 100 mL graduated cylinder. An initial volume of 50 mL of distilled water was added to the sample, after which the contents were agitated and left to settle for 4 h. The sample was then re-agitated, topped up with distilled water to the 100 mL mark, and left to settle for an additional 24 h. The free swelling was calculated as the percentage ratio between the increase in sediment volume and the initial volume of the soil sample. For interpretation and inter-formation comparison, FS values were grouped into swelling-activity classes according to NP 126:2010 [5], using the threshold intervals specified by the standard: low-activity clays (FS < 70%), medium-activity clays (FS = 70–100%), active clays (FS = 100–<140%), and very active clays (FS ≥ 140%).

In addition to the free swell (FS) determinations, swell susceptibility was evaluated using three empirical classification systems: Van der Merwe (1964) [10], Seed et al. (1962) [9], and Dakshanamurthy and Raman (1973) [11]. For each specimen, the swelling-potential class was assigned by plotting the relevant index parameters on the method-specific charts: (i) Van der Merwe (PI vs. clay fraction, <2 µm) [10], (ii) Seed et al. (AI vs. clay fraction) [9], and (iii) Dakshanamurthy and Raman (LL vs. PI) [11]. For cross-formation comparison, the outcomes were summarized as percentage distributions across swelling-potential classes. In parallel, the same specimens were grouped into FS-based activity categories (NP 126:2010 thresholds [5]), and the plotted points were coded according to these FS categories to enable a direct comparison between the empirical assessments and the measured volumetric response.

While these empirical schemes provide categorical estimates of swell susceptibility, they do not explain why specimens assigned to the same class may display different measured swelling responses.

Therefore, the empirical outcomes were complemented by statistical association analyses between FS and the main index parameters (PI, LL, clay fraction, AI) and CaCO₃ content, to identify the dominant controls on swelling intensity and to evaluate the role of microstructural constraints across formations.

For interpretative purposes, specimens were additionally grouped into carbonate-content classes according to EN ISO 14688-2:2017 [29], i.e., non-calcareous (CaCO₃ < 1%), slightly calcareous (1–5%), calcareous (5–25%), highly calcareous (25–50%), and very highly calcareous (CaCO₃ > 50%).

3.2.2. Mineralogical Analyses

XRD was performed on bulk (random powder) mounts and on oriented clay preparations (air-dried, ethylene glycol-solvated, and heat-treated at 400 °C and 550 °C). Specimens were prepared at the Department of Geology at Babeș-Bolyai University (Cluj-Napoca, Romania), according to the glass slide technique described by Moore and Reynolds [30].

XRD analyses were performed on a limited subset of specimens selected to be representative of each formation in terms of lithological level and index-property variability (plasticity range and carbonate-content classes). “Natural” (bulk) mounts were used to identify the main crystalline phases in the whole material, while oriented preparations (silt + clay fraction obtained by decantation/centrifugation) combined with glycolation and heating were used to confirm the presence of expandable clay minerals and to support clay-mineral identification. Because only a subset was analyzed by XRD and the XRD subset does not fully overlap with the specimens subjected to FS testing, XRD results are used primarily as a mineralogical benchmark that constrains the interpretation qualitatively, rather than as a basis for direct quantitative mineralogy–FS regressions.

Carbonate content was determined by two independent approaches: Routine gasometric measurements reported in the geotechnical dataset and semi-quantitative bulk XRD (RIR) estimates of carbonate phases. Because XRD was performed on a reduced representative subset and the XRD specimens do not fully overlap with the specimens used for the geotechnical determinations, the two datasets cannot be compared on a one-to-one sample basis. At formation scale, Brebi shows consistent evidence of carbonate enrichment across both approaches, supporting the interpretation that high carbonate content is a robust characteristic of this unit. In contrast, for the Mera and Moigrad formations, carbonate proportions inferred from the XRD subset may differ from the gasometric dataset, which is interpreted primarily as a heterogeneity and sampling-coverage effect (carbonate occurring in variable interbeds and/or cemented horizons that may be under- or over-represented in the reduced XRD subset).

Air-dried oriented specimens for XRD were obtained using the decantation method based on Stokes’ law and extraction of silt and clay-sized fractions from the first 5 cm of the suspension prepared from each sample. The suspension containing the silt and clay fraction had been centrifuged at 3500 rotations/minute for 15 min. The concentrate of silt and clay from the bottom of the centrifuge jar was used to prepare the air-dried oriented specimen with 24 h sedimentation time. After the X-ray diffraction, the oriented samples were moved into the exicator for ethylene glycol solvation in vacuum for 24 h. Subsequently, each specimen was subjected to thermal treatment at 400 °C and then at 550 °C, each stage lasting 60 min; after each thermal treatment stage, the specimen was analyzed by X-ray diffraction (XRD).

The X-ray diffraction (XRD) investigations were performed using a Bruker D8 Advance diffractometer with Cu Kα radiation (λ = 1.54060 Å), Fe 0.01 mm filter, and a LynxEye one-dimensional detector. The working XRD parameters were 40 kV and 40 mA. The data were collected between 3.8° and 64° 2θ, with a step size of 0.02° 2θ with a counting time of 0.2 s/step. Both the Diffrac. Suite EVA program and PDF2 (version 2.1202) database from ICDD were used to identify the bulk mineral composition, and the clay minerals were identified in accordance with Moore and Reynolds [30], Poppe et al. [31], and Azizi et al. [32]. The mineral abbreviation is according to [33].

Semi-quantitative mineralogical analysis was performed using the Reference Intensity Ratio (RIR) method, following the formulation proposed by de Visser and Wolff [34]. Crystalline phases were identified by comparing the experimental diffractograms with reference entries from the Powder Diffraction File (PDF-2) database, based on the agreement of the positions and relative intensities of characteristic reflections. RIR quantification uses the I/Ic (RIR) values, defined as the ratio between the intensity of the most intense reflection of phase X and the intensity of the most intense reflection of corundum (standard): I/Ic = Imax(X)/Imax(corundum). I/Ic values can be determined experimentally under identical measurement conditions or obtained from databases; in this study, the I/Ic values were taken from PDF-2 (version 2.1202). Weight percentages (wt%) were estimated from the intensities of the reflections of the identified phases by determining their scale factors (least-squares fitting between the calculated and experimental intensities/profile) and normalized to the total of the identified crystalline phases. No internal standard was used; therefore, any amorphous component is not included in the quantitative balance. The results were interpreted as semi-quantitative, considering known factors that can affect the intensity–weight fraction relationship in XRD (e.g., preferred orientation, microabsorption, particle-size effects, and peak overlap). To provide qualitative constraints on mineralogical tendencies, two complementary index-property–based approaches were applied. For each specimen, indirect mineralogical indications were obtained by: (i) plotting the data on the Casagrande plasticity chart [6] and interpreting the specimen position relative to the indicative clay-mineral fields proposed by Holtz and Kovacs (1981) [7]; and (ii) applying Skempton’s (1953) activity concept [8], whereby the activity index (AI) was calculated and interpreted using the activity chart to classify clays as inactive, normal active, or active.

For cross-formation comparison, the outcomes of both approaches were summarized as distributions of specimens across the inferred fields/classes, and the plotted points were additionally coded according to the FS-based swelling-activity categories adopted in this study. This enabled a coherent visual and statistical comparison between index-property–based indications and the measured volumetric response captured by FS. Because both empirical approaches provide indirect inferences and may be influenced by material texture/fabric, cementation (e.g., carbonate bonding), and water chemistry, they were used strictly as qualitative indicators and interpreted jointly with the FS results and XRD-based mineralogical determinations.

4. Results

Classification of the soil samples was performed based on physical properties determined through laboratory testing, in accordance with EN ISO 14688-1 [35] and 14688-2 standards [29]. In this paper, the results obtained for each of the identified geological formations—Brebi, Mera, and Moigrad—are presented in terms of minimum, maximum, average, and both lower and upper characteristic values.

4.1. Geotechnical Parameters

The grain size distribution, liquid limit (LL), plasticity index (PI), and CaCO₃ content used in this study were previously determined and reported for the same sites and lithological units by Crișan et al. (2026) [28]. The corresponding testing methodology and laboratory procedures are described in detail in the cited publication; therefore, only the relevant data are summarized here for completeness. The previously obtained results were integrated into the present study to support the interpretation of newly acquired experimental and numerical data.

To obtain a comprehensive evaluation of expansive behavior, the geotechnical dataset was supplemented by determining the activity index (AI) and the free swell index (FS). These parameters allow for the quantification of the reactivity of the clay fraction and the soil’s sensitivity to moisture variations, constituting essential elements in analyzing the mechanisms of swelling and shrinkage and in correlating them with mineralogical composition and microfabric structure.

4.1.1. Brebi Formation

From a geotechnical and lithological perspective, the Brebi Formation is mainly composed of gray clays, predominantly of medium plasticity (CIM), with local intercalations of high-plasticity clays (CIH) and isolated occurrences of low-plasticity clays (CIL) (Figure 3). The deposits exhibit a very stiff consistency and are predominantly very highly calcareous, with intervals of highly calcareous material and locally slightly calcareous horizons (data following Crișan et al. (2026) [28]).

For the Brebi Formation, the activity index (AI) ranges from 0.52 to 1.51, with an average of 0.78, while the free swell (FS) varies between 50% and 135%, with an average value of 86.7% (

Table 1. Statistical evaluation of physical parameters for the Brebi Formation.

	Clay	Liquid Limit	Plasticity Index	Free Swell	Carbonate Content	Activity Index
	A2u	LL	PI	FS	CaCO3	AI
	[%]	[%]	[%]	[%]	[%]	-
Number of analyzed samples	48	83	83	85	88	47
Vx	0.211	0.179	0.231	0.263	0.423	0.265
Minimum value	25.35	28	15	50	14.33	0.52
Average value	40.26	44	29	87	53.75	0.78
Maximum value	66.88	62	47	135	97.58	1.51
Xk,inf	37.62	42	27	80	46.70	0.72
Xk,sup	42.89	47	31	94	60.81	0.85

Values for A_2u, LL, PI, and CaCO₃ are according to Crișan et al. (2026) [28].

). According to the NP 126:2010 classification [5], these values correspond to soils ranging from low to active, indicating a moderate to high expansive behavior within the clayey deposits of this formation.

The assessment of the swelling potential of the Brebi Formation based on empirical classification systems indicates the following distributions: according to Van der Merwe (1964) [10], 47% of the samples are classified as high expansive, 40% as very high expansive, and 13% as medium expansive (Figure 3d); according to Seed et al. (1962) [9], 83% of the samples fall in the high-swelling category, 15% in the medium category, and 2% in the very high category (Figure 3b); and according to Dakshanamurthy and Raman (1973) [11], 70% of the samples exhibit medium swelling potential, 19% high, and 11% low (Figure 3c).

When grouped by FS classes (NP 126:2010 [5]), the statistical distribution indicates that 22.4% of the samples fall within the low-activity category (FS < 70%), 41.2% exhibit medium activity (FS = 70–100%), and 36.5% are classified as active (FS = 100–140%). There are no samples classified as very active soils, with FS ≥ 140%, in the case of the Brebi Formation.

Regarding the relationship between calcium carbonate (CaCO₃) content and expansive indices, samples classified as highly calcareous (CaCO₃< 50%) exhibit average values of approximately 0.96 for the activity index (AI) and 91.8% for free swell (FS), while soils classified as very highly calcareous (CaCO₃ ≥ 50%) display lower values, with an activity index of 0.78 and a free swell of 80.4% (Figure 4a,b). Given the limited number of samples falling into the calcareous class (CaCO₃ between 5% and 25%), they were grouped, for statistical purposes, with the highly calcareous class (25–50%).

4.1.2. Mera Formation

The Mera Formation is composed mainly of gray, high-plasticity clays (CIH), with subordinate layers of very-high- and medium-plasticity clays (CIV and CIM) and locally developed low-plasticity (CIL) horizons (Figure 5a). Thin sandy intercalations are also present, typically a few tens of centimeters thick, but the sandy samples were excluded from the analysis in this paper. The deposits exhibit a very stiff consistency and are predominantly calcareous, with horizons ranging from slightly to highly calcareous, whereas very highly calcareous intervals occur only sporadically. The formation also includes a basal packstone-type limestone layer, known as the Hoia Limestone. For the geotechnical characterization, only the clayey deposits were considered; sandy and limestone intercalations were excluded due to their limited thickness and lack of statistical representativeness (data from Crișan et al. (2026) [28]).

For the Mera Formation, the activity index (AI) ranges from 0.50 to 4.02, with an average value of 1.75, while the free swell index (FS) ranges from 30% to 170%, with an average of 100.33% (

Table 2. Statistical evaluation of physical parameters for the Mera Formation.

	Clay	Liquid Limit	Plasticity Index	Free Swell	Carbonate Content	Activity Index
	A2u	LL	PI	FS	CaCO3	AI
	[%]	[%]	[%]	[%]	[%]	-
Number of analyzed samples	46	41	41	45	35	40
Vx	0.338	0.485	0.616	0.331	0.581	0.562
Minimum value	13.36	28	16	30	2.84	0.50
Average value	30.96	71	53	100	23.88	1.75
Maximum value	68.59	160	135	170	51.06	4.02
Xk,inf	27.71	61	43	90	19.58	1.44
Xk,sup	34.20	82	63	111	28.18	2.05

Values for A_2u, LL, PI, and CaCO₃ are according to Crișan et al. (2026) [28].

). According to the classification defined in NP 126:2010 [5], these results correspond to soils ranging from low active to very active, indicating a wide variability of expansive behavior within the clayey deposits of the formation.

The assessment of the swelling potential of the Mera Formation based on empirical classification systems indicates the following distributions: according to Van der Merwe (1964) [10], 45.95% of the samples are classified as high swelling potential, 37.84% as very high swelling potential, and 16.22% as medium swelling potential (Figure 5d); according to Seed et al. (1962) [9], 57.5% of the samples fall in the high swelling potential range, 27.5% in the very high swelling potential range, and 15% in the medium swelling potential range (Figure 5b); and according to Dakshanamurthy and Raman (1973) [11], 41.46% of the samples exhibit high swelling potential, 24.39% medium swelling potential, 21.95% extreme swelling potential, 9.76% very high swelling potential, and 2.44% low swelling potential (Figure 5c).

Regarding carbonate content, samples were grouped into calcareous (CaCO₃ < 25%) and highly calcareous (CaCO₃ ≥ 25%) classes. Mean values decrease from AI = 2.31 and FS = 114.7% (calcareous) to AI = 1.54 and FS = 100.5% (highly calcareous) (Figure 6a,b).

4.1.3. Moigrad Formation

The Moigrad Formation is composed primarily of reddish silty clays, with colors ranging from reddish to violet, brown, and brick-red, and generally exhibiting medium plasticity (CIM). Subordinate layers of high- and very high-plasticity clays (CIH and CIV) also occur, along with local horizons of low-plasticity clays (CIL) (Figure 7a). Sandy intercalations are present, with thicknesses varying from a few centimeters to several meters, but the sandy samples are excluded. The deposits exhibit a very stiff consistency. The formation is predominantly calcareous, with carbonate contents ranging from slightly to highly calcareous across frequent horizons (data following Crișan et al. (2026) [28]). Given the specific aims of this study, only the clayey samples were considered relevant for assessing swelling potential; consequently, sandy intercalations were excluded from the analysis due to their non-expansive behavior and limited geotechnical relevance in this context.

For the Moigrad Formation, the activity index (AI) ranges from 0.33 to 3.74, with an average value of 1.09, while the free swell index (FS) varies between 20% and 190%, with an average of 91% (

Table 3. Statistical evaluation of physical parameters for the Moigrad Formation.

	Clay	Liquid Limit	Plasticity Index	Free Swell	Carbonate Content	Activity Index
	A2u	LL	PI	FS	CaCO3	AI
	[%]	[%]	[%]	[%]	[%]	-
Number of analyzed samples	263	254	254	245	214	253
Vx	0.362	0.278	0.404	0.310	0.613	0.504
Minimum value	12.06	31	13	20	0.40	0.33
Average value	35.58	52	35	91	15.81	1.09
Maximum value	68.65	115	98	190	43.04	3.74
Xk,inf	31.58	48	30	82	12.81	0.92
Xk,sup	39.57	57	39	99	18.82	1.26

Values for A_2u, LL, PI, and CaCO₃ are according to Crișan et al. (2026) [28].

). According to the NP 126:2010 classification [5], these values correspond to soils ranging from low active to very active, indicating a broad variability of swelling potential within the clay deposits of the formation.

The assessment of the swelling potential of the Moigrad Formation based on empirical classification systems indicates the following distributions: according to Van der Merwe (1964) [10], 44.49% of the samples are classified as high swelling potential, 29.53% as very high swelling potential, 19.69% as medium swelling potential, and 6.30% as low swelling potential (Figure 7d); according to Seed et al. (1962) [9], 51.18% of the samples exhibit high swelling potential, 27.56% medium swelling potential, 19.29% very high swelling potential, and 1.97% low swelling potential (Figure 7b); and according to Dakshanamurthy and Raman (1973) [11], 55.12% of the samples exhibit medium swelling potential, 30.31% high swelling potential, 8.66% very high swelling potential, 3.15% low swelling potential, and 2.76% extremely high swelling potential (Figure 7c).

Regarding the relationship between calcium carbonate (CaCO₃) content and expansive indices, samples classified as slightly calcareous (CaCO₃ < 5%) exhibit average values of AI = 1.49 and FS = 82%, whereas samples in the calcareous class (CaCO₃ = 5–25%) show lower average AI (AI = 0.99) but higher free swell (FS = 94%). Samples classified as highly calcareous (CaCO₃ = 25–50%) display the same average activity index as the calcareous class (AI = 0.99) and lower free swell (FS = 78%) (Figure 8a,b). Samples with CaCO₃ contents below 1% were included in the slightly calcareous class for statistical evaluation.

4.2. Mineralogical Analysis

A mineralogical study was performed on samples collected from the Brebi, Mera, and Moigrad formations. In total, eight XRD analyses were carried out for the Brebi Formation, twelve for the Mera Formation, and eleven for the Moigrad Formation. In addition, a representative (characteristic) XRD diffractogram for each formation is provided. The XRD subset was selected to capture the main mineralogical end-members expected within each unit (carbonate-richer vs. carbonate-poorer intervals and variability in plasticity), while keeping the dataset feasible for detailed, oriented-sample treatments (EG solvation and heating). As such, XRD is used to support formation-scale mineralogical interpretation and to contextualize FS-based swelling behavior, rather than to provide specimen-by-specimen mineralogical prediction of FS.

4.2.1. Brebi Formation

Semi-quantitative XRD data for the Brebi Formation point to a carbonate-dominated mineral assemblage with a relatively limited clay-mineral component. Calcite is the major phase (35.54–76.25 wt%, mean 53.52 wt%), accompanied by quartz (6.71–18.46 wt%, mean 12.39 wt%) and minor plagioclase (0–4.69 wt%, mean 1.19 wt%). The clay-mineral pool (montmorillonite + montmorillonite–chlorite interstratifications + illite/muscovite + clinochlore + kaolinite) averages 31.53 wt% (range 12.38–51.34 wt%) and is largely controlled by illite/muscovite (8.62–45.18 wt%, mean 24.82 wt%), with chlorite (clinochlore) and kaolinite occurring in minor amounts (means 1.98 wt% and 1.65 wt%, respectively). Expandable phases are consistently minor in bulk terms: montmorillonite (0.56–1.51 wt%, mean 1.02 wt%) and montmorillonite–chlorite interstratifications (1.00–2.69 wt%, mean 2.06 wt%), giving total potentially expandable phases of 1.67–3.78 wt% (mean 3.08 wt%) (

Table 4. Relative abundance of major minerals in the Brebi Formation: statistical summary from XRD data.

	Montmorillonite	Montmorillonite–Chlorite Interstratifications	Clinochlore	Illite/Muscovite	Kaolinite	Quartz	Calcite	Plagioclase Feldspar	Gypsum
	Mnt	Mnt-Chl	Ccl	Ilt/ Ms	Kln	Qz	Cal	Pl	Gyp
	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]
Number of analyzed samples	8	8	8	8	8	8	8	8	8
Vx	0.294	0.308	0.690	0.511	0.678	0.366	0.262	1.527	2.362
Minimum value	0.56	1.00	0.00	8.62	0.00	6.71	35.54	0.00	0.00
Average value	1.02	2.06	1.98	24.82	1.65	12.39	53.52	1.19	1.37
Maximum value	1.51	2.69	3.54	45.18	3.15	18.46	76.25	4.69	9.22
Xk,inf	0.82	1.63	1.07	16.33	0.90	9.35	44.13	0.00	0.00
Xk,sup	1.22	2.48	2.90	33.32	2.39	15.42	62.92	2.41	3.53

).

The X-ray diffraction patterns obtained from the natural, oriented, and ethylene glycol-solvated specimens of sample P6 from the Brebi Formation (Figure 9) indicate the presence of basal reflections typical of smectite (14.33 Å in the oriented preparation and ≈16.92 Å in the ethylene glycol-solvated specimen), illite (≈10.00 Å), kaolinite (7.15 Å), and chlorite (7.06 Å). Significant amounts of calcite and quartz are also present in the analyzed sample (Table 4).

The ethylene glycol-solvated preparation of the clay fraction from sample P6 shows a shift in the montmorillonite (001) reflection from approximately 14.33 Å to about 16.92 Å, while the (001) reflections of illite and kaolinite (≈7.15 Å), as well as the (002) reflection of chlorite (≈7.07 Å), remain unchanged.

The diffraction patterns obtained from the clay fraction of sample P6 heated at 400 °C and 550 °C, respectively, indicate the collapse of the montmorillonite reflection from ≈16.92 Å to ≈10 Å (accompanied by an increase in illite peak intensity), a slight increase in the intensity of the chlorite reflection at ≈14.20 Å when heated to 550 °C, the persistence of the 7.10–7.15 Å kaolinite and chlorite reflections at 400 °C, and their disappearance upon heating to 550 °C.

The indirect interpretation of the mineralogical composition based on physical geotechnical parameters indicates that most samples plot within the montmorillonitic or montmorillonite-dominated domains on the Casagrande (1948) plasticity chart [6] supplemented with Holtz and Kovacs (1981) clay mineral zones [7] (Figure 10a), whereas the Skempton (1953) activity chart [8] yields an average activity index of approximately 0.78, typically associated with illitic clays (Figure 10b).

4.2.2. Mera Formation

The semi-quantitative XRD dataset for the Mera Formation indicates a bulk mineralogy dominated by framework silicates, with an average clay-mineral pool broadly comparable to Brebi, yet contrasting sharply in carbonate abundance and in the dispersion of expandable components. Orthoclase is the prevailing phase, spanning 22.12–79.46 wt% (mean ≈ 55.31 wt%), with quartz as the main accessory constituent (2.91–21.75 wt%, mean ≈ 10.04 wt%) and locally elevated plagioclase contents (0–18.92 wt%, mean ≈ 1.58 wt%). Crystalline calcite is generally scarce (0–15.21 wt%, mean ≈ 1.33 wt%) and is absent from most specimens, except for a single carbonate-bearing outlier.

The clay-mineral pool (montmorillonite + montmorillonite–chlorite interstratifications + illite/muscovite + chlorite + kaolinite) averages ≈ 31.68 wt% but ranges widely from 6.39 to 53.25 wt%. It is typically dominated by illite/muscovite (1.66–49.22 wt%, mean ≈ 23.24 wt%), accompanied by variably developed chlorite (1.19–7.74 wt%, mean ≈ 2.80 wt%) and minor kaolinite (0–6.13 wt%, mean ≈ 1.42 wt%).

Among the three investigated formations, Mera exhibits the greatest variability in expandable phases. Montmorillonite varies from 0.34 to 3.50 wt% (mean ≈ 1.63 wt%), while montmorillonite–chlorite interstratifications span 0.76–6.64 wt% (mean ≈ 2.60 wt%). Consequently, the combined potentially expandable fraction ranges from 1.10 to 9.90 wt% (mean ≈ 4.23 wt%) (

Table 5. Relative abundance of major minerals in the Mera Formation: statistical summary from XRD data.

	Montmorillonite	Montmorillonite–Chlorite Interstratifications	Clinochlore	Illite/Muscovite	Kaolinite	Quartz	Calcite	Plagioclase Feldspar	Gypsum	Orthoclase Feldspar
	Mnt	Mnt-Chl	Ccl	Ilt/ Ms	Kln	Qz	Cal	Pl	Gyp	Or
	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]
Number of analyzed samples	12	12	12	12	12	12	12	12	12	12
Vx	0.663	0.649	0.696	0.643	1.141	0.663	3.295	3.464	3.464	0.293
Minimum value	0.34	0.76	1.19	1.66	0.00	2.91	0.00	0.00	0.00	22.12
Average value	1.63	2.60	2.80	23.24	1.42	10.04	1.33	1.58	0.06	55.31
Maximum value	3.50	6.64	7.74	49.22	6.13	21.75	15.21	18.92	0.77	79.46
Xk,inf	1.04	1.69	1.74	15.14	0.54	6.43	0.00	0.00	0.00	46.52
Xk,sup	2.21	3.51	3.85	31.34	2.30	13.64	3.70	4.54	0.18	64.11

).

The XRD analysis of the natural, oriented, and ethylene glycol-treated preparations of the samples (Figure 11) reveals basal reflections typical of smectites (approximately 14.60 Å in the oriented preparation and ≈16.88 Å in the ethylene glycol-solvated specimen), illite (≈10.00 Å), and chlorite (7.10 Å)/kaolinite (7.17 Å). Significant quantities of feldspars and quartz are also present.

The ethylene glycol-solvated specimens confirm the typical expansion of smectite (montmorillonite) from ≈14.45–14.60 Å to ≈16.88 Å, while the (001) reflections of illite and kaolinite/chlorite remain unchanged. Heating treatments at 400 °C and 550 °C reveal the complete collapse of the smectite peak to ≈10 Å, accompanied by an increase in illite peak intensity. The kaolinite and chlorite reflections (≈7.10–7.17 Å) remain stable at 400 °C but are destroyed upon heating to 550 °C.

The indirect interpretation of the mineralogical composition based on physical–geotechnical parameters indicates that most Mera Formation samples plot in the transition zone between the illite and montmorillonite domains on the Casagrande (1948) plasticity chart [6], supplemented with the Holtz and Kovacs (1981) clay mineral reference fields/lines [7], clustering mainly along the montmorillonite trend and, subordinately, within the montmorillonite domain (Figure 12a). In contrast, the Skempton (1953) activity chart [8] yields an average activity index of approximately 1.75, typically associated with active clays and suggesting a stronger smectitic (montmorillonitic) tendency than in the Brebi Formation (Figure 12b).

4.2.3. Moigrad Formation

The semi-quantitative XRD results for the Moigrad Formation are consistent with a siliciclastic framework characterized by a comparatively enlarged clay-mineral pool and systematically low carbonate contents in the analyzed subset. Quartz is abundant, ranging from 10.32 to 66.10 wt% (mean ≈ 33.82 wt%), accompanied by substantial plagioclase (7.15–29.97 wt%, mean ≈ 15.87 wt%) and orthoclase with pronounced variability (3.88–37.18 wt%, mean ≈ 10.71 wt%). Crystalline calcite is largely absent (0–9.19 wt%, mean ≈ 0.84 wt%), with only a single minor carbonate outlier.

The clay-mineral pool is, on average, higher than in Brebi and Mera (mean ≈ 38.76 wt%; range 16.70–66.83 wt%) and is typically dominated by illite/muscovite (6.26–59.17 wt%, mean ≈ 28.12 wt%), supplemented by chlorite (2.69–6.75 wt%, mean ≈ 4.17 wt%) and minor kaolinite (0–2.23 wt%, mean ≈ 1.12 wt%).

Expandable phases are consistently present, with montmorillonite at 0.92–3.11 wt% (mean ≈ 1.78 wt%) and montmorillonite–chlorite interstratifications at 1.56–6.50 wt% (mean ≈ 3.57 wt%), yielding a combined potentially expandable fraction of 2.48–9.55 wt% (mean ≈ 5.35 wt%) (

Table 6. Relative abundance of major minerals in the Moigrad Formation: statistical summary from XRD data.

	Montmorillonite	Montmorillonite–Chlorite Interstratifications	Clinochlore	Illite/Muscovite	Kaolinite	Quartz	Calcite	Plagioclase Feldspar	Orthoclase Feldspar
	Mnt	Mnt-Chl	Ccl	Ilt/ Ms	Kln	Qz	Cal	Pl	Or
	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]	[%]
Number of analyzed samples	11	11	11	11	11	11	11	11	11
Vx	0.439	0.529	0.341	0.813	0.822	0.579	3.317	0.447	0.885
Minimum value	0.92	1.56	2.69	6.26	0.00	10.32	0.00	7.15	3.88
Average value	1.78	3.57	4.17	28.12	1.12	33.82	0.84	15.87	10.71
Maximum value	3.11	6.50	6.75	59.17	2.23	66.10	9.19	29.97	37.18
Xk,inf	1.34	2.51	3.37	15.30	0.61	22.83	0.00	11.89	5.39
Xk,sup	2.22	4.63	4.96	40.95	1.64	44.81	2.39	19.85	16.03

).

The X-ray diffraction analysis of the natural, oriented, and ethylene glycol-solvated preparations of sample P12 (Figure 13) reveals the presence of the smectite (001) reflection (d ≈ 15.21 Å) in the oriented sample, which expands to ≈16.98 Å after ethylene glycol-solvation. Illite (d ≈ 9.97 Å) and kaolinite (d ≈ 7.15 Å) are also identified, together with quartz, feldspars, and calcite. The reduced intensity of the quartz 3.34 Å peak in the oriented and ethylene glycol-solvated preparations compared with the natural sample is attributed to its rapid sedimentation.

The diffraction patterns of the clay fraction of sample P12 heated to 400 °C and 550 °C (Figure 13) show the complete collapse of the montmorillonite peak from ≈16.98 Å to ≈10 Å, accompanied by an increase in the illite (001) reflection. The (001) kaolinite reflection at ≈7.16 Å disappears upon heating.

The indirect interpretation of the mineralogical composition based on physical–geotechnical parameters indicates that most Moigrad Formation samples plot in the transition zone between the illite and montmorillonite domains on the Casagrande (1948) plasticity chart [6], supplemented with the Holtz and Kovacs (1981) clay mineral reference lines [7], with subordinate clustering within the montmorillonite domain and isolated occurrences within the illite domain (Figure 14a). In contrast, the Skempton (1953) activity chart [8] yields an average activity index of approximately 1.09, typically associated with illitic clays, with a subordinate contribution of montmorillonite and kaolinite (Figure 14b).

5. Discussion

To characterize the geotechnical behavior of the soil formations in terms of swelling potential, an integrated approach was applied, combining empirical methods of indirect assessment and direct experimental geotechnical parameter determinations. In this context, the classification systems proposed by Dakshanamurthy and Raman (1973) [11], Van der Merwe (1964) [10], and Seed et al. (1962) [9] were employed, which correlate the liquid limit (LL), plasticity index (PI), clay fraction (<2 µm), and activity index (AI) with swelling susceptibility. The direct assessments included free swell (FS) measurements, providing a quantitative characterization of expansive behavior.

Indirect techniques included plotting of the samples on the Casagrande (1948) plasticity chart [6], supplemented with characteristic lines for major clay minerals (Holtz and Kovacs, 1981 [7]), and correlating the activity index (AI) with clay content, following the Skempton (1953) classification [8]. Indirect indices (LL, PI, AI, and clay fraction <2 µm) provide a behavioral proxy for clay–water interactions, but they are non-unique: they integrate mineralogy with fabric, cementation, and dilution by non-clay phases. Therefore, mineralogical inference from Casagrande [6]/Skempton [8] charts should be treated as probabilistic rather than deterministic. XRD offers an independent benchmark by directly identifying the crystalline phases and by quantifying (semi-quantitatively) the relative abundance of expandable components. To make the two approaches comparable, the most relevant metric is not the bulk wt% of smectite alone (which can be strongly diluted by quartz/feldspars/carbonates), but the expandable proportion within the clay-mineral pool, i.e., (montmorillonite + mixed-layer montmorillonite–chlorite) normalized by total clay minerals (expandables + illite/muscovite + chlorite + kaolinite). This normalization links directly to AI, which is also defined per unit clay fraction. Within this framework, agreement (or divergence) between XRD and indirect indices can be interpreted mechanistically: high swelling may reflect either (i) a higher expandable fraction and/or (ii) a higher volumetric continuity of the clay matrix, while carbonates primarily act as microstructural constraints (cementation/aggregation) that decouple mineralogical potential from the expressed free swell.

Accordingly, the mismatch between index-based swelling susceptibility and measured free swell (FS) reflects physico-chemical and microstructural mediation of clay–water interaction rather than lithology alone. Index property charts (LL–PI, PI–clay, and AI–clay content) are proxies that combine mineralogical predisposition with fabric state, cementation, and dilution by non-clay phases; they can therefore indicate a high susceptibility even when the macroscopic strain is partly restrained. This is particularly relevant in carbonate-affected deposits, where CaCO₃-related bonding/cementation and aggregate stiffening reduce the volumetric continuity and deformability of the active clay matrix and may limit water access at the specimen scale. Consequently, FS captures the effective expression of swelling under the existing pore-scale conditions and fabric constraints, whereas empirical indices primarily reflect a potential of swelling.

A brief link to experimental work also supports this distinction between mineralogical potential and its effective expression. In a laboratory model test on desiccated expansive clay, Hamza and Ikin (2020) [36] showed that electro-osmotic rehydration can rapidly increase average water content (by more than 100% within 8 h) and reduce suction (by about 93% within 5 h, ultimately to ~1 kPa by the end of treatment), while the desiccation-induced shrinkage was mostly recovered. These macroscopic changes occur over hours through moisture transfer and suction neutralization; therefore, the observed volume-change response can be interpreted primarily in terms of moisture redistribution and suction neutralization, without requiring mineralogical transformation to explain the short-timescale response. This behavior is consistent with our formation-scale interpretation that FS integrates mineralogical predisposition with clay-matrix continuity and microstructural constraints (including carbonate bonding/cementation) that modulate how swelling strains are expressed.

Mixed-layer montmorillonite–chlorite (Mnt–Chl) interstratifications indicate the presence of partially expandable phyllosilicate assemblages, in which smectitic interlayers alternate with chloritic (non-expandable) layers. Such mixed-layer structures may still contribute to swelling because hydration of the smectitic component promotes interlayer expansion; however, the chloritic component tends to impose microstructural restraint, such that the expressed swelling is commonly lower than for separate montmorillonite phases at comparable bulk abundance. X-ray diffraction (XRD) is suitable for identifying Mnt–Chl through characteristic basal-reflection behavior in oriented mounts, including shifts upon ethylene glycol solvation and collapse/thermal response upon heating. Nevertheless, in semi-quantitative bulk (RIR-based) datasets, the precise proportion of smectitic layers within Mnt–Chl cannot be robustly constrained due to peak overlap, preferred orientation effects, and variability in mixed-layer ordering. Accordingly, Mnt–Chl is grouped here with montmorillonite as part of the potentially expandable fraction, and its relevance is interpreted at formation scale rather than used for specimen-level prediction of free swell (FS).

5.1. Brebi Formation

For the Brebi Formation, free swell values range from 50% to 135% (Table 1), indicating a low-to-active swelling activity, without reaching the “very active” domain (FS ≥ 140%) defined by NP 126:2010 [5]. According to NP 126:2010 [5], based on free swell behavior the formation comprises soils with low activity (FS < 70%), medium activity (FS = 70–100%), and active soils (FS = 100–140%), while no very active soils (FS ≥ 140%) were identified. The distribution across FS classes shows that most specimens fall within the medium- and active-activity classes (41.2% and 36.5%, respectively), whereas the low-activity category is less represented (22.4%). This distribution suggests that expansivity is not an isolated behavior controlled by a few particular strata, but rather a recurrent property of the formation, expressed to varying degrees depending on internal deposit variations (fine-grained lithology, degree of cementation, and matrix homogeneity/heterogeneity).

The classification of specimens by FS classes shows that, as free swell increases, the materials tend to occur more frequently within higher-plasticity domains (from CIL/CIM towards the CIM–CIH range). This trend is consistent with an increased water-retention capacity and the more pronounced role of the fine parts in governing volumetric behavior. However, the activity index (AI) does not exhibit the same pattern: the mean values remain relatively similar across the FS classes and generally fall within the range proposed by Skempton (1953) [8] for inactive to normally active clays (Figure 10b). This contrasting evolution of FS and AI suggests that the intensification of swelling within the Brebi Formation does not necessarily reflect a major shift in the dominant clay mineralogy, but rather the combined effects of clay-fraction content (continuity of the fine matrix and water distribution within the mass) and microstructural controls (e.g., localized stiffening/cementation), which may either amplify or constrain the volumetric response to wetting.

Building on this interpretation, it is useful for the Brebi Formation to distinguish between: (i) the influence of the clay fraction and (ii) the structural conditions that govern how this potential is expressed at the material scale. In practical terms, the activity index (AI) (Skempton, 1953) [8] serves as an indicator of the plastic behavior of the <2 µm fraction normalized by clay content, and thus primarily reflects the nature and reactivity of this fraction. By contrast, free swell (FS) directly quantifies macroscopic volumetric deformation, integrating not only mineralogical effects but also the amount of fine matrix, its distribution within the soil skeleton, and intergranular bonding (cementation) that may impose constraints on deformation. In this context, for Brebi, the fact that AI remains on average within the range of inactive to normally active clays, whereas FS locally indicates active soil behavior (NP 126:2010 [5]), supports the view that expansivity variability is controlled mainly by clay-fraction proportion and microstructure, rather than by a major change in the expansive mineralogy (Figure 10b).

Against this background, calcium carbonate (CaCO₃) should be interpreted as a factor that attenuates the manifestation of expansivity through structural effects, rather than as a direct predictor of swelling potential. The statistical correlations indicate a tendency for both FS and AI to decrease with increasing CaCO₃, which is mechanistically plausible through two complementary processes: (i) intergranular carbonate cementation, which increases skeleton stiffness and limits volume changes of the fine matrix; and (ii) localized coating/impregnation of clay surfaces by carbonate phases, which reduces water access and diminishes the effective hydration of aggregates. As a result, more carbonate-rich specimens tend to exhibit a more constrained volumetric response, even where the clay fraction remains sufficient to generate swelling. Therefore, in the Brebi Formation, increasing continuity of the clay matrix (i.e., a higher <2 µm fraction) promotes higher FS, whereas carbonates act in the opposite sense by reinforcing the microstructure and reducing volumetric deformability (Figure 4a,b; [37]).

In parallel with the direct assessment provided by FS, empirical classification schemes also confirm the expansive character of the Brebi Formation, while revealing systematic differences among methods (Figure 3b–d). Approaches that explicitly account for clay content tend to yield higher estimates of swell susceptibility: the Van der Merwe (1964) method [10], based on the PI–clay fraction relationship, and the Seed et al. (1962) method [9], which relates AI to clay content, place a substantial proportion of specimens in the higher expansivity classes. By contrast, the Dakshanamurthy and Raman (1973) classification [11], founded solely on Atterberg limits (LL, PI), results in a more moderate rating for the same materials.

This divergence is geotechnically explainable and, for the Brebi Formation, appears relevant to the distinction between swelling potential and its effective expression. Methods that are sensitive to the clay fraction (Seed [9]; Van der Merwe [10]) better capture the “volumetric” component of susceptibility, i.e., the capacity of a more abundant fine matrix to retain water and generate larger strains. However, for Brebi, the direct FS response indicates a medium- to highly swelling activity without reaching the “very active” domain (FS ≥ 140%), suggesting the presence of microstructural constraints on swelling. In a material with elevated carbonate content, cementation/stiffening effects may limit dilation of the fine matrix and reduce the magnitude of volumetric deformation. Consequently, clay-inclusive classifications can be interpreted as more conservative—reflecting a higher potential—whereas the LL–PI-based classification (Dakshanamurthy and Raman [11]) may better approximate the effective behavior evidenced by FS under the existing fabric/structure. Therefore, for the Brebi Formation, the combined use of these methods is justified: Seed [9] and Van der Merwe [10] characterize susceptibility associated with the fine fraction, while Dakshanamurthy and Raman [11] provide an estimate closer to the observed global response when microstructure (including the influence of CaCO₃) modulates the expression of swelling (Figure 3b–d and Figure 4a,b).

Building on the interpretation that distinguishes between the volumetric response (FS) and clay-fraction activity indicators (AI), the mineralogical inferences drawn from empirical charts should be understood as having different levels of specificity. For Brebi, the key observation is a partial decoupling between swelling expression (FS) and clay-fraction “reactivity” indicators (AI): while FS spans 50–135% (NP 126:2010 [5] low-to-active range; no very active soils, FS ≥ 140%), AI remains, on average, within the inactive-to-normally active range. This pattern is consistent with a relatively stable clay-mineral assemblage across the formation, where differences in FS are driven predominantly by the amount/continuity of the fine matrix and fabric-related constraints, rather than by a formation-scale increase in expansive clay mineralogy. In carbonate-bearing clays, this decoupling is expected because swelling strain integrates mineralogical potential with the stiffness and bonding of the skeleton.

For the Brebi Formation, the Casagrande (1948) plasticity chart [6], supplemented with the indicative fields proposed by Holtz and Kovacs (1981) [7] (Figure 10a), suggests a smectitic influence on plastic behavior. However, this representation remains essentially behavioral: the LL–PI position is jointly controlled by overall plasticity, the amount of fines, and fabric-related features (aggregation/dispersion), and in carbonate-rich deposits it may also be influenced by microstructural effects (stiffening/cementation) that modify the expression of plasticity. Therefore, for Brebi, the signal in Figure 10a should be interpreted primarily as indicating that a smectitic component (even if subordinate) may affect plasticity, rather than implying that smectites mineralogically dominate the clay fraction at the formation scale. The indirect mineralogical fields on the Casagrande chart should therefore be interpreted cautiously for Brebi: LL–PI position reflects bulk plasticity behavior, but carbonate cementation and aggregate structure can shift the apparent plasticity response without requiring a major change in clay mineral species.

By contrast, for estimating the mineralogical composition of the clay fraction, Skempton’s activity chart [8] (Figure 10b) is arguably more informative and relevant, because it relates plasticity (PI) to clay content (<2 µm) and thus captures the reactivity of the clay fraction more directly. In Brebi, the placement of AI within the inactive-to-normally active clay range is consistent with the interpretation discussed above: a predominantly illitic mineralogical background, with local/subordinate smectitic contributions, upon which the expansive behavior measured by FS is amplified by the proportion and continuity of the fine matrix and modulated by structural effects associated with CaCO₃. Within this framework, the two charts are not contradictory, but they carry different evidential weight: the Casagrande chart [6] (with the Holtz and Kovacs fields [7]) provides a broad indication of plastic behavior, whereas the Skempton chart [8] offers a more robust basis for mineralogical inference regarding the <2 µm fraction (Figure 10b). The relatively modest AI values argue against a smectite-dominated clay fraction and are consistent with subordinate expandable contributions within an overall illitic background. Accordingly, for Brebi, Skempton’s activity concept [8] provides a closer formation-scale match to the XRD mineralogy than the Casagrande mineralogical fields (Casagrande [6] complemented by Holtz and Kovacs (1981) [7]).

For the Brebi Formation, X-ray diffraction (XRD) analysis (Brebi subset n = 8; Section 4.2) provides an independent mineralogical benchmark for the interpretations discussed above. The semi-quantitative XRD results show carbonate-rich samples (calcite 35.54–76.25 wt%, mean 53.52 wt%) and a clay-mineral pool dominated by illite/muscovite (8.62–45.18 wt%, mean 24.82 wt%) with minor clinochlore and kaolinite (means 1.98 wt% and 1.65 wt%, respectively), whereas expandable components are minor in bulk terms (montmorillonite 0.56–1.51 wt%, mean 1.02 wt%; montmorillonite–chlorite interstratifications 1.00–2.69 wt%, mean 2.06 wt%; total expandables 1.67–3.78 wt%, mean 3.08 wt%). To make the mineralogical signal comparable with AI (defined per unit clay fraction), the most relevant metric is the expandable proportion within the clay-mineral pool, i.e., (montmorillonite + montmorillonite–chlorite interstratifications) normalized by total clay minerals; under this normalization, the expandable fraction remains modest (mean expandable/clay minerals ≈ 0.10), consistent with AI clustering in the inactive-to-normally active range and with the absence of very active FS values (≥140%). At the same time, the wide variability in calcite supports the interpretation of carbonates as a microstructural control factor, capable of decoupling mineralogical potential from expressed free swell through cementation/stiffening effects.

It should be emphasized, however, that the XRD dataset does not overlap with the specimens tested geotechnically (including FS). Therefore, the mineralogical results can be used as a qualitative validation of the interpretative framework (illitic background with chloritic/kaolinitic contributions, subordinate smectites, and a modulating role of CaCO₃), but not as a basis for direct quantitative correlations between mineralogical percentages and swelling levels. In this respect, XRD serves as a mineralogical “benchmark” reinforcing the idea that, in Brebi, expansive behavior variability is more robustly controlled by the proportion of the fine fraction and microstructural constraints (including CaCO₃) than by large formation-scale variations in smectite content.

5.2. Mera Formation

Unlike the Brebi Formation—where FS remains within the low-to-active range—Mera spans the full NP 126:2010 spectrum [5], including very active soils. This broader distribution provides a clearer basis for tracking how plasticity, clay-fraction activity, and microstructural constraints jointly control swelling intensity.

For the Mera Formation, FS values range from 30% to 170%, spanning all four activity categories defined by NP 126:2010 [5]. The distribution across FS classes shows a predominance of active soils (FS = 100–140%), which account for 40.0% of all specimens, followed by medium-activity soils (FS = 70–100%) at 33.3%. Low-activity soils (FS < 70%) and very active soils (FS ≥ 140%) are equally represented, each comprising 13.3%. This distribution indicates that, at the formation scale, expansive behavior is a recurrent characteristic expressed over a broad range of intensities, rather than a response confined to a few isolated lithological levels.

As FS increases, specimens tend to occur more frequently within higher-plasticity domains, shifting from low-plasticity fields (predominantly CIL for FS < 70%) towards very high plasticity (CIV for FS ≥ 140%) (Figure 5a). This association does not imply a direct causal relationship between plasticity and FS; rather, it suggests that strata developing larger swell are generally characterized by a more plastic, water-sensitive fine matrix, which provides favorable conditions for the development of volumetric strains. In parallel, increasing FS is accompanied by a systematic rise in clay-fraction activity: mean AI values increase from approximately 0.65 (low-activity class) to 1.39 (FS = 70–100%), 1.93 (FS = 100–140%), and 2.63 for FS ≥ 140%. This increase in AI should be interpreted as an indicator of enhanced clay-fraction reactivity rather than as a direct statement of mineralogical dominance: even secondary smectite-group minerals or illite–smectite mixed-layer components can produce a disproportionate effect on swelling behavior, particularly when carbonate-related constraints are less pronounced than in Brebi and the fine matrix becomes volumetrically continuous. Because the mean clay content remains relatively stable for FS < 140% (≈27.5–30.6%), the increase in AI below the 140% threshold primarily indicates enhanced reactivity of the <2 µm fraction (i.e., mineralogical shifts toward a greater smectitic contribution and/or more favorable hydration/dispersion conditions), rather than merely an increase in total clay content. The FS ≥ 140% threshold, however, marks a different regime in which the mean clay content rises to ≈42.7%, suggesting that a volumetrically continuous fine matrix is reached—capable of amplifying mineralogical reactivity into macroscopic deformation.

An important control on how swelling potential is manifested is the calcium carbonate (CaCO₃) content, through carbonate cementation and associated microstructural constraints. Similar to the Brebi Formation, in the Mera Formation CaCO₃ can be interpreted as a proxy for the degree of carbonate cementation, with the potential to limit free swell by increasing structural stiffness. However, carbonation levels in Mera are markedly lower than in Brebi; therefore, CaCO₃ acts primarily as a modulating factor of the expansive response rather than as a formation-scale dominant control. Within this framework, the decrease in CaCO₃ with increasing FS (from ≈29% in the FS = 70–100% class to ≈21% for FS ≥ 140%) implies a reduction in carbonate cementation, which may allow a more pronounced expression of swelling. The analysis by carbonate-content classes supports the same trend, suggesting that at lower CaCO₃ contents the volumetric response is less constrained, whereas at higher contents stiffening effects become more evident (Figure 6a,b).

With respect to the indirect assessment of swell susceptibility, the empirical classification schemes applied to the Mera Formation show a consistent convergence toward high swelling-potential domains (Figure 5b–d). Methods that explicitly incorporate the clay fraction (Seed et al., 1962 [9]; Van der Merwe, 1964 [10]) classify a very large proportion of specimens within the high–very high categories (≈84–85%). The Dakshanamurthy and Raman (1973) [11] method, which is based solely on Atterberg limits, confirms the same tendency by assigning a high proportion of specimens to the high class and to the upper classes (very high + extra high), which together account for more than 70%. From an interpretative standpoint, the agreement between a plasticity-controlled method (Dakshanamurthy and Raman [11]) and methods that incorporate the volumetric effect of the fine fraction (Seed [9]; Van der Merwe [10]) suggests that, in Mera, swelling potential arises from a combination of high plasticity and the capacity of the fine matrix to retain water and facilitate moisture transfer through the material mass.

Regarding mineralogy inferred indirectly from physico-geotechnical parameters, the systematic FS–AI trend noted above indicates a progressively more reactive <2 µm fraction in the higher-FS classes. This behavior is consistent with an increasing contribution of expandable components (smectite and/or mixed-layer smectitic structures) and/or a progressive shift toward hydration-favorable fabric states (dispersion, reduced bonding), rather than being explained solely by changes in total clay percentage. Importantly, the Casagrande (1948) plasticity chart [6] captures this evolution as a shift toward higher-plasticity domains and toward the montmorillonitic trend; however, as in all mixed-mineral soils, the Casagrande position is behavioral and does not uniquely define mineralogy.

On the Casagrande (1948) plasticity chart [6], complemented by the clay-mineral reference fields/lines of Holtz and Kovacs (1981) [7], most specimens cluster in the transition zone between the illitic and montmorillonitic fields—typically along the montmorillonite line and, secondarily, within the montmorillonitic field (Figure 12a). Skempton’s activity concept [8] provides a more diagnostic indication for the <2 µm fraction because it normalizes PI by clay content; in Mera, the systematic AI increase with FS supports a progressively more reactive clay fraction in the higher-FS classes (Figure 12b).

The semi-quantitative XRD dataset provides an independent mineralogical benchmark at formation scale (Mera subset n = 12; Section 4.2). Bulk mineralogy is dominated by framework silicates, with orthoclase as the prevailing phase (22.12–79.46 wt%, mean ≈ 55.31 wt%), quartz as the main accessory constituent (2.91–21.75 wt%, mean ≈ 10.04 wt%), and locally elevated plagioclase (0–18.92 wt%, mean ≈ 1.58 wt%). Crystalline calcite is generally scarce (0–15.21 wt%, mean ≈ 1.33 wt%) and is absent from most specimens, with one carbonate-bearing outlier (Table 5). The clay-mineral pool (montmorillonite + montmorillonite–chlorite interstratifications + illite/muscovite + chlorite + kaolinite) averages ≈ 31.68 wt% (range 6.39–53.25 wt%) and is typically dominated by illite/muscovite (1.66–49.22 wt%, mean ≈ 23.24 wt%), accompanied by variably developed chlorite (1.19–7.74 wt%, mean ≈ 2.80 wt%) and minor kaolinite (0–6.13 wt%, mean ≈ 1.42 wt%). Among the studied formations, Mera exhibits the largest dispersion in expandable components: montmorillonite varies from 0.34 to 3.50 wt% (mean ≈ 1.63 wt%), while montmorillonite–chlorite interstratifications span 0.76–6.64 wt% (mean ≈ 2.60 wt%), yielding a combined potentially expandable fraction of 1.10–9.90 wt% (mean ≈ 4.23 wt%) (Table 5).

Taken together, the agreement between (i) AI increasing with FS and (ii) the presence and variability of expandable components in XRD supports the interpretation that the highest swelling in Mera reflects the combined effect of a reactive clay fraction and the attainment of a volumetrically continuous fine matrix in the highest-FS domain. The carbonate effect is best treated as locally dependent on occurrence and cementation state rather than as a uniform formation-scale control—especially given that the XRD subset is mostly low-calcite, while the broader geotechnical dataset documents variable CaCO₃ at formation scale (Table 2; Figure 6a,b).

5.3. Moigrad Formation

Moigrad also spans the full range of FS-based activity classes defined by NP 126:2010 [5]; however, its internal controls differ from Mera: a practical threshold emerges around FS ≈ 100%, above which the volumetric dominance of the fine matrix becomes decisive, while carbonate effects appear non-linear across compositional classes.

For the Moigrad Formation, FS values range from 20% to 190% (Table 3), spanning all four activity categories defined by NP 126:2010 [5]. The FS-class distribution is dominated by medium-activity soils (FS = 70–100%), which account for 37.1% of all specimens, followed by active soils (FS = 100–140%) at 33.9%. Low-activity soils (FS < 70%) represent 21.6%, whereas the very active category (FS ≥ 140%) is less common (7.4%). This distribution indicates that, at the formation scale, expansive behavior is a recurrent characteristic expressed predominantly within the moderate-to-high swell range, while very large swell values occur less frequently and are associated with particular lithological levels.

The classification across FS domains highlights a practical threshold at around FS ≈ 100%, which marks a shift in the degree of control exerted by the fine fraction on swelling response. For FS < 100%, the mean clay content remains nearly constant (≈27–28%), and CaCO₃ shows mean values on the order of ≈16%. Under these conditions, the transition from low-FS to medium-FS levels cannot be attributed to a quantitative increase in clay content or to a systematic change in carbonate content. Therefore, differences in swelling are more likely related to internal features of the fine matrix and the nature of the clay component, which can vary substantially even at comparable total clay percentages: lithological alternations involving a more reactive fine fraction, differences in aggregation/dispersion state, and variations in the distribution and connectivity of clay domains within the mass.

For FS ≥ 100%, however, a robust change emerges: mean clay content increases sharply to approximately 43–46%. From an interpretative standpoint, this is the clearest indication that, in Moigrad, high swelling no longer depends solely on the nature of the fine fraction, but also on its amount and continuity. Beyond this threshold, the clay matrix becomes sufficiently dominant to control the material’s volumetric response and to sustain large macroscopic expansive strains, through an increased water-retention capacity.

The relationship between FS and the activity index (AI) provides an important mechanistic clarification. Mean AI values remain relatively similar across the FS < 70%, FS = 70–100%, and FS = 100–140% classes—around ≈1—with a more evident increase observed only in the FS ≥ 140% domain. At the same time, the AI ranges are very wide, indicating pronounced internal heterogeneity: within the same FS domains, levels with less active fines can coexist with levels exhibiting much higher reactivity, including cases exceeding the AI > 1.25 threshold (Figure 14b). Technically, AI expresses the ratio between plasticity (PI) and the <2 µm clay content; thus, a large dispersion in AI indicates that, for comparable clay contents, the potential of the fine fraction to generate plasticity and wetting sensitivity differs substantially among specimens. This observation supports the interpretation that the transition into the active range (FS ≥ 100%) is primarily controlled by the volumetric effectiveness/continuity of the fine matrix, whereas very high swelling (FS ≥ 140%) additionally requires the occurrence of locally more reactive clay fractions and favorable microstructural conditions for strain expression.

For the Moigrad Formation, calcium carbonate (CaCO₃) can influence how swelling is expressed, primarily through its effects on microstructure and on the pathways of water ingress and redistribution within the material mass. Classifying the specimens into slightly calcareous (CaCO₃ < 5%), calcareous (5–25%), and highly calcareous (25–50%) groups highlights that the relationship between CaCO₃ and swelling-related indices is non-linear and cannot be described by a single trend over the full compositional range. Instead, it varies with CaCO₃ content and with how carbonates are distributed within the soil fabric (Figure 8a,b).

In the slightly calcareous class, the highest mean activity index is recorded (AI ≈ 1.49), together with a free swell of about 82%. This combination suggests that, when carbonate input is low, the response is dominated by the properties of the clay fraction, with minimal carbonate-related modification of the structure. In the calcareous class (5–25%), AI decreases to ≈0.99, whereas FS increases to ≈94%. This association shows that a decrease in AI does not automatically translate into reduced swelling; at this CaCO₃ level, the primary effect is likely related to fabric organization and local wetting permeability, such that water can penetrate and redistribute efficiently throughout the mass, promoting a larger overall swell (Figure 8a,b).

In the highly calcareous class (25–50%), AI remains essentially unchanged relative to the calcareous class (≈0.99), but FS drops to ≈78%. The persistence of AI alongside a decrease in free swell indicates the emergence of more pronounced structural constraints, consistent with intensified carbonate bonding that stiffens the skeleton and limits volumetric change of the fine matrix. In this compositional range, CaCO₃ becomes relevant mainly through its effect on fabric and stiffness, reducing the ability of the clay matrix to dilate upon wetting even under similar fine-fraction activity (Figure 8a,b).

Overall, the results suggest that the influence of CaCO₃ in Moigrad arises from the superposition of mechanisms whose effects vary across the compositional range: at lower to intermediate CaCO₃ contents, microstructural configuration and wetting conditions may enable high swelling even at lower AI, whereas at higher CaCO₃ contents the stiffening/bonding effect becomes dominant and constrains free swell. The dispersion of values among classes further indicates that swelling response remains controlled by multiple factors, with CaCO₃ acting in conjunction with fine-fraction heterogeneity and deposit-specific microstructural features (Figure 8a,b).

In addition to the direct FS assessment, the empirical classification schemes applied to the Moigrad Formation confirm formation-scale swell susceptibility, but allocate specimens differently across susceptibility classes (Figure 7b–d). The Dakshanamurthy and Raman (1973) [11] method assigns most specimens to the medium class (55.12%), followed by high (30.31%), whereas very high (8.66%) and extremely high (2.76%) are subordinate and low is marginal (3.15%) (Figure 7c). By contrast, the Van der Merwe (1964) [10] and Seed et al. (1962) [9] methods place a larger share of the dataset in the high and very high domains: 44.49% high and 29.53% very high for Van der Merwe [10], and 51.18% high and 19.29% very high for Seed [9] (Figure 7b,d), with the remaining specimens distributed between the medium and low classes.

From an interpretative standpoint, this discrepancy reflects distinct methodological sensitivities. Relative to the direct FS evaluation, the Dakshanamurthy and Raman (1973) [11] classification appears, at formation scale, the closest to the overall swelling signal, because its ranking is governed by plasticity indicators (LL, PI) that directly condition wetting-induced response across both low–medium activity regimes and active–very active regimes (NP 126:2010 [5]) (Figure 7c). At the same time, the transition into FS ≥ 100% regimes (active and very active soils in NP 126:2010 [5]) coincides with a robust increase in clay fraction. Under these conditions, clay-sensitive criteria (Seed [9]; Van der Merwe [10]) tend to highlight more clearly the elevated predisposition to swelling (Figure 7b–d). However, within these higher-FS regimes, the magnitude of FS may still be modulated by microstructural features and CaCO₃-related effects (Figure 8a,b), such that the plasticity signal expressed by Atterberg limits can separate specimens differently than criteria that explicitly incorporate clay content. Within this framework, Seed [9] and Van der Merwe [10] are particularly useful for identifying interbeds where the clay fraction becomes decisive, whereas Dakshanamurthy and Raman [11] provide a formation-scale reference that more closely tracks the FS response across the full activity range.

Regarding mineralogy inferred indirectly from physico–geotechnical parameters, the charts applied to the Moigrad Formation lead to a broadly convergent interpretation, albeit with different emphasis depending on the indicator considered. On the Casagrande (1948) plasticity chart [6], supplemented with the characteristic montmorillonite, illite, and kaolinite lines proposed by Holtz and Kovacs (1981) [7], most specimens cluster in the transitional zone between the illitic and montmorillonitic fields, with secondary occurrences within the montmorillonitic field and isolated cases within the illitic field (Figure 14a). This distribution is consistent with plastic behavior influenced by a smectitic component, while still allowing for a persistent illitic background at the formation scale.

At the same time, Skempton’s (1953) activity chart [8], based on the relationship between plasticity index and clay content (<2 µm), yields a mean of AI ≈ 1.09 for Moigrad (Figure 14b), which supports a predominantly illitic mineralogical background with secondary contributions from montmorillonite and kaolinite. Overall, the indirect evidence points to a mixed clay matrix in which smectitic influence is present but not dominant.

For the Moigrad Formation, semi-quantitative XRD analyses (Moigrad subset n = 11; Section 4.2) provide an independent mineralogical benchmark at formation scale (Table 6). The bulk mineralogy is predominantly siliciclastic, with quartz as a major constituent (mean ≈ 33.8 wt%, up to 66.1 wt%), while the total clay-mineral pool remains substantial (mean ≈ 38.76 wt%). Calcite is generally low in the analyzed subset (typically near detection, with one minor-carbonate specimen), indicating that carbonate cementation is not a pervasive control in the XRD subset, even though carbonate-related fabric effects may still be relevant across the broader geotechnical dataset.

Within the clay-mineral pool, the physico–geotechnical signal (mean Skempton activity AI ≈ 1.09 [8]; Figure 14b) is consistent with an overall illitic background (with chlorite/kaolinite-group contributions) coupled with subordinate but persistent expandable components. In bulk terms, expandable phases occur at low absolute proportions—e.g., montmorillonite ranges from 0.92 to 3.11 wt%, while mixed-layer expandable components range from 1.56 to 6.50 wt%—yet bulk wt% alone understates their relevance because XRD is performed on whole-rock powders. When the expandable fraction is normalized to the total clay-mineral pool (i.e., expandable/clay minerals), the proportion becomes markedly more informative and variable (up to ≈0.37 in the analyzed subset). This reconciliation is critical for Moigrad: a few wt% of expandable clays in bulk composition can correspond to a sizeable share within the clay-mineral pool, disproportionately increasing surface area, bound water, diffuse double-layer effects, and therefore PI/AI, even when the formation retains an illitic “background” signature in mean activity.

This mixed mineralogical framework helps explain why Moigrad is dominated by moderate-to-high FS classes at formation scale while still exhibiting large internal scatter. Up to FS < 140%, the near-constant mean AI (≈1) combined with the sharp clay-content increase beyond FS ≈ 100% supports the interpretation that swelling intensity is primarily governed by the volumetric effectiveness and continuity of the fine matrix (i.e., clay volume/connectivity), rather than by a systematic formation-scale increase in intrinsic clay reactivity. By contrast, the emergence of very active behavior (FS ≥ 140%) is consistent with the joint requirement of (i) high clay content (a volumetrically continuous fine matrix) and (ii) locally more reactive clay fractions/fabrics, reflected by the high-AI tail (Figure 14b), which can be mineralogically enabled by higher relative shares of expandable components within the clay-mineral pool. As for carbonate effects, the low calcite levels in the XRD subset are compatible with the non-linear CaCO₃ behavior observed across compositional classes in the geotechnical dataset: carbonate influence is best treated as a fabric- and occurrence-dependent microstructural modifier rather than a uniform formation-scale mineralogical control.

6. Conclusions

This study integrates direct free swell measurements (FS; STAS 1913/12-88 [12]) with index-property–based swelling classifications (Seed et al. [9], Van der Merwe [10], Dakshanamurthy and Raman [11]) and semi-quantitative XRD (RIR) constraints to assess the swelling potential of the Brebi, Mera and Moigrad formations. The main conclusions are:

• For the Brebi Formation, FS indicates low-to-active behavior (50–135%), with most specimens in the medium (70–100%) and active (100–<140%) classes and no very active values (≥140%). Clay-fraction–sensitive charts (Seed [9]; Van der Merwe [10]) systematically assign higher susceptibility than the LL–PI method (Dakshanamurthy and Raman [11]), consistent with their sensitivity to the volumetric role of fines. However, the measured FS distribution suggests that this higher “susceptibility” is not fully expressed as free swell, which is interpreted as a microstructural constraint effect in carbonate-influenced/cemented materials. XRD supports a carbonate-dominated bulk composition (calcite-rich) with an illite/muscovite-dominated clay-mineral pool and only minor expandable phases in bulk terms; therefore, variability in FS is more consistently explained by the proportion/continuity of the fine matrix and carbonate-related fabric stiffening than by large formation-scale shifts in smectite abundance.
• For the Mera Formation, FS spans the full NP 126:2010 [5] activity range (30–170%), with a predominance of active soils and the occurrence of very active behavior. Empirical classifications converge towards high to very high susceptibility across methods, and the systematic increase in activity index with higher FS supports a progressively more reactive fine fraction in the higher-swelling domain. XRD indicates a framework-silicate–dominated bulk mineralogy (orthoclase-rich) and an illite-dominated clay-mineral pool with the largest dispersion of expandable components among the three formations, consistent with the wider FS range. Carbonate effects are best treated as a secondary/modulating control in this unit, with swelling expression primarily reflecting the combined action of fine-fraction reactivity and the volumetric effectiveness of the clay matrix.
• For the Moigrad Formation, FS also spans the full activity spectrum (20–190%) but is dominated by medium-to-active classes, with very active values occurring less frequently. A practical transition is observed around FS ≈ 100%, above which the marked increase in clay fraction indicates a shift to a regime where the continuity/volumetric dominance of the fine matrix becomes a major control on swelling intensity. Differences within the active–very active domain are further modulated by fine-fraction characteristics and microstructural constraints. Carbonate influence is non-monotonic across CaCO₃ classes, supporting a fabric/occurrence-controlled effect rather than a single formation-wide trend. XRD indicates a substantial clay-mineral pool with an illitic background and persistent, though bulk-minor, expandable components; when considered relative to the clay-mineral pool, these expandables can be locally meaningful and help explain the upper end of high swelling values.

Across all formations, the empirical charts diverge in a systematic way: clay-fraction–sensitive schemes (Seed [9]; Van der Merwe [10]) tend to assign higher swelling susceptibility than the LL–PI approach (Dakshanamurthy and Raman [11]), especially in carbonate-influenced or cemented materials. In this context, Seed [9] and Van der Merwe [10] are best interpreted as conservative susceptibility screens that highlight horizons where the volumetric role of fines is decisive, whereas Dakshanamurthy and Raman [11] can provide a useful plasticity-based reference that may track the expressed FS more closely where microstructural stiffening attenuates swelling. Consequently, empirical classifications should be applied jointly and calibrated against FS, rather than treated as interchangeable predictors.

XRD confirms the presence of expandable components (including montmorillonite occurring as a separate phase and mixed-layer montmorillonite–chlorite) in all formations, but the dataset is semi-quantitative (RIR, no internal standard) and based on a reduced subset that does not fully overlap with geotechnical specimens. Mixed-layer phases, in particular, carry an intrinsic quantification uncertainty (ordering/overlap/orientation effects). Therefore, XRD results are used as formation-scale mineralogical constraints that support the mechanistic interpretation of FS and index-based classifications, rather than as specimen-by-specimen predictors of swelling.