Influence of Clay Minerals Nature on the Hydromechanical and Fracture Behaviour of Stones

The first clay-bearing stone under study is the Villarlod molasse, from the Burdigalien formation, extracted from the Fribourg region (Switzerland), where it is used as a traditional building material (Felix 1983). Typical decay of this stone is related to small scales around 1 cm thick (Fig. 1). Scaling effect of this specific stone has been under study for several years (Felix 1977; Felix et al. 2000).

The second sedimentary clay-bearing stone considered for this research is the Thüringer sandstone, known as Schilfsandstein from the Superior Trias. This stone has also been used as monumental material, for instance, at the Drei Castle (Stück 2013).

The supposed anisotropy due to the natural bedding plane of these to sedimentary stones is identified by measuring the elastic wave propagation velocities in the three directions of blocks (35 × 35 × 35 cm²) from quarry. Ten measures are carried out on each direction. The lowest value indicates the direction perpendicular to this plane, along the \(\mathbf {e_n}\) axis. The highest velocity helps to identify the preferred orientation of minerals and inherent microcracks along \(\mathbf {e_t}\). As \(V_t\) and \(V_l\) are close, the two rocks are considered isotropic transverse. Plugs of 40 mm in diameter and 80 mm in length are cored from blocks in the two principal directions of the stone (Fig. 2). They are designed as perpendicular and parallel samples.

Samples of sedimentary rock may present heterogeneities due to the fabrics and mineralogical composition. They may influence the mechanical properties (Baecher and Einstein 1961; Guy et al. 2018). Wave propagation measurements on the set of samples confirm that cores and specimens used for experimental testings are homogeneous. No macroscopic scattering is identified (Table 1).

Mineralogical composition and microstructural properties are studied. Particular attention is paid to clay minerals identification by X-ray diffraction (XRD) using a Brücker D8 \(\hbox {Advance}^\circledR\) Bragg–Brentano diffractometer. Extraction of clay fraction is performed for the two sedimentary stones, followed by two treatments on natural clay fraction: a 490 °C oven heating during 4 h and a conditioning in ethylene glycol saturated atmosphere. Identification of mineralogical phases is confirmed by optical microscopy (OM) and observations of fresh fractures by FESEM observations. Relative orientation and organization of the phases are observed, and the open porosity is measured by Mercury Injection Porosimetry (MIP) (ASTM 2010). Total porosity and pore size distribution are quantified.

Elastic properties of the molasse and the sandstone are measured to study the potential modification induced by the various exposure conditions at fixed RH during humidification and drying. A specific choice of a non-destructive acoustical method results from a large number of the previous researches, indicating that dynamic methods can measure the elastic Young moduli of stones even if differences may appear with static mechanical results (Homand et al. 1993; Amadei 1996). Non-destructive dynamical testings of moduli are developed in cultural heritage field as only few samples can be cored on monuments. As relative differences are evaluated in this research—between the two stones, the two principal directions and under RH variations—acoustic method seems relevant for this analysis. A \(\hbox {GrindoSonic}^\circledR\) device is used for the testing. The measurement is based on two modes resonance frequencies on cores of 40 mm in diameter. This core geometry is appropriate for such dynamical testing (Siggins 1993). As wave propagation velocities in materials are related to the elasticity constants through the Christoffel’s equation, the Young moduli can be measured (Sarout 2006). A plane of transverse isotropy has been identified for the two sedimentary stones, so only five components of the stiffness matrix are required to characterize the material (Amadei 1996; Aristégui and Baste 1997). Thus, only two moduli are measured, perpendicular and parallel to the bedding plane, plane of transverse isotropy. The transverse samples prepared perpendicular to the natural bedding plane allow to measure \(E_n\), while axial samples parallel to the plane give \(E_t\) (Homand et al. 1993). Evolution of these two moduli with various conditions is under study. Eight samples cored on both directions are tested using this non-destructive method.

To evaluate the influence of RH variations on the mechanical and fracture behaviour, fracture toughness is also measured under various exposure conditions (presented in the section below). Samples adapted for cultural heritage studies have been proposed thanks to the previous investigations (Tiennot and Bourgès 2016; Tiennot 2017). Semi-circular Bending (SCB) samples are used for this research. They are prepared from the cores of 40 mm in diameter. Several researches used this method in geomaterial field (Lim et al. 1993; Tutluoglu and Keles 2011), where the sample is loaded under a three-point bending. The semi-circular specimen is sliced such as the thickness t is equal to the radius R, so \(t/R = 1\) (Lim et al. 1993). An initial notch of length a is prepared using a diamond saw and its final dimensions are 1 mm in width and 4 mm in length (Fig. 3). This initial notch length is fixed, so that \(a/R = 0.2.\) The distance between the two rollers is fixed as \(S/R=0.6\)

The notch is not sharpened to study the minerals in which the initiation occurs, because it seems more representative of the natural conditions (Tiennot 2017).

As the only mode I is considered, the initial notch is straight aligned with the loading direction (Fig. 3). For these SCB samples, the stress intensity factor is calculated using the expression (1) depending on the applied stress \(\sigma\), the initial notch length a, and a geometrical factor Y, depending of the sample geometry: withwhere F is the applied force, R is the radius, and t is the thickness of the sample.

For the used parameter \(S/R = 0.6\), \(t/R = 1\), and \(a/R = 0.2\), Lim et al. (1993) calculated the value \(Y = 3.260\). Knowing the maximum loading of the specimen allows calculating the critical stress intensity factor in mode I, the toughness \(K_{\text {IC}}\).

The aim of this research is to better understand the fracture behaviour of two sedimentary stones. Thus, toughness has to be measured with respect to the natural minerals orientation. The identified plane of transverse isotropy is related to the natural bedding plane of the Villarlod molasse and the Thüringen sandstone. Three configurations of toughness measurements with respect to this plane are described in Fig. 4. Perpendicular SCBperp configuration is prepared using plugs cored along \(\mathbf {e_n}\). SCBt and SCBn samples are sliced on cores along the bedding plane. For SCBn and SCBt configuration, the initial notch required is, respectively, orthogonal and along the preferred orientation of minerals and microcracks along \(\mathbf {e_t}\) (Fig. 4).

Hygric swelling ability of the two stones is measured using the dilatometric system adapted at the laboratory (Mertz et al. 2012) on the two directions perpendicular and parallel to the bedding plane. Dimensional behaviour is measured by LVDT sensors TWK (accuracy 0.3 μm). Free swelling strains are measured under isothermal condition (T = 20 °C) from the initial dry state to the choosen RH level 33–65–97%, during humidification and drying. Each step within the dilatometric device lasts 96 h (Fig. 8). These free swelling strains are related to the natural orientation of the inherent microcracks and of the mineralogical phases, especially the clay minerals, sensitive to moisture.

For each testing, the initial state items designed samples oven-dried until constant mass. For elastic moduli and toughness measurements, the humidification phase designed samples first oven-dried and then placed on enclosures saturated at 33%, 65%, or 97% of RH until constant mass. Samples used to characterize the drying phase are first oven-dried, then conditioned at 97% of RH and finally placed on the defined enclosure at 65% or 33% of RH until constant mass. These conditions are relevant with environmental conditions applied on monumental stone.

For each condition, initial dry state and RH levels, four cores parallel and four cores perpendicular are used to measure Young moduli in the two directions. Eight samples SCBt, SCBn, and SCBperp are used to characterize the two stones at the initial state. Four samples in each configuration are tested for the five RH levels previously defined for the humidification and the drying phase.

Young moduli are measured along the two principal directions of the sedimentary stones, \(\mathbf {e_n}\) and \(\mathbf {e_t}\). At the initial dry state, the moduli of Villarlod molasse are \(E_n = 8.43 (0.39) \;\hbox {GPa}\) and \(E_t = 10.7 (0.17) \;\hbox {GPa}\) (Table 2). For the sandstone, \(E_n = 13.2 (0.70) \;\hbox {GPa}\) and \(E_t = 14.4 (0.94) \;\hbox {GPa}\) (Table 3). The modulus \(E_n\) is lower than \(E_t\), which is relevant with the orientation of microcracks and minerals (Homand et al. 1993).

Under moisture, a significant loss of moduli in the two principal directions is identified and results indicate a decrease up to 69% at RH = 97% for the Villarlod molasse. However, the evolution is quite different for the two stones. Indeed, both elastic moduli of Villarlod molasse are affected as soon as the RH increases (Table 2). At RH = 33%, \(E_n\) modulus decreases up to 47%. Few modifications are then measured between 33 and 65% of RH, and a new significant decrease is measured between 65 and 97% of RH.

On the contrary, Thüringer sandstone is not affected in the same manner (Table 3). The moduli are barely affected under 65% of RH, and the significant decrease is measured between 65 and 97% of RH.

These results quantify the known evolution of Young moduli with the increase of moisture content. Moreover, they illustrate the differences in the elastic behaviour of the two stones and the fact that Villarlod molasse is affected even for low moisture content. This is related to the nature of the mineralogical phases. Indeed, molasse main clay phases, smectite and glauconite, are both very sensitive to moisture. It appears with these results that they are involved in the loss of elastic properties at very low moisture content. On the contrary, the clay phases in the sandstone are not affected at the same humidity state and samples have to be conditioned at RH higher than 65% to suffer a significant moduli loss.

To characterize the elastic properties evolution over a complete cycle of moisture content variations, the influence of RH decrease on samples first conditioned at 97% of RH is studied.

For the Villarlod molasse, the two moduli increase as a result of RH decrease from 97% to the dry state. It should be noted that the value of both moduli \(E_t\) and \(E_n\) is more important for a given RH state during the drying phase than the humidification. For instance, \(E_{65\%{\text {drying}}}\) is higher than \(E_{65\%{\text {humid}}}\). This can be related to the free swelling strains (Fig. 8). As a matter of fact, it is underlined that \(\varepsilon _{s-{\text {drying}}}\) reaches lower values during the shrinkage. The recovering of this mechanical property is in close correlation with the elimination of water molecules from the network and the minerals sensitive to moisture. As the swelling strains in the two directions indicate the important gradient induced during drying, the slightly improved values of both moduli during drying seem relevant.

On the other hand, the Thüringer sandstone behaviour is more complicated. Indeed, the first result shows the inversion of its anisotropy. At 65% of RH of the drying phase, \(E_{n}\) becomes higher than \(E_{t}\). Moreover, at 33% of RH, \(E_{n}\) remains lower despite the decrease of moisture content. This indicates that deformation of this sandstone phases are highly different—for instance, between quartz and clay phases—and requires a microcracking evolution. Initiation of a drying-induced microcracking is highlighted. As \(E_{n}\) is more affected, these induced microcracks are generated along \(\mathbf {e_t}\), the inherent microcracks plane (Bargellini et al. 2006). This is relevant with the result as obtained after several cycles of RH variations (Tiennot et al. 2017). Such microcracking can appear to relax the stresses induced during the drying shrinkage (Bazǎnt and Raftshol 1982). Thus, clay phases nature and distribution influence the swelling and consequently the elastic behaviour of stones.

Fracture behaviour and toughness are studied with respect to the natural anisotropy of the stones. Both stones show a quasi-brittle and brittle behaviour (Figs. 9, 10) with a slow crack propagation for the Villarlod molasse and a really fast for the Thüringer sandstone.

From the initial dry state, samples present different fracture behaviours. Toughness values reached in each configuration are lowest for the molasse than for the sandstone (Tables 4, 5). First, Villarlod molasse shows an anisotropic behaviour at the initial state, as SCBperp, SCBn, and SCBt values differ from each other (Fig. 9 and Table 4). SCBt is lower, indicating that initiation occurs more easily along the isotropy transverse plane. On the contrary, the Thüringer sandstone toughnesses SCBn and SCBt are identical at the initial state, so that no preferred orientation is highlighted for crack initiation (Fig. 10 and Table 5). Evolution of the stones toughness under various RH conditions is then studied.

Several papers presented the influence of water vapour pressure on toughness of stone (Utagawa et al. 1999; Nara et al. 2011, 2012). Some of them proposed to follow the dependency of \(K_{IC}\) with the water vapour pressure p using the equation:where \(\beta\) is a constant for \(p=1\) Pa and m the slope of the line on the logarithmic graph (Kuruppu et al. 2010; Kataoka et al. 2014). The resistance to the initiation of crack is more affected by water vapour pressure when m reaches high values.

In this research, when RH increases, the three toughnesses SCBperp, SCBn, and SCBt of the molasse and the sandstone decrease. This phenomenon has already been observed, and is due to the water condensation effect at the crack tip (Atkinson 1982).

However, the influence of moisture content and of the induced swelling on the mechanical behaviour of the two stones is different. The evolution of the sandstone toughness has already been presented in Tiennot et al. (2017). SCBn and SCBt values measured during humidification indicate that an anisotropy of fracture toughness appears when the moisture content increases (Fig. 11). Indeed, values of SCBn and SCBt remain equal at RH = 33%, and when 65% are reached, anisotropy appears and remains at 97% of RH. SCBt samples are more affected by the RH increase than SCBn, as the m parameter is higher. The opening of microcracks, as previously described, along the \(\mathbf {e_t}\) axis is relevant with the identified anisotropy of toughness. Thus, oriented swelling influences the toughness with respect to the orientation of the phases.

Villarlod molasse toughness evolution is described in Table 4 and Fig. 12. At the initial dry state, the values of \(K_{\text {IC}}\) in the three configurations differ from each other. When water vapour pressure increases SCBn and SCBt both decrease in the same manner until 65% of RH. However, when the samples are conditioned at RH = 97%, the three fracture toughnesses decrease and reach the same value of 0.05 \({\text {MPa}}.\sqrt{m}\) (Table 4). Under this high water vapour pressure and thus this high free swelling of the material, the ability of the stone to withstand crack propagation is the same perpendicular and parallel to the natural orientation of minerals and microcracks. The swelling of the sensitive minerals, especially smectite and glauconite, annihilates the initial anisotropy of the fracture behaviour of molasse (Fig. 12).

It should also be noted that molasse m parameters reach higher values than the sandstone. This indicates that the increase of moisture content affects more significantly the first stone. This may also be explained by the mineralogical differences and by the swelling ability of each phase. Thus, the two stones show a different behaviour in close correlation with the nature and accessibility of the mineralogical phases and their swelling ability has a direct impact on the macroscopic mechanical behaviour.

Evolution of toughness during drying is then verified. The reversibility of the behaviour is observed for the Villarlod molasse. Toughness values are recovered, while RH decreases (Table 4).The temporary isotropy at RH = 97% of the molasse, and moreover, the reversibility of this behaviour is related to the highly sensitive to moisture phases smectite and glauconite.

On the contrary, the sandstone anisotropy measured for RH values higher than 65% remains during the drying. A preferred initiation orientation appears within the sandstone after a complete RH cycle. This indicates that the phases affected and involved in the swelling are not able to recover their initial conditions. Indeed, as previously discussed in Tiennot et al. (2017), the evolution of SCBt and SCBn illustrates the creation of drying shrinkage microcracks along the \(\mathbf {e_t}\) axis. Anisotropy, in close correlation with the mineralogical phase distribution and nature, evolves and influences the fracture behaviour of the two stones.

While the clay matrix of molasse is able to sustain the free hygric swelling, the sandstone matrix needs to relax the stresses caused by drying, inducing a modification of the microcracks network. Thus, mineralogical phases have direct consequences on the fracture behaviour of the stone during the drying.

The SCB samples tested to measure the toughness of the stones are observed after crack propagation. Clay phases are assumed to be the weakness phases of the stones. This is confirmed by the crack paths followed within the stone.

Optical microscope and FESEM images analysis indicates that intergranular paths are followed within the two sedimentary stones (Figs. 13, 14). Crack gets around and along the most rigid grain boundaries as quartz and feldspars of both stones, and propagates within the smectite phases of molasse (Figs. 6, 13) and the calcitic and clay matrix of the sandstone (Fig. 14). In the flexible smectitic matrix of molasse, grain boundaries marks can even be observed after the crack propagation (Fig. 15). Intergranular decohesion of glauconite spots when the principal crack propagates along them also indicates the weakness of these clay minerals (Fig. 16).

These results and observations illustrate that failure occurs in the plane, where clay phases are located and that fracture patterns follow these minerals orientation. Clay phases are the preferred planes of propagation.