Determining the formation temperature of minerals using fluid inclusions is a crucial step in understanding rock-forming scenarios. Unfortunately, fluid inclusions in minerals formed at low temperature, such as gypsum, are commonly in a metastable monophase liquid state. To overcome this problem, ultra-short laser pulses can be used to induce vapor bubble nucleation, thus creating a stable two-phase fluid inclusion appropriate for subsequent measurements of the liquid-vapor homogenization temperature, Th. In this study we evaluate the applicability of Th data to accurately determine gypsum formation temperatures. We used fluid inclusions in synthetic gypsum crystals grown in the laboratory at different temperatures between 40 °C and 80 °C under atmospheric pressure conditions. We found an asymmetric distribution of the Th values, which are systematically lower than the actual crystal growth temperatures, Tg; this is due to (1) the effect of surface tension on liquid-vapor homogenization, and (2) plastic deformation of the inclusion walls due to internal tensile stress occurring in the metastable state of the inclusions. Based on this understanding, we have determined growth temperatures of natural giant gypsum crystals from Naica (Mexico), yielding 47 ± 1.5 °C for crystals grown in the Cave of Swords (120 m below surface) and 54.5 ± 2 °C for giant crystals grown in the Cave of Crystals (290 m below surface). These results support the earlier hypothesis that the population and the size of the Naica crystals were controlled by temperature. In addition, this experimental method opens a door to determining the growth temperature of minerals forming in low-temperature environments.


A common method to obtain information on mineral growth conditions relies on the analysis of the mineral-forming solutions that were trapped during their growth (Roedder, 1984; Bodnar, 2003a). During the formation of a crystal, several processes may provoke the formation of voids, which are filled by the fluid from which the crystal is growing, the so-called primary inclusions. Provided that these inclusions have not been subjected to postentrapment modifications of their composition and/or volume (Bodnar, 2003b), they can be used to assess the pressure-temperature (P-T) conditions of crystal growth. This is done by determining the density of the trapped solution, based on measurements of the liquid-vapor homogenization temperature, Th, and locating the corresponding fluid isochore using an appropriate equation of state. This technique has been extremely valuable, particularly in the case of hydrothermal mineral deposits (Bodnar et al., 1985; Wilkinson, 2001) and diagenetic environments (Goldstein and Reynolds, 1994); i.e., when the formation temperature of the crystals was high enough to provoke the segregation of the vapor bubble during its cooling history to surface temperature. However, for most near-surface (i.e., low temperature) crystal formation, most inclusions remain monophase in the metastable liquid state even after cooling in the laboratory (Roedder, 1967, 1971; Zheng et al., 1991). In all these cases, when cooling, either in nature or in the laboratory, fails to induce the formation of a gas bubble, the fluid inclusion technique cannot provide precise information on mineral growth temperature (Goldstein, 2001). The inability to determine the mineral formation temperature may provide obstacles and limit the applicability of using fluid inclusions to infer past climates.

One mineral that could yield important and decisive thermal information for understanding the history of seas and atmosphere (Orti, 2010), as well as the history of water on Mars in future explorations (McCollom and Hynek, 2005; Tosca et al., 2008), is gypsum (CaSO4·2H2O). Unfortunately, most gypsum crystals (Sabouraud-Rosset, 1976; García-Guinea et al., 2002; García-Ruiz et al., 2007) contain typically monophase fluid inclusions or, in the best of the cases, a small percentage of two-phases inclusions that make the microthermometric measurements unreliable. This situation could be changed because femtosecond laser pulses are able to induce vapor bubble nucleation in metastable monophase inclusions, thus creating a stable two-phase state for subsequent Th measurements (Krüger et al., 2007). Nevertheless, the technique needs to be carefully tested because gypsum deforms easily and presents perfect cleavage along (010), making inclusions susceptible to postentrapment volume alterations (Bodnar, 2003a). We undertook this study to evaluate the applicability of laser-supported Th measurements to decode the thermal information contained in fluid inclusions in gypsum. First we tested the method with synthetic gypsum crystals grown in the laboratory at known temperatures under ambient atmospheric pressure conditions. Based on the experimental findings, we then analyzed homogenization temperatures of gypsum crystals grown in a natural environment. For the field test, we chose the giant gypsum crystals of Naica (Chihuahua, México) (García-Ruiz et al., 2007), first because the window of temperatures at which these crystals formed is rather narrow, and second because the geological information suggests that these crystals had a rather simple cooling history.


Synthetic and Natural Gypsum Crystals

Synthetic gypsum crystals were grown in the laboratory from solutions under atmospheric pressure at different temperatures, and natural gypsum crystals were collected at different sites in the Naica mine. (For more details, see the GSA Data Repository1.)

Liquid-Vapor Homogenization Method

We used an amplified Ti:sapphire femtosecond-laser system (Coherent, Inc.) to induce the vapor bubble in metastable monophase inclusions (Krüger et al., 2007). Subsequent microthermometric measurements were performed on a Linkam THMSG 600 heating-freezing stage, calibrated by means of synthetic H2O and H2O-CO2 fluid inclusion standards. Based on cross validations, we estimated the precision of the Th measurements to be ±0.3 °C at 40 °C, and ±0.8 °C at 80 °C. Repeated nucleation and homogenization cycles on the same inclusions revealed a reproducibility of the Th measurement within 0.1 °C, confirming that the high-intensity laser pulses did not change the inclusion volume.


Figure 1 shows the results of Th measurements from monophase and two-phase inclusions in synthetic gypsum, presented as a probability distribution for each growth temperature, Tg. The diagrams illustrate that Th of initially monophase inclusions [Th(1ph)] are typically lower than Tg. Note that the small number of Th(1ph) values greater than Tg likely results from partial leakage of the inclusions and can be disregarded from further analysis. The variation of Th(1ph) values increases with increasing growth temperature of the host crystal. Within the 10%–90% interval denoted in the diagrams, the variation of Th(1ph) increases from 2.5 °C at a Tg of 40.0 °C, to 8.0 °C at a Tg of 78.9 °C. The same trend can be observed for the deviation of the median value of Th(1ph) from the actual growth temperature Tg; i.e., an increase of ΔT [TgTh(1ph)(median)] from 1.3 °C to 6.0 °C. Figure 1 also shows that homogenization temperatures of two-phase inclusions [Th(2ph)] in crystals grown at 78.9 °C are generally higher than the corresponding Th(1ph) values. The variation of Th(2ph) within the 10%–90% interval is only 2.0 °C, compared to 8.0 °C for the monophase inclusions, while ΔT [Tg − Th(median)] is 1.1 °C and 6.0 °C, respectively.

The measured homogenization temperatures of initially monophase and two-phase inclusions are systematically lower than the Tg of the gypsum crystals. The observed deviations are caused by two mechanisms: (1) the effect of surface tension on liquid-vapor homogenization (Fall et al., 2009; Marti et al., 2012), and (2) the postentrapment volume changes by plastic deformation of the inclusion walls due to high internal tension occurring in the metastable state of the inclusions.

Effect of Surface Tension on Th

In fluid inclusion research the pressure of the liquid phase is usually regarded as equal to the pressure of the coexisting vapor phase, i.e., the saturation pressure. Under this assumption (that strictly applies only to infinitely large systems), liquid-vapor homogenization proceeds by a continuous decrease of the vapor bubble volume to zero. The homogenization temperature of such a hypothetical, infinitely large inclusion is defined as the nominal homogenization temperature Th∞ (Marti et al., 2012) and would equal Tg in our experiments. In real inclusions of finite size, however, the situation is different due to the effect of surface tension. The surface tension at the liquid-gas interface of the vapor bubble results in a pressure difference, ΔP, between the liquid and the vapor phase, and forces the vapor bubble to collapse from a nonzero radius at a temperature below Th∞. Thus, Th is systematically lower than Th∞ and therefore lower than Tg. The temperature difference, ΔTh, between Th and Th∞ increases with increasing density (i.e., with decreasing Tg) and decreasing volume of the inclusions. In the case of low-temperature inclusions, and particularly for small inclusions, the effect of surface tension on Th must be taken into account for an accurate determination of mineral growth temperatures (cf. Krüger et al., 2011). The light gray bands in Figure 1 indicate the range of expected Th values considering only the effect of surface tension and inclusion volumes between 102 and 104 μm3. ΔTh ranges from −1.5 to −0.5 °C for inclusions formed at 40.0 °C, and from −0.9 to −0.3 °C for inclusions formed at 78.9 °C, assuming a 5 wt% NaCl solution (Fig. DR2 in the Data Respository).

Postentrapment Volume Changes

Upon cooling the inclusions from growth to room temperature, the fluid pressure inside the inclusions becomes negative (tensile stress), following a nearly isochoric P-T path. The inclusions thereby pass from the stable into the metastable liquid state, where the trapped water becomes stretched (cf. Roedder, 1967; Diamond, 2003). This applies not only to monophase inclusions that remain in the metastable state, but also to two-phase inclusions. Recall that spontaneous bubble nucleation in two-phase inclusions was observed ∼30–45 °C below Tg, which implies that the inclusions reached internal negative pressures of as much as −400 bar, though only for a short time. At room temperature (25 °C), the tensile stress acting on the walls of monophase inclusions reaches −120 bar in inclusions formed at 40.0 °C, and as much as −540 bar in inclusions formed at 78.9 °C (see Fig. DR3). This internal tension may cause a decrease of the initial inclusion volume by plastic and viscoelastic deformation of the surrounding gypsum host. As a result, the fluid density increases and Th decreases. For example, a volume change of −2‰ results in a decrease of Th by 7.0 °C for inclusions formed at 40.0 °C, and by 3.7 °C for inclusions formed at 78.9 °C (see Fig. DR4). Note that the reversible viscoelastic volume change of the inclusions was always found to be smaller than the total deformation and manifests by a slow increase of Th over several days and weeks after relaxation of the internal tensile stress, i.e., when the inclusions were in the stable two-phase state.

The results of Th measurements from initially monophase inclusions shown in Figure 1 indicate a clear increase of both the variation of Th(1ph) and ΔT with increasing gypsum Tg, i.e., with increasing internal tension. Burnley and Davis (2004) showed that additional factors may influence the resulting deformation, such as size and shape of the inclusions, the thickness of inclusion walls, and the orientation of inclusions within an anisotropic crystal. The interaction of all these parameters causes the large variation of the Th(1ph) values. Furthermore, the comparatively small variation of Th(2ph) values indicates that the duration of the metastable state in the inclusions has a major influence on the resulting volume change.

In summary, we have shown that the effect of surface tension on Th is comparatively small in our experiments, but increases with decreasing volume and formation temperature of the inclusions, whereas the reduction of Th due to deformation increases with increasing tensile stress. Our results indicate that Th measurements from two-phase inclusions allow for a close approximation of the gypsum formation temperature, while Th data from monophase inclusions can be affected strongly by postentrapment volume changes. However, monophase inclusions that formed below 50 °C, and thus underwent only low tensile stress, can still provide an accurate determination of Tg based on the characteristic distribution pattern of Th(1ph) data.


Gypsum also occurs at the later stages of hydrothermal systems, and in very odd cases forms colossal single and twinned crystals, up to 11 m in length, the explanation of which is a challenge (García-Ruiz et al., 2008). The most outstanding site for these gigantic crystals is the underground mine of Naica, México. According to the proposed model (García-Ruiz et al., 2007; Van Driessche et al., 2011), Naica’s gypsum crystals have grown from calcium sulfate–rich solutions that formed by dissolution of sedimentary and hydrothermal anhydrite when the system cooled below ∼58 °C, i.e., the temperature at which the solubilities of anhydrite and gypsum become equal. Two-phase fluid inclusions in a large twin crystal from the Cave of Crystals (290 m below surface) indicate that the crystal precipitated from low-salinity solutions at a temperature of ∼54 °C. Based on calculations derived from the classical nucleation theory, it was postulated (García-Ruiz et al., 2007) that gypsum crystals from the Cave of Swords (120 m below surface) have grown at lower temperatures than crystals from the deeper Cave of Crystals, but the low frequency of two-phase inclusions in these crystals did not allow confirmation of this.

After validating the laser-induced method in the lab, we measured Th values of monophase and two-phase inclusions in a gypsum crystal from the Cave of Swords (upper left photo, Fig. 2), and we complemented the fluid inclusion data for the Cave of Crystals (upper right photo, Fig. 2) by additional measurements of monophase inclusions. Figures 2A and 2C display the probability distributions of Th(1ph) and Th(2ph) values from the two caves. As in the case of synthetic samples, Th(1ph) values of initially monophase inclusions are generally lower than Th(2ph) values of two-phase inclusions. However, in comparison to synthetic inclusions formed at 50.3 °C, the variation of Th(1ph) is significantly larger. Furthermore, Figures 2B and 2D display a clear inverse correlation between Th(1ph) and the volume of the inclusions, particularly for large volumes. In synthetic gypsum samples this correlation was not observed, because of the smaller size of the inclusions (102 to 104 μm3). Our measurements confirm that the relative decrease of the inclusion volume due to deformation depends also on the inclusion volume (cf. Burnley and Davis, 2004), and we conclude that small monophase inclusions are less affected by postentrapment volume alterations than large ones. The volume and the Th∞ of the inclusions have been calculated based on the measured Th values and measurements of the vapor bubble radius at known temperatures (cf. Marti et al., 2012). The temperature difference between Th∞ and the measured Th value never exceeds 0.8 °C, and is therefore of little importance for the interpretation of the data.

We assume that gypsum crystals from the Cave of Swords formed close to the groundwater table, i.e., close to atmospheric pressure; we therefore expect that the Th∞ equals the gypsum Tg. Based on the distribution of the measured Th data, we estimate Tg to be 47 ± 1.5 °C. However, for gypsum crystals from the Cave of Crystals, we assumed crystal growth under a hydrostatic pressure of ∼16.7 bar, corresponding to a water column of 170 m. In this case, the Tg does not equal Th∞ but is ∼1.5 °C higher, due to the pressure correction. We determined a gypsum Tg of 54.5 ± 2 °C (Th∞ = 53 ± 2 °C) for the Cave of Crystals. These results demonstrate that the gypsum growth temperatures in the Cave of Swords were ∼7 °C lower than in the Cave of Crystals; this supports the idea that the larger number and smaller size of crystals in the Cave of Swords are due to an increase of supersaturation upon faster cooling of the solution (García-Ruiz et al., 2007).


The use of monophase fluid inclusions opens the possibility of determining the growth temperature of gypsum and other minerals forming at low temperature under late hydrothermal, diagenetic, or even evaporitic environments. However, differential stress (internal tension or fluid overpressure) may result from natural temperature variations during uplift or burial processes, or can be artificially induced during sampling, when collected crystals are not preserved at their actual temperature on site. Therefore, the crystals should be maintained close to their present-day ambient temperature during transport, storage, and sample preparation. We also recommend choosing fluid inclusions as small as possible in order to reduce plastic deformation. In those cases, however, when the fluid inclusions have been naturally exposed to a very complex postcrystallization thermal history with large fluctuations, and/or have been deformed by external tectonic stress, the high accuracy demonstrated here probably cannot be assured.

We are grateful to Roberto Villasuso, Roberto Carlos Reyes, and Compañía Peñoles for the facilities provided during the field studies. We are also grateful for support from the Consolider-Ingenio 2010 project “Factoría Española de Cristalización,” the Ministerio de Ciencia y Innovación (MICINN-Project CGL2010-16882), and the Swiss National Science Foundation (grant 200021–119966).

1GSA Data Repository item 2013029, materials and methods, and Figures DR1–DR4, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.