6.1 Processes governing the chemical and isotopic composition of waters
Water-rock interaction at relatively high temperature may be considered as the main process influencing the chemical composition of the collected waters. The Na-Cl(SO4) composition of thermal springs at Nhawóndòc, Tenta and Niaondive suggests an interaction with low solubility rocks, i.e. gneisses and granitoid rocks of the Crystalline Basement and mafic rocks of the Tete Suite. In fact, the relatively low salinity of waters seems to exclude the interaction with soluble evaporitic rocks and/or mixing with highly saline (connate) waters, which are not reported to be present in the study area. So, the relatively high Cl- concentration may result from the dissolution of minerals in which the Cl- ion occurs as a vicariant of OH-, such as micas (mainly biotite, as accessory mineral in the gneisses of the study area), amphiboles, apatite, etc. Consistently, the relatively high SO42- concentrations may be produced through pyrite dissolution under O2-free conditions. Owing to its deleterious environmental impact, the reaction mechanisms and kinetics of O2-driven pyrite dissolution has received considerable attention (e.g., Moses et al. 1987; Moses and Herman 1991; Williamson and Rimstidt, 1994; Gleisner et al. 2006), whereas pyrite dissolution in O2-free systems has never been investigated to the best of our knowledge. Since the formal oxidation state of S in pyrite is -1, it is reasonable to hypothesize, from a purely theoretical point-of-view, the occurrence of the following disproportionation reaction in the absence of O2:
4FeS2 + 4 H2O ® H2S(aq) + 6 HS- + SO42- + 4 Fe2+. (3)
According to reaction (3), seven of the eight S atoms initially hosted in pyrite are reduced to aqueous sulphide species and only one is oxidized to sulphate, whereas Fe is expected to maintain the formal oxidation state of +2, under strongly reducing conditions. Sulphide and sulphate are considered in reaction (3) because they are the thermodynamically most stable sulfur species (Langmuir, 1997), although several aqueous S species with formal oxidation state intermediate between those of sulphide and sulphate are known (Williamson and Rimstidt, 1992) and could be involved in pyrite disproportionation in the absence of O2.
It must be underscored that reaction (3) produces aqueous H2S, which is an acid somewhat weaker than aqueous CO2 (or carbonic acid) below 115°C. However, H2S becomes an acid stronger than aqueous CO2 at higher temperatures. Therefore, pyrite dissolution in O2-free systems might actually generate the acidity needed for the occurrence of silicate dissolution, thus playing a fundamental role in the overall water-rock interaction process. The good correlation between SO4 and Na for the three Na-Cl(SO4) samples could be explained by coupled pyrite-albite dissolution, as schematically indicated by the following reaction:
NaAlSi3O8(s) + 4FeS2(s) + 4.5H2O® Na+ + SO42- + 7HS- + 4Fe2+ + 2SiO2(s) + 0.5Al2Si2O5(OH)4(s) (4)
Reactions (3) and (4) must be considered working hypotheses at this stage. To confirm their occurrence, we suggest to investigate the distribution of pyrite in the rocks of the area of interest and to carry out sulphur isotopic analyses, as done for several thermal sites of Malawi (unpublished reports by LM). The good correlation between HCO3- and Ca2+ for the three Na-Cl(SO4) samples suggests the occurrence of calcite dissolution/precipitation which, in turn, is probably controlled by acquisition/loss of CO2. The good correlation between K+ and Mg2+ is probably explained by acquisition of both constituents upon cooling or occurrence of exchange reactions also involving Na+ ion. The possible occurrence of these processes casts some uncertainties on theoretical geothermometric evaluations.
The Maiura Na-Cl(HCO3) spring (#3) is located in an area marked by the presence of rocks belonging to the Karoo and Post-Karoo Formations, largely made of sedimentary rocks with coal-bearing horizons. In this framework, the availability of CO2 is expected to be higher than in the mafic rocks of the Tete Suite and the underlying crystalline rocks of the Precambrian basement. Therefore, CO2-driven plagioclase dissolution, at saturation with calcite is expected to be the main reaction leading to the production of Na-Cl(HCO3) waters (Eq. 5):
CaxNa1-xAl1+ xSi3-xO8(s) + CO2(aq) + (3+x)/2 H2O ®
(1-x) Na+ + (1-x) HCO3- + x CaCO3(s) + (2-2x) SiO2(s) + (1+x)/2 Al2Si2O5(OH)4(s) (5)
These waters are typically characterized by low Ca2+ concentrations, owing to saturation with calcite and high HCO3- concentrations. Consequently, they may acquire high F- concentrations, up to the maximum value constrained by fluorite saturation, through dissolution of minerals in which the F- ion occurs as vicariant of OH- ion, such as apatite, micas, amphiboles, etc. Highly saline waters (over 8000 mg/L) were found in the Karoo and Post-Karoo Formations (Groundwater Consultants Bee Pee (Pty) Ltd and SRK Consulting (Pty) Ltd, 2002) but the low salinity of the Maiura spring seems to exclude the hypothesis of mixing with similar brines.
The d18O and dD values of the collected waters, indicating a meteoric origin, are consistent with the lack of saline (connate) waters from the deeper circuits.
As far as the dissolved gases are concerned, the classical triangular diagram of N2-Ar-He (Fig. 6; from Giggenbach and Goguel 1989) shows that N2 and Ar are mostly of atmospheric origin and are driven to depth by infiltrating meteoric waters. The relatively high He contents of samples 1 and 3 point to a prolonged residence time in crustal rocks, leading to accumulation of radiogenic 4He produced from a-decay of U- and Th-bearing minerals. The crustal origin of He is confirmed by the low R/Ra ratios (0.13-0.21 R/Ra), based on which it is reasonable to conclude that there is no evidence of primary mantle 3He degassing (Marty and Jambon, 1987). Carbon dioxide shows very low concentrations in the collected waters. The isotopic composition suggests a shallow origin from plant-root respiration and aerobic decay of organic matter contained in soils and/or a deeper one through anaerobic decay of organic matter contained in the sedimentary rocks of the Karoo Supergroup (Cerling et al., 1991), whereas a deep (inorganic) magmatic-metamorphic provenance seems to be ruled out. Shallow alteration of organic matter and/or interaction with coal deposits of the Lower Karoo Formation may be responsible for the relatively high CH4 content of the Maiura spring relative to the other samples. This hypothesis is consistent with the carbon and hydrogen isotopic composition of CH4, which suggests a probable derivation of methane from biogenic sources.
6.2 Geothermometric estimations
Silica geothermometers are potentially affected by dilution, but the effects of this process are expected to be negligible for the Na-Cl(SO4) thermal springs at Nhawóndòc, Tenta and Niaondive (samples 1, 2 and 4, respectively) because they have very similar chloride concentrations, suggesting that these springs discharge the same pure thermal endmember. This assumption is fortified by the low concentrations of Mg recognized in the samples and suggesting the negligible presence of shallow fluids. The limited differences in SiO2 concentrations are probably due to variable re-equilibration upon cooling during their upflow. Chalcedony solubility (Fournier, 1977) indicates aquifer temperatures of 110, 118, and 108°C for samples 1, 2 and 4, respectively, and a somewhat lower aquifer temperature, 93°C, for the Maiura spring, (sample 3) (Table 4). Corresponding quartz temperatures (Fournier, 1977) are 26-28°C higher than chalcedony temperatures and, therefore, are probably less likely than latter ones (see section 4.2). The K-Mg geothermometer (Giggenbach, 1988) provides aquifer temperatures of 119, 112, and 113°C for samples 1, 2 and 4, respectively, in good agreement with chalcedony temperatures, whereas the K-Mg temperature of sample 3, 153°C, is significantly higher than the chalcedony temperature and is probably less reliable for the reason given below.
As shown by Cioni and Marini (2020), all the different Na-K geothermometric functions proposed by different authors appear to be plausible and, therefore, there is an infinite number of Na-K geothermometers which are controlled by the exchange reaction between low-albite and variably ordered adularia, from fully ordered maximum-microcline (with order-disorder degree, ZAdl = 1) to completely disordered high-sanidine (with ZAdl = 0). Equilibrium coexistence of low-albite and fully ordered maximum-microcline indicates aquifer temperatures of 210, 132, and 167°C for samples 1, 2 and 4, respectively, whereas equilibrium coexistence of low-albite and completely disordered high-sanidine indicates aquifer temperatures of 92, 27, and 55°C for samples 1, 2 and 4, respectively (Table 4). The limiting Na-K aquifer temperatures of sample 3 compare with those of sample 1, that is, 210 and 93°C. Since there is an infinite number of Na-K geothermometers, there is also an infinite number of other cation geothermometers which are controlled by exchange reactions involving adularia. For this reason, cation geothermometers (e.g., K-Mg, Na-K-Ca, etc.) appear to be less reliable than silica geothermometers.
Following Cioni and Marini (2020) and using the Na-K ratio as indicator of the degree of order-disorder of hydrothermal adularia, in hypothetical equilibrium with the thermal waters of interest at the chalcedony temperature, it turns out that ZAdl is 0.132, 0.848, and 0.422, for samples 1, 2 and 4, respectively, and 0.422 for sample 3. These ZAdl values can then be used to compute K-Ca and Na-Ca temperatures, controlled by clinozoisite (Czo), prehnite (Prh), laumontite (Lmt), and wairakite (Wrk). For the three Na-Cl(SO4) thermal springs at Nhawóndòc, Tenta and Niaondive, the most plausible K-Ca and Na-Ca temperatures (i.e., closest to the chalcedony temperature) are those controlled by equilibrium with clinozoisite, namely 115, 127, and 124°C and 116, 134, and 137°C, for samples 1, 2 and 4, respectively. For the Na-Cl(HCO3) Maiura spring (sample 3), the most plausible K-Ca and Na-Ca temperatures are those controlled by equilibrium with laumontite, namely 99 and 105°C, respectively. All in all, these K-Ca and Na-Ca temperatures are somewhat higher than chalcedony temperatures. Among the three Na-Cl(SO4) thermal springs, sample 1 from Nhawóndòc is the one with the K-Ca and Na-Ca temperatures closest to the chalcedony temperature, with differences of 5 and 7°C only. Since this is the sample with the highest Ca concentration, 98.1 mg/kg (vs. 77.1 and 73.7 mg/kg of samples 2 and 4, respectively), the deviations of the K-Ca and Na-Ca temperatures from the chalcedony temperature might be due to variable loss of Ca, probably through precipitation of calcite, as suggested by Ca-HCO3 relationships (see above).
Regarding multicomponent geothermometry, the obtained SI-T curves (Figures 7-9) are very different from each other, for clinozoisite, prehnite, muscovite, and clinochlore-7A, whereas the SI-T curves of quartz, chalcedony, the two adularias, low-albite, laumontite, wairakite, heulandite, and anhydrite exhibit smaller shifts.
In all three samples, clinozoisite, prehnite, wairakite, and anhydrite do not attain the equilibrium condition within the saturation range of SiO2 minerals. The deviation from this condition for the Ca-Al silicates could be due to variable loss of Ca and/or the fact that the SI of Ca-Al silicates is strongly influenced by the pH and partial pressure of CO2, which are poorly defined parameters, at high temperatures, for the thermal springs. The occurrence of hydrothermal anhydrite requires SO4 concentrations higher than those of the thermal waters of interest. These minerals are therefore poor geothermometric indicators, at least in the considered waters.
Clinochlore-7A and muscovite achieve saturation at different temperatures, either close to the quartz equilibrium temperature in some cases or at even higher temperatures in other cases. The geothermometric indications provided by these two minerals have to be considered with caution due to the variable chemistry of hydrothermal chlorites and illites.
Laumontite attains saturation close to the quartz equilibrium temperature in samples 1 and 2 and near the chalcedony saturation temperature in sample 3. Heulandite achieves saturation near the chalcedony equilibrium temperature in samples 1 and 2 and at somewhat lower temperatures in sample 3.
Low-albite, the two adularias, and heulandite (as well as, obviously, calcite in the second and third models) attain the equilibrium condition within the saturation temperatures of quartz and chalcedony. Since equilibrium with low-albite and calcite (in the second and third models) was forced and the order parameter of the two adularias was suitably chosen, the only independent geothermometric results are given by chalcedony, quartz, and heulandite. Since quartz probably overestimate the aquifer temperature below 180°C (Arnórsson et al., 1983), the results given by multicomponent geothermometry are essentially based on chalcedony and heulandite saturation, with mean values of 109.5 ± 2.0 (1s) °C for sample 1, 117.1 ± 1.5 °C for sample 2, and 87.2 ± 2.2 °C for sample 3. These temperatures are slightly lower than those given by the chalcedony geothermometer of Fournier (1977), 110, 118, and 93°C, respectively, because pH effects (and the heulandite saturation temperature) are considered in multicomponent geothermometry whereas they are neglected by the chalcedony geothermometer.
Anyway, the temperature estimated by multicomponent geothermometry and based on chalcedony and heulandite saturation are fairly consistent with those provided by chalcedony solubility, the K-Mg geothermometer, and the K-Ca and Na-Ca theoretical geothermometers for samples 1, 2 and confirm that these two thermal waters discharge from a thermal aquifer at temperature of 110-120°C, whereas sample 3 comes from an aquifer at 90-100°C.
6.3 Conceptual circulation model
By considering the chemical and isotopic composition of the waters, the geothermometric evaluations and the available geological and hydrogeological information, a conceptual circulation model is proposed for the study area (Fig. 10). Meteoric waters infiltrating through the widespread fracture network of the SSZ circulate for more than 2 km deep in the crust, namely, into the Precambrian Basement. The crustal origin of He (as indicated by the low R/Ra ratios) and the relatively high He contents in the dissolved gas phase provide further evidence for the long circulation underground. The heavier isotopic values of the rainwater sample relative to the thermal springs suggest that the recharge area may be located at higher altitudes than the sampling quote (~100 m asl). The recharge area may correspond to the hills and mountains (up to 1000 m asl) located N and/or E of the study area (Fig. 10) towards Malawi (Fig. 2a), but this inference is poorly constrained due to the lack of an isotope-altitude relation for the study area. Meteoric waters flowing down into the crust are heated up to maximum temperatures of 110-120°C below sites 1, 2, and 4, and 90-100°C below site 3. Afterward, the heated fluids rise up towards the surface, via fractures and faults. In fact, rock permeability is mostly related to fractures and faults, which drive at the same time the meteoric water downwards and deep waters upwards, according to the position of the fluids in the convective cells.
Water-rock interaction processes at high temperature between the fluids and the metamorphic and mafic rocks control the chemical composition of the deep waters that are finally discharged from the considered thermal springs. The described hydrothermal system does not show connections with any active magmatic system but it seems to be related to a paleo-suture tectonic structure, as observed in intra-cratonic systems (Minissale et al., 2000). This hypothesis agrees with: 1) the low CO2 concentrations and the negative d13C value of CO2, 2) the isotopic value of CH4, 3) the relatively high He contents and the low R/Ra ratios, indicating low 3He concentrations, and 4) the N2/Ar ratio close to the meteoric value.