Introduction

Crystalline aquifers are composed of hard rocks through which groundwater flows and is stored in fissures formed by various processes, including weathering, tectonics, thermal stress, and unloading (e.g., Lachassagne et al. 2021; Duranel et al. 2021; Yuguchi et al. 2021; Mao et al. 2022). Groundwater that fills these fissures undergoes complex chemical interactions with the host rock, known as rock–water interaction processes. The progression of rock–water interaction begins with groundwater recharge, which involves the input of low-ionic solutes that circulate through the interconnected network of rock fissures into the aquifers, gradually altering their initial composition (e.g., Almeida et al. 2022; Akurugu et al. 2022; Zango et al. 2023).

Chemical interactions between rock and water involve various distinct processes, such as dissolution/precipitation, ion exchange, oxidation, and reduction (e.g., Adabanija et al. 2020; Kumar and Kumar 2020; Fuoco et al. 2022; Ju et al. 2023; Mizuno et al. 2023). However, the dominant mechanism is undoubtedly represented by silicate hydrolysis (e.g., Banks and Frengstad 2006; Elango and Kannan 2007; Teramoto et al. 2019; Fuoco et al. 2022; Ghalit et al. 2023).

Numerous studies have investigated the evolution of hydrochemistry in groundwater stored in fractured rock, particularly focusing on incongruent silicate dissolution (e.g., Warnner et al. 2017; Sunkari and Abu 2019; Rashid et al. 2020; Abbas et al. 2021; Kouser et al. 2022; Cho and Choo 2019; Nakayama et al. 2022). The quantitative relationship between the progress of rock–water interaction and water quality in crystalline aquifers were explored in several previous works (e.g., Sung et al. 2012; Olofinlade et al. 2018; Ackerer et al. 2021; Roy et al. 2020; Choi et al. 2020; Ojoki et al. 2021; Fuoco et al. 2022). Furthermore, rock–water interaction also leads to the precipitation of new mineral phases, primarily clay minerals such as kaolinite, chlorite, and illite, while more stable phases like K-feldspar, quartz, and zircon accumulate as residual minerals (Procházka et al. 2018).

An essential aspect of rock–water interaction is comprehending the potential release of chemical species that may pose risks to human health. A prominent example is the fluoride ion (F−), which is predominantly released into water through rock–water interaction, rendering it unsuitable for consumption. Numerous previous studies have reported the loss of potability resulting from the release of fluoride due to rock–water interaction in crystalline aquifers worldwide (e.g., Pauwels et al. 2015; Nagaraju et al. 2016; Martins et al. 2018; Sunkari and Abu 2019; Rashid et al. 2020; Cuccuru et al. 2020; Nakayama et al. 2022; Fuoco et al. 2022).

In addition to evaluating water quality, understanding rock–water interaction is also relevant for elucidating the dynamics of groundwater movement within bedrock fissures (e.g., Jaunat et al. 2012; Mao et al. 2022; Stober et al. 2022). For instance, dissolved silica derived from rock–water interaction can serve as a reliable indicator for estimating residence time in the crystalline aquifer (Benettin et al. 2015; Marçais et al. 2018; Mao et al. 2022). Furthermore, experimental data demonstrate that rock–water interaction can strongly impact the natural permeability of fissured rocks. It can increase permeability through the dissolution of certain minerals or decrease it through the precipitation of clay minerals along fractures (Sanchez-Roa et al. 2021).

One of the challenges in studying rock–water interaction is measuring the extent of this interaction, as only a few metrics have been proposed for its assessment. For example, Wanner et al. (2017) propose the use of Li as a proxy to determine the degree of interaction between water and rock, while Négrel (2006) suggests utilizing Sr, Ne, and rare-earth elements to quantify rock–water interaction. The quantification of rock–water interaction remains an ongoing issue, as different approaches and indicators provide valuable insights into this complex process. Wanner et al. (2017) propose the use of Li as a proxy, recognizing its potential in indicating the degree of interaction between water and rock. In contrast, Négrel (2006) suggests incorporating Sr, Ne, and rare-earth elements as alternative indicators for quantifying rock–water interaction.

As an alternative to simple metrics for estimating the extent of rock–water interaction, the application of geochemical modeling can provide quantitative and robust results, taking into account the complex interactions between various reactions (e.gAppelo and Postma 2004; Stradioto et al. 2020; Milesi et al. 2023). Geochemical modeling enables us to assess the coherence of anticipated dominant reactions and offers quantitative insights into the interaction between rocks and groundwater. It also facilitates establishing a connection between the evolution of groundwater hydrochemistry and the advancement of the rock–water interaction process. Despite the capabilities of geochemical models, only a limited number of studies have effectively employed them to replicate the water quality resulting from the interaction between groundwater and rock in crystalline aquifers, considering various reaction pathways (e.g., Fuoco et al. 2022; Manu et al. 2023).

The Precambrian crystalline basement outcrop and crystalline aquifers in São Paulo State serve as significant water sources due to their widespread occurrence in a substantial portion of Southeastern Brazil. In addition to São Paulo, the largest city in Latin America, other major urban centers in São Paulo State, such as Guarulhos and Campinas, also rely on the utilization of crystalline aquifers. The remarkable diversity in geotectonic contexts, lithology, geochemistry, and mineralogy observed in the crystalline basement of São Paulo State presents a significant opportunity to advance our quantitative understanding of rock–water interaction processes and their impact on groundwater hydrochemistry. Investigating these factors can provide valuable insights into the complex dynamics between rock and water, ultimately enhancing our knowledge of groundwater quality in the region. However, to date, a comprehensive study characterizing and modeling the hydrochemical evolution of crystalline aquifers in São Paulo State has not been undertaken. Therefore, this study has two main objectives. First, we aim to enhance our understanding of the mechanisms involved in the quantitative evaluation of hydrochemical evolution resulting from rock–water interactions. To achieve this, it is crucial to examine the underlying mechanisms and variables that influence the water quality of crystalline aquifers. Second, our objective is to develop a reliable model capable of accurately predicting the extent of rock–water interaction.

Materials and methods

Study area and regional geology

The study area is situated in the Crystalline Basement of the State of São Paulo, located in southeastern Brazil (Fig. 1). The rocks of the crystalline basement that occupies about 30% of the area of the State of São Paulo in its eastern region.

Fig. 1
figure 1

The location of the compiled groundwater samples and of the region where the geochemistry of rock is available

Full size image

The focus of this study will be on the Mantiqueira Province as most of the samples are located within this unit. The Mantiqueira Province (Almeida et al. 1981) constitutes a complex Brazilian orogenic system (Fuck et al. 2008) that extends parallel to the east coast of Brazil in an approximate northeast direction. It stretches from southern Uruguay to southern Bahia, covering a length of over 3000 km (Heilbron et al. 2004). The province is delimited by the São Francisco, Tocantins, and Paraná provinces, as well as the basins of the East Continental Margin Province (Hasui 2010). It is compartmentalized into three main segments and five distinct orogens, as described by Heilbron et al. (2004): (a) the northern segment, which includes the Araçuaí Orogen; (b) the central segment, comprising the southern portion of the Brasília Orogen, and the Ribeira and Apiaí orogens; and (c) the southern segment, encompassing the Dom Feliciano and São Gabriel orogens.

Stratigraphically, the Mantiqueira Province can be divided into Archean and Paleoproterozoic basement rocks, metasedimentary sequences deposited in Paleoproterozoic to Mesoproterozoic intracontinental basins, Neoproterozoic metasedimentary and metavulcano-sedimentary sequences, as well as Neoproterozoic granites and Neoproterozoic covers. The metamorphic grade recorded in the lands of the Mantiqueira Province ranges from greenschist facies to granulite facies (Heilbron et al. 2004).

Data collection and exploratory analysis

We analyzed a comprehensive dataset comprising 96 hydrochemical records extracted from six previously conducted studies carried out in the crystalline basement of São Paulo State, Southeastern Brazil. The sources of these data include Kiang et al. (2003), Iritani et al. (2011), DAEE (2011), Ezaki et al. (2014), Martins et al. (2018), and Teramoto et al. (2019).

The selection of data prioritized samples from the crystalline basement with plutonic or hypabyssal magmatic rocks, as well as rocks subjected to metamorphism in amphibolite facies or higher. Aquifers composed of metasedimentary/metabasic rocks with low metamorphic grade were excluded from the dataset. This ensured that the analyzed aquifers were predominantly composed of quartz, potassic feldspars, plagioclases, biotite, and amphibole. Additionally, we compiled available geochemical data to assess the compatibility of groundwater with the interacting rock. The locations of the sampled groundwater are depicted in Fig. 1, and the complete dataset is presented in Table S1 of the Supplementary Materials.

To setting the diverse geochemistry of rocks found in the crystalline basement of São Paulo, we compiled X-ray diffraction analyses from various rock samples published in previous works. The data sources include Spinelli and Barros (2009), Oliveira et al. (2008), Janasi et al. (2009), Sobrinho et al. (2011), Melo and Oliveira (2013), Duffles et al. (2013), and Toledo et al. (2018).

To conduct an initial assessment of compositional variations in the water samples, we performed hydrochemical typology and calculated the saturation index of the dissolved phase in groundwater. Additionally, we constructed a Cox Diagram to explore the distinct geochemistry of the rocks. Where both groundwater and rock analyses were conducted within the same aquifer, we endeavored to establish a correlation between them, seeking to identify any connections or relationships.

Geochemical conceptual model

We assume that the natural hydrochemistry of crystalline aquifers is governed by chemical interactions with the modal mineralogy of rocks. Consequently, the rock–water interaction is controlled by the dissolution of K-feldspar, plagioclase, pyroxenes, amphiboles, and biotite. Na+, K+, Ca2+, Mg2+, and HCO3− are assumed to be ions released into the water, while Cl−, SO42−, and NO3− are presumed to be derived from anthropogenic contamination.

Potassium feldspars (orthoclase and microcline) represent abundant mineral phases in granites and gneisses. The hydrolysis of potassium feldspars is responsible for the release of potassium into the water, as described in Eq. 1 (Appelo and Postma 2004)

$$2KAl{Si}_{3}{O}_{8}+2{CO}_{2}+11{H}_{2}O\to {Al}_{2}{Si}_{2}{O}_{5}{(OH)}_{4}+2{K}^{+}+2{HCO}_{3}^{-}+4Si{(OH)}_{4}.$$
(1)

Plagioclases are significant mineral phases found in granitic/gneissic rocks, especially those of the tonalite and granodiorite types. The hydrolysis reactions of the terminal members of the plagioclase series, anorthite and albite, are presented in Eqs. 2 and 3, respectively (Appelo and Postma 2004)

$$CaAl{Si}_{3}{O}_{8}+2{CO}_{2}+3{H}_{2}O\to {Al}_{2}{Si}_{2}{O}_{5}{(OH)}_{4}+{Ca}^{2+}+2{HCO}_{3}^{-}+4Si{(OH)}_{4}$$
(2)
$$2NaAl{Si}_{3}{O}_{8}+2{CO}_{2}+11{H}_{2}O\to {Al}_{2}{Si}_{2}{O}_{5}{(OH)}_{4}+2{Na}^{+}+2{HCO}_{3}^{-}+4Si{(OH)}_{4}.$$
(3)

Although amphiboles typically constitute a small percentage of granitoids (< 5%), these minerals can have a significant impact on groundwater composition due to their reactivity. Given the mineralogical composition of the examined rocks, it is likely that hornblende hydrolysis is the dominant source of magnesium in the water, as outlined in Eq. 4 (Helms et al. 1987)

$${Na}_{\mathrm{0,5}}{Ca}_{2}\left({Fe}_{\mathrm{1,3}}{Mg}_{\mathrm{2,6}}{Al}_{\mathrm{1,1}}\right)\left({Al}_{\mathrm{1,6}}{Si}_{\mathrm{6,4}}\right){O}_{2}+15{H}^{+}+{H}_{2}O\to \mathrm{0,5} Na+2{Ca}^{2+}+\mathrm{2,6}{Mg}^{2+}+\mathrm{1,3}{Fe}^{2+}+\mathrm{2,7}Al{(OH)}_{2}^{+}+\mathrm{6,4}{H}_{4}Si{O}_{4}\left(aq\right).$$
(4)

In the case of an open system with respect to CO2, which is expected to be true for fissured crystalline rocks, the pH is buffered by high PCO2 levels, and HCO3− serves as the main anion produced by silicate hydrolysis, as demonstrated in Eqs. 1–3. We interpret the variations in major cations and HCO3− as distinct stages of rock–water interaction, forming a continuum reflecting the extent of reaction with the rock.

Kaolinite is widely acknowledged as the primary mineral phase resulting from the weathering of crystalline rocks, such as granite and gneiss, in tropical and subtropical climates (Eqs. 1–4) (e.g., Jeong 2000; Gurumurthy et al. 2012; Liu et al. 2016; Fuoco et al. 2022).

Geochemical simulation

To assess the extent of rock–water interaction, we categorized certain samples into geological groups and ranked them based on the concentrations of major cations and anions. All numerical geochemical simulations in our study were performed using Geochemist's Workbench 10.0 (GWB 10) software (Bethke and Yeakel 2018).

Speciation calculations were conducted using the Espec8 program within GWB 10 to determine the saturation state with respect to dissolved minerals. The saturation index (SI) of a specific mineral is obtained by dividing the chemical activities of the dissolved ions (ion activity product, IAP) by their solubility product (Ksp), as described in Eq. 5

$$SI={log}_{10}\left(\frac{IAP}{{K}_{sp}}\right).$$
(5)

When the saturation index (SI) is lower than 0, the solution is unsaturated with respect to the evaluated mineral. An SI equal to 0 indicates that the solution is saturated with respect to the evaluated mineral. In the case of SI > 0, the mineral phase is supersaturated, leading to the precipitation of the mineral phase.

The reaction paths related to rock–water interaction was performed following the procedures described by Teramoto et al. (2019). The simulation of water composition variations as a function of rock–water interaction was conducted using the React program within GWB 10 and thermodynamic database thermo.tdat. These simulations aimed to reproduce changes in the chemical composition of water as water–rock interaction intensifies. Additionally, the simulations provided estimates of the mass balance of chemical species released into the water and the mass of secondary minerals produced from this interaction. Some reference samples were selected as criteria for adjusting the reacted mineralogy.

The initial solution was assumed to be pure water, with an average concentration of Cl−, SO42−, NO3−, and K+ based on the reference samples. The selection of mineral phases for reference was based on the geological context of the water sources, and the masses of each reacted phase were manually adjusted. Since the GWB database did not include calcio-magnesian hornblende, we used tremolite as a proxy for magnesian amphibole. To enable the precipitation of supersaturated mineral phases, we activated the precipitation function in GWB. The resulting precipitated mineralogy was also calculated.

Results

Geochemistry of rocks

The geochemistry of the Precambrian basement exhibits a wide range of diversity, which aligns with its genetic context. Metamorphic rocks found in amphibolite and granulite facies include ortho- and para-derived gneiss, migmatites, granulites, and amphibolites. Additionally, there are scattered occurrences of sin- or post-orogenic granites, predominantly classified as monzogranite and syenogranite according to the QAP classification. The lithological diversity reflects the varied geochemistry of the rocks, as illustrated in Fig. 2. The samples of Amparo Complex (Oliveira et al. 2008) fall within gabbro field, since these rocks are derived from the metamorphism of tholeiitic basalts. The samples from the Amparo Complex (Oliveira et al. 2008) fall within the gabbro field, as these rocks originate from the metamorphism of tholeiitic basalts. The geochemistry of the Guaxupé Complex exhibits broad variation (Melo and Oliveira 2013), encompassing the paleosome of migmatites composed of tholeiitic basaltic orthogranulite, as well as the neosome formed through partial melting. The neosome exhibits a calc-alkaline composition, ranging from granitic to tonalitic, with high SiO2 content. Additionally, samples from charnockite intrusions are also present.

Fig. 2
figure 2

Cox diagram illustrating the percentage of Na2O + K2O plotted against SiO2 in rock samples from selected lithostratigraphic units of the Crystalline Basement in São Paulo State. The diagram highlights the distinctiveness of rock geochemistry based on data acquired from literature sources

Full size image

Regarding the geochemistry of the rocks, our study focused on the major metallic elements that comprise the modal mineralogy of granitoids and gneiss, namely potassium, sodium, calcium, and magnesium. Generally, the granitoids exhibit a composition range from diorite to syenite, as shown in Fig. 2. The dominant element is K2O (> 4.5 wt%), followed by Na2O (0.28–7.11 wt%), CaO (0.05–4.22 wt%), and MgO (with average values below 0.5 wt%) (e.g., Janasi et al. 2009; Spinelli and Barros Gomes 2009; Duffles et al. 2013; Toledo et al. 2018). However, there are some rock formations that deviate from the typical granitoid composition. For instance, the Amparo Complex, which spans a large area in São Paulo State, is derived from the metamorphism of tholeiitic basalts. The average content of CaO, MgO, Na2O, and K2O in this complex is 10.66%, 11.89%, 1.24%, and 0.35% wt, respectively (Oliveira et al. 2008). Similarly, in the case of the paleosome of migmatites within the Guaxupé Complex, the average composition of CaO, MgO, Na2O, and K2O is 11.37%, 9.88%, 1.87%, and 0.85% wt, respectively (Melo and Oliveira 2013).

Characterization of evaluated samples

As illustrated in Eqs. 1–3, bicarbonate is the predominant anion resulting from rock–water interaction in an open system with respect to CO2. Consequently, all water samples in our study are classified as either Ca–Mg–HCO3 or Na–HCO3 types, as depicted in Fig. 3. Three samples deviate from this classification and are categorized as Na-SO4 types, indicating anthropogenic influence. The pH values of the water samples are consistently close to neutral, and the electrical conductivity (EC) exhibits a wide range of variation, spanning from 28 to 1,121 µS/cm.

Fig. 3
figure 3

Groundwater samples projected on the Durov diagram. The evaluated samples were classified into five distinct groups, according to TDS values

Full size image

The relationship between Electrical Conductivity (EC) and Total Dissolved Solids (TDS) is well established, with the former often used as an indicator for estimating the latter. By plotting the TDS values against the corresponding EC data (Fig. 4), we observed a linear relationship between the two parameters, which aligns with expectations. Through linear regression analysis with the intercept constrained at EC = 0 and TDS = 0, we determined that TDS can be estimated by multiplying EC by 0.82 in crystalline aquifers in the studied area.

Fig. 4
figure 4

Scatterplot of TDS versus EC values as a linear model

Full size image

To assess the formation of authigenic minerals resulting from rock–water interaction, we constructed a stability diagram focusing on Ca-rich and Na-rich silicates. However, due to missing data on SiO2 and Al3+ concentrations for some samples, they were not included in the diagram. Figure 5 illustrates that the majority of samples are located within the kaolinite field, indicating the prevalence of kaolinite formation. However, a few samples are classified under gibbsite and montmorillonite categories. It is worth noting that the samples falling within the montmorillonite field are characterized by high TDS values, suggesting a significant level of interaction with the rocks.

Fig. 5
figure 5

Stability diagram regarding aluminum silicate minerals

Full size image

Geochemical simulation

To gain insights into the hydrochemical evolution of groundwater resulting from rock interaction in crystalline aquifers in São Paulo, we examined two theoretical endmembers: the Ca–Mg–HCO3 and Na–HCO3 types. In the first set of simulations, we focused on samples from the Amparo Complex, which consists of metamafic and amphibolites from the Amparo and Itapira complexes. For the second set of simulations, we chose samples representing sodium-rich rocks and those with an alkaline affinity. The quantities of precipitated kaolinite and quartz per liter of water are presented in Table 1, providing valuable information on mineral precipitation during rock–water interaction.

Table 1 Mass of precipitated minerals calculated by simulations per liter of groundwater
Full size table

Figures 6 and 7 depict the Stiff diagrams comparing the results of the first and second sets of simulations, respectively, highlighting their similarities. The plagioclase series encompasses various intermediate types, spanning from predominantly calcium-based compositions (anorthite) to predominantly sodium-based compositions (albite). To replicate plagioclases with varying proportions of calcium and sodium in their compositions, we employed variable masses of anorthite and albite as the interacting minerals with water. Dissolved CO2 values were adjusted to represent different PCO2 conditions in the studied aquifers, resulting in hydrogen concentrations and alkalinities that closely resembled the observed values.

Fig. 6
figure 6

Comparison of stiff diagram of simulated solution with distinct stages of rock–water interaction and data from Amparo complex samples

Full size image
Fig. 7
figure 7

Comparison of stiff diagram of simulated solution with distinct stages of rock–water interaction and data of sodium-rich samples

Full size image

Figure 8 illustrates the evolution of CE, pH, TDS, HCO3− concentration, and the masses of precipitated kaolinite and quartz as a function of the reacted mass of rock (sum of reacted minerals). The figure shows that, for most of the presented parameters, there is a linear increase as the mass of reacted minerals increases. However, the pH exhibits a non-linear increase followed by a gradual rise, indicating the buffering effect of high PCO2. This behavior reflects the system's ability to maintain pH within a certain range despite variations in the reacted mineral mass.

Fig. 8
figure 8

Variation of pH, CE, TDS, HCO3−, and the amount of precipitated quartz and kaolinite as a function of the reacted rock (the sum of all dissolved minerals)

Full size image

Due to the buffering effect caused by high PCO2, water exhibits elevated concentrations of CO2, which undergo hydrolysis to form HCO3− during interactions with rock. Taking this into consideration, we calculated an increase in HCO3− concentration corresponding to the reacted rock (Fig. 8). As depicted in the figure, the HCO3− concentration shows a linear increase in relation to the equivalent mass of the reacted rock.

Based on the linear correlation between HCO3− concentration and the mass of reacted minerals, we developed two linear equations for predicting the theoretical mass of reacted rock for the Ca–Mg–HCO3 type (Eq. 6) and the Na–HCO3 type (Eq. 7). The distinction between these equations lies in the proportion of reacted amphibole, as these mineral lacks alumina

$${M}_{interacted\,rock}=0.002{\cdot HCO}_{3}^{-}-0.0408$$
(6)
$${M}_{interacted\,rock}=0.0032{\cdot HCO}_{3}^{-}-0.0647.$$
(7)

Discussion

Our analysis was based on previously published works. Figure 1 illustrates a significant data gap in the southern portion of the study area, which limits our analysis in that region. The concentration of samples in specific regions introduces a potential bias into our analyses; however, we have confidence in the representativeness of our samples. They were collected from the principal lithostratigraphic groups encompassing the crystalline basement of the State of São Paulo.

During the hydrochemical characterization, we observed that all groundwater samples exhibited variations between the Ca, Mg–HCO3 and Na–CO3 types, as depicted in Fig. 3. This is likely due to the high annual precipitation in the studied area (> 1200 mm/year) and the subsequent leaching process. Most samples fall within the kaolinite and gibbsite fields, as shown in Fig. 5. These samples align along two reaction paths: one composed of orthogneisses from the Amparo Complex and the other involving sodium-rich rocks. To simulate the rock–water interaction within the Amparo Complex, we utilized amphiboles and calcium-rich plagioclases as reactant phases (Fig. 6). Conversely, to replicate the evolution of Na-HCO3 types, the presence of sodium-rich plagioclase alone was sufficient (Fig. 7). In our simulations, the secondary mineralogy resulting from the dissolution of primary silicates consists exclusively of quartz and kaolinite (Table 1).

In addition to rock mineralogy, the climatic and hydrological conditions also exert an influence on rock–water interaction. Therefore, by conducting a comprehensive analysis of our study along with previous research that employed similar methodologies, we significantly advance the current understanding of the mechanisms that govern rock–water interaction in crystalline aquifers. Despite the recent publication of several relevant works deducing the mechanisms controlling water quality in crystalline rocks (e.g., Cho and Choo 2019; Cuccuru et al. 2020; Kumar and Kumar 2020; Abbas et al. 2021; Mao et al. 2022; Kouser et al. 2022; Ghalit et al. 2023), only a few of them (e.g., Fuoco et al. 2022; Manu et al. 2023) have utilized geochemical modeling to reproduce the reaction paths of rock–water interaction, enabling direct comparisons with our results.

When comparing our findings with previous studies that utilized similar modeling approaches (Fuoco et al. 2022; Manu et al. 2023), certain similarities arise, particularly in recognizing the significance of plagioclases as a key mineral group influencing hydrochemistry. From this perspective, the concentrations of Na+ and Ca2+ in groundwater can be predominantly influenced by the Na/Ca ratio in the reacted plagioclases. However, it is important to exercise caution when interpreting this statement, as the dissolution of plagioclases, as established by Stober and Bucher (1999), is notably incongruent. This means that while Ca is released into the groundwater, Na remains relatively unaffected within the crystal structure. Consequently, the proportions of reactive albite and anorthite adjusted in geochemical simulations may not accurately reflect the true composition of plagioclases.

In contrast to our findings, Fuoco et al. (2022) demonstrated in their geochemical simulations that biotite plays a significant role as the second most important reacted mineral, releasing magnesium into the groundwater. The use of biotite as a source of Mg2+ in crystalline aquifers is supported by several studies (e.g., Soulsby et al. 1998; Elango 2007; Jeong 2001; Venkatramanan et al. 2016), considering its abundance and reactivity. The average concentrations of Na+, K+, Ca2+, and Mg2+ are 3.48 mmol/L, 0.056 mmol/L, 2.50 mmol/L, and 0.94 mmol/L, respectively. Thus, our samples display a relatively high proportion of magnesium, comparable to that of calcium and sodium, while showing a significantly lower concentration of potassium. This challenges the notion of biotite being the primary source of magnesium. Based on our model, amphiboles appear to be a more suitable candidate as the main source of magnesium.

According to the simulations by Fuoco et al. (2022), K-feldspar is postulated as the third most important mineral contributing to water quality. Since K-feldspar is one of the most abundant minerals in granitoids and gneiss, it is theoretically expected to have a significant impact on the hydrochemistry of crystalline aquifers. However, during our simulations, we observed that the contribution of K-feldspar to the hydrochemistry of crystalline aquifers is minimal, despite its abundance in granite/gneiss. This is indicated by the low activity of K+ in almost all samples. We ruled out the possibility of potassium-rich clays, such as montmorillonite and illite, precipitating due to their low saturation indices. The preferential dissolution of plagioclases over K-feldspar can be attributed to the low reactivity of the latter mineral, with dissolution rates at least two-to-three orders of magnitude lower than those of plagioclases (e.g., White et al. 2001; Kampman et al. 2009; Zhu 2005).

In some locations where both the rock and geochemistry data were available, a comparative evaluation was possible. We note that groundwater samples collected at the occurrence of Amparo and Guaxupé Complex in the north portion of studied area, the concentration of calcium and magnesium is high, agreeing with rock geochemistry. On the other hand, in the locations markedly by occurrence of granitoids with composition varying between syenogranite and granodiorite, the sodium comprises the most important dissolved cation, with minor but important concentration of calcium and magnesium. In this case, our study reveals a discrepancy between groundwater and rock compositions, highlighting the need for a more nuanced interpretation.

As rock–water interaction is the primary mechanism influencing the water quality of crystalline aquifers, various metrics have been proposed to measure the extent of these processes. Our research demonstrates the suitability of HCO3− as a proxy for quantifying the extent of rock–water interaction, as this anion is exclusively generated through silicate hydrolysis in the studied area. However, it is important to note that this holds true under specific conditions, particularly in an open system with respect to CO2, where the pH remains close to neutral and there is no carbonate precipitation or another external source of bicarbonate (e.g., calcite).

Conclusions

Through the compilation of hydrochemical data from groundwater in Precambrian crystalline rocks and the application of geochemical modeling, we have gained valuable insights into the primary mechanisms governing the chemical composition of these waters in the Southeast region of São Paulo State, Brazil. In our approach, we employed a reacted mineralogy adjustment in numerical simulations to accurately reproduce the two main hydrochemical evolution trends.

Our simulations revealed that the groundwater in crystalline aquifers in São Paulo consists of Na–HCO3, Ca–Mg–HCO3, or intermediate hydrochemical types. These types can be attributed to two distinct evolutionary trends in hydrochemistry. The Ca–Mg–HCO3 types can be linked to the dissolution of amphiboles and plagioclases, while the sodium bicarbonate type can be reproduced by the dissolution of sodium-rich plagioclases. Consequently, in the studied area, the rock–water interaction can be simplified as involving the hydrolysis of plagioclase and amphiboles, leading to the precipitation of secondary mineralogy, such as kaolinite and quartz.

It is worth noting that despite the abundance of K-feldspar in the granitoids and its high percentage in the rocks, the concentration of K+ in groundwater is very limited. While our results align with other studies that have reproduced the hydrochemical evolution resulting from rock–water interaction in crystalline aquifers, it is important to recognize that important features can vary significantly, and models should be applied with caution, considering the specificity of each site.

Another significant finding is that contrary to common belief, the hydrochemistry of the evaluated samples does not mimic the mineralogy or geochemistry of the granitic/gneiss rocks. Instead, the dissolved cations in groundwater mainly originate from unstable and reactive minerals, primarily represented by amphiboles and plagioclases. Finally, our simulations theoretically demonstrate that HCO3− can be utilized as a parameter to measure the extent of rock–water interaction in the studied area, as it is the main species produced through silicate hydrolysis under an open system with respect to CO2.