Experimental study on the water–rock interaction mechanism in a groundwater heat pump reinjection process

In recent years, global warming and climate change have been changing the human lifestyle and have also had a profound impact on the global economy, ecological environment and groundwater resources (Pandey et al. 2020; Khan et al. 2021; Valipour et al. 2021). The carbon emissions caused by the large consumption of traditional petrochemical energy sources is one of the most critical factors. Sustainable and clean energy, such as solar energy, wind energy and geothermal energy, has become a good measure to solve the energy crisis and alleviate environmental pollution (Sanner et al. 2003; Zhou et al. 2015). In particular, geothermal energy has attracted more and more attention due to its abundant reserves, stable reliability and low carbon dioxide emissions (Ni et al. 2015; Xu et al. 2016; Wang et al. 2017). Groundwater heat pump (GWHP), which is a shallow low-grade geothermal energy technology that uses heat pump units to exchange heat between groundwater and the ground space for the heating or cooling of buildings, as well as for valuable water/waste treatment applications, such as groundwater desalination, has been widely developed and applied in many countries (Luo et al. 2016; Makasis et al. 2018; Panagopoulos 2021; Yang et al. 2021). However, during the long pumping and recharge operations of GWHP, physical, chemical and biological blockage occur in the recharge wells and surrounding aquifers (Hepbasli et al. 2014; Athresh et al. 2016; Fetzer et al. 2017), resulting in a sharp decline in the efficiency of the heat pump system, which has become a key problem that affects the project operation. The chemical blockage usually refers to the chemical precipitation caused by the oxidation and corrosion of iron and manganese ions present in groundwater. Nevertheless, the chemical blockage of aquifers induced by the water–rock reaction is difficult to repair, and research on the underlying mechanism has begun to attract more and more attention.

At present, research on the water–rock interaction mechanism in our country and abroad has mainly focused on dam rock engineering, mineral exploitation and artificial groundwater recharge (Hashemi et al. 2015; Shi et al. 2016; Hu et al. 2019; Li et al. 2019). Dou et al. (2021) have studied the influence of the water–rock interaction on the mechanical properties of granite, analysed the evolution of the shear damage volume with the immersion time and proposed a shear volume rough crack model considering the immersion time. Wang et al. (2021) investigated the microfracture-pore structures and water–rock interactions in the lithofacies from the Lower Eagle Ford Formation and compared their porosity and analysed their pore-throat size distribution, pore size distribution, pore connectivity, water adsorption and desorption behaviour. In mining engineering, the dissolution, precipitation and redox reactions of clay minerals, pyrite and carbonate minerals in flowing groundwater are a water–rock interaction geochemical process (Du et al. 2018; Phan et al. 2018; Zhang et al. 2019). Marcon et al. (2017) carried out static simulation experiments on shale-containing calcite and pyrite with brine under high temperature and pressure and the results showed that carbonate and clay were dissolved, anhydrite and secondary clay were precipitated and there was a balance between the montmorillonite and barite contents. Zhang et al. (2021) used a fluorescence excitation–emission matrix and parallel factor analysis to characterize the mine water flows in goaf and showed that clay minerals in rocks had significant effects on the binding characteristics of dissolved organic matter and metal ions under the water–rock interaction. Cui et al. (2020) developed a simple conservative mixing model and a progressive conservative mixing model to evaluate the impacts of geochemical processes on the indicators of hydraulic fracturing flowback fluids and suggested that the fluid–shale interactions that occurred during hydraulic fracturing and flowback were critical for hydrochemical evolution. When surface water is reinjected into the ground, the aquifer may be affected by hydrochemical reactions. Vanderzalm et al. (2010) used urban pluvial flood water to recharge a carbonate aquifer and found that the reactivity of the recharge water itself was the main factor affecting the process of mineral dissolubilisation and precipitation. Walter et al. (2017) studied the Precambrian shield rocks and Pleistocene deposit aquifers in Quebec province, Canada, and suggested that the groundwater in the crystalline bedrock naturally evolved from a recharge-type to a brackish groundwater due to water–rock interactions, while groundwater evolution in confined aquifers is dominated by water–clay interactions.

However, in geothermal energy extraction, the recharge of surface water disrupts the water–rock balance of aquifers, leading to new mineral dissolution and precipitation reactions, and changing the permeability of reservoirs. Tarcan et al. (2016) analysed scale samples of water from the Kızıldere geothermal field in Turkey, and found that the scale types included calcite, artesite, dolomite, amorphous silica and barite. The relationship between low-temperature recharge and scale formation was also explored. Bai et al. (2012) compared the effects of the flow rate and temperature on the enhanced geothermal system, and analysed the changes in pH, TDS, reaction rate and dissolution equilibrium under the action of the water–rock interaction. As computer technology matures, numerical simulations based on hydrogeochemical theory have been applied to study the water–rock interaction mechanism (Rinck-Pfeiffer et al. 2000; Bedrikovetsky et al. 2009). Nitschke et al. (2017) used the temperature–hydrodynamic–chemical coupling model to simulate the volume and thickness of the scaling layer and found that the initial reservoir brine composition had a great influence on the scope of scaling. García-Gil et al. (2016) established two transient reaction transport models coupled with groundwater flow, heat, solute transport and chemical reactions to predict changes in the calcite precipitation rate and porosity in aquifers caused by temporal and spatial variations.

The above research on the mechanism of the water–rock interaction was mostly done based on experimental analysis and numerical simulation at the geochemical macro-scale, while the analysis of the rock sample chemical composition, micro morphology and crystal structure at the micro-scale has been relatively scarce. In particular, in the process of geothermal water reservoir reinjection, research on the combination of the micro-mechanism of the water–rock interaction and on the macro-geochemistry is still lacking. Here, aiming to solve the chemical blockage of geothermal aquifers induced by GWHP reinjection, the microscopic scale changes in the contents of mineral components in aquifer sand and the evolution of the ion contents in recharge water are explored and the water–rock interaction process of recharge blockage is revealed. In this paper, an indoor seepage sand column was designed to simulate the aquifer of a GWHP system and geothermal water and aquifer sand samples of the actual heat pump project were collected. Through the analysis of the microscopic morphology, mineral composition and group structures of the sand before and after the experiment, combined with the geochemical reaction of the recharge water, the water–rock interaction mechanisms, such as the dissolution and precipitation of minerals and the cation exchange induced by the recharge using the heat pump, were explored, providing a scientific basis and technical support for the prevention and control of the recharge blockage of GWHP systems.

The experimental sand samples of the GWHP recharge simulation were taken from a drilling well site in Qiaocheng District, Bozhou City, Anhui Province, China, with a sampling depth of about 350 m. In order to avoid the influence of sunlight on the composition of sand, the collected samples were placed in the shade indoors to remove the particles with a size greater than 2.0 mm. Then, the quantitative sand was passed through the standard sieves with pore sizes of 2.0, 1.0, 0.5, 0.25, 0.1 and 0.075 mm, respectively. The percentage of particles of each size was calculated according to the quality of the retained sieve soil for each pore size, and the particle gradation of sand was determined. Finally, the average particle size of sand was 0.34 mm, belonging to the fine sand category.

The experimental recharge water samples were collected from a GWHP pumping well in Yingzhou District, Fuyang city, Anhui Province, China, at a depth of about 1,500 m and a temperature of 53 °C. Table 1 shows that the main cations present in recharge groundwater were Na⁺, Ca²⁺ and Mg²⁺ and the main anions were Cl⁻, and ⁠, which corresponds to a Cl–Na hydrochemical type. In addition, the turbidity of the water sample was 27.20 NTU, the total hardness was 1,322 mg L⁻¹ and the TDS value was 12,678 mg·L⁻¹, all of which were higher than the limits imposed for class V of the ‘Standard for groundwater quality’ (GB/T 14848-2017); thus, the content of chemical components was high.

The GWHP reinjection simulation experimental device is shown in Figure 1. The seepage sand column was made of plexiglass, with a length of 2,000 mm and a pipe diameter of DN120 × 10. At a distance of 100 mm from the two endpoints, a set of sampling ports, pressure measuring pipes and temperature sensor interfaces were set within 200 mm intervals. Moreover, the three interfaces set on the left and right sides and above the outer wall of the column, respectively, were located in the same section, with a total of ten groups, which were marked P1, P2,… and P10 from the inlet to the outlet. Furthermore, the outer wall of the sand column was covered with a heating plate for continuous constant temperature heating, and a low-temperature thermostatic bath was used to cool the recharge water, so that the process of water seepage in the sand column simulated the recharge of the aquifer by the GWHP system.

• (1)

  Filling and saturation. The sieved sand sample was used to fill the sand column in equal layers and was compacted evenly section by section, and both ends were tightened using a flange after uniform compaction. Then, the test water was filled slowly from the bottom to the top and saturated for a week to fully empty the gas in the sand layer.

• (2)

  Experimental operation. The closed water tank was placed in the low-temperature thermostatic bath, whose working temperature was set to 4°C. Then, the geothermal water was pumped to the cooling water tank at a constant flow rate (20 mL·min⁻¹) using a peristaltic pump, and then supplied to the sand column after low-temperature cooling. Meanwhile, the heating plate was used to heat the column to a constant temperature (50°C) and cooling water was used to control the gradual rise in temperature during the process of sand column seepage. Finally, the effluent was returned to the circulating water tank to perform the operation of heat pump recirculation continuously.

• (3)

  Recording and sampling. When a stable state was obtained, the water pressure and temperature at the monitoring points were measured using a pressure tube and temperature sensor, and the water quality at each point of the sand column could be sampled and detected via the sampling port. According to the actual operation of the GWHP project, the peristaltic pump worked for 12 h in each recharge cycle and stopped for 12 h, while the heating plate was continuously used for 24 h to maintain a constant temperature. The water was sampled every 72 h, and a sampling volume of 5 mL was collected at each monitoring point, and the test operation cycle lasted for 63 days in total.

In the experiment, 44 sand samples, 214 water samples and 2,140 water quality indexes were collected and analysed.

(1) Scanning electron microscope (SEM) and energy dispersive spectrum (EDS)

High-resolution field emission SEM (Regulus 8230, Hitachi, Japan) was used to analyse the micro morphological characteristics, point and regional spectra of sand powder samples and to determine the contents of K, Ca, Na, Mg, Fe, C, Al, Si and Mn. The test conditions used were the following: secondary electron resolution ≤0.6 nm (15 kV); magnification range: 200–2,000,000.

(2) X-ray diffraction (XRD) analysis

A fixed XRD (X-Pert Pro MPD, PANalytical, Netherlands) target was used to qualitatively and quantitatively analyse the mineral phase in the sand samples. The test conditions were as follows: Cu target, standard velocity scanning, step length of 0.02°, residence time of 20 s, 2θ starting angle of 5° and a cut-off of 70°.

(3) Fourier transform infrared spectroscopy (FT-IR)

FT-IR (Nicolet 6700, Thermo Nicolet, USA) was used to detect the types of molecular chemical bonds or functional groups of minerals in the sand samples. The test conditions used were a spectral range of 7,800–50 cm⁻¹, a resolution higher than 0.09 cm⁻¹ and an ASTM standard linearity higher than 0.07%.

(4) Inductively coupled plasma mass spectrometry (ICP-MS) and ion chromatography

ICP-MS (ICAP-RQ, Thermo Fisher, USA) was used to detect the levels of cations, such as K⁺, Na⁺, Ca²⁺, Mg²⁺, Fe²⁺ and Fe³⁺, in the recharge water samples; the levels of anions, such as F⁻, Cl⁻, and were determined using ion chromatography (Diane ICS-5000 + , Thermo Fisher, USA), and the level was determined via acid–base titration. A TDS tester (HI98130, Hanna, Italy) was used to measure the TDS and pH value.

A worn mineral surface usually has edges and micro cracks; therefore, atomic reconstruction and relaxation make the mineral surface have characteristic activity and adsorption qualities (Hird et al. 1996; Fuller et al. 2015). The original sand and the sand samples after the reinjection test were characterized using SEM, as shown in Figure 2(o1) and 2(o2). The original sand samples had different forms of scattered massive particles, among which broken small particles had a few to tens of microns, and the large particles had a size of about 100–600 μm. There was a layer of dense irregular banded or flake particles on them, and a massive quartz crystal plane could be seen locally. By comparing the morphology of the sand samples collected at different points of the column after reinjection, as shown in Figure 2(a1)–2(d1), there were a few broken small particles, the surface of the large ones was still covered with broken banded particles, and the wrinkle texture was thickened and deepened. It can be seen from Figure 2(a1) and 2(b1) that pores or holes appeared on the surface of many particles, whose size was about 5–10 μm. In Figure 2(c1), the local surface of particles presented a fracture gully, while the crystal plane was mostly filled with unevenly broken particles, as seen in Figure 2(d2). Therefore, there were certain differences in the internal structure of the test sand samples before and after reinjection; in particular, in the front of the sand column, the morphological characteristics of the particles changed significantly between some sampling points.

After the reinjection test, the element types and relative contents were analysed through the EDS of each sampling point of the sand column and the original one. It can be seen from Figure 3 and Table 2 that the main elements of the test sand were C, Si and Al, accounting for more than 80% of the total contents. At P1, P2, P5, P8 and P10, the element content order was C > Si > Al, while the element content order was Si > C > Al at P3, P4, P6, P7 and P9. Secondly, Ca and K also accounted for a large proportion of the elements in some test points; for example, the content of Ca at P5 was 26.54% and the content of K at P7 and P9 was 10.68% and 10.89%, respectively. The contents of Fe, Mg and Na were relatively low but there were significant differences at some points. Furthermore, the order of the contents of elements in the original sand was: Si > C > Al > K > Fe > Na > Mg > Ca, and the total content of C, Si and Al was higher than 90%. Therefore, it was speculated that the sand samples were mainly aluminosilicate and carbonate minerals, and the changes in the relative content of element at each point in the sand column might be caused by the geochemical reaction of minerals.

The XRD spectra corresponding to the main detection points in the sand column and the original sand are shown in Figure 4. Combined with the spectrum of all the sampling points, we observed that the mineral component contents changed to some extent after reinjection. Firstly, except for P10 at the end of the column, the mineral composition of the other points decreased. Among them, chlorite had no corresponding diffraction peak in P1–P3 and at P5, hematite had no diffraction peak at P1, P5 and P8, and vermiculite had no diffraction peak in the whole P1–P9 section. Meanwhile, the spectra of albite, orthoclase and calcite showed great changes at individual sampling points, and the changes or even disappearance of diffraction peaks indicated that the above minerals might have undergone incomplete dissolution during reinjection. Secondly, along the flow direction of the sand column, the diffraction peak intensity at each point showed a weakening trend. The maximum diffraction intensity of quartz at P1, P3, in P5–P6 and at P8 was about 6,000, being especially high at P2, where it reached 12,550, while the diffraction peak intensity at P7 and in P9–P10 was about 3,000, and the diffraction intensity of other minerals was also relatively weakened. Furthermore, according to the comparison of the diffraction patterns observed in Figure 4(e) and 4(f), the phase composition was consistent, while chlorite, hematite and vermiculite were not found in the front of the sand column. Therefore, it was speculated that the water chemical reaction induced by heat pump recharge mainly occurred in the front end.

Table 3 shows the relative contents of the mineral components at each point of the sand column and of the original sand after reinjection. Except for the decrease in the mineral content in P4, the quartz content at the other points increased, especially at P10, where it increased by 13.28%. The content of albite decreased at all points, except for P4 and P6, with a maximum decrease of 13.97%. In addition, the calcite content in P3–P4 and P7–P10 was increased, with a maximum increase of 1.2%, while it decreased at other points, with a maximum decrease of 0.54%. Moreover, the anhydrite content decreased in the whole recirculation zone, with a maximum decrease of 0.8% at P2, while the dolomite content increased at P4, P7, P9 and P10 and decreased at other points by a maximum of 0.37%.

Combined with Figure 5, it can be seen that quartz was basically formed in the whole section of the sand column during the recharge process. While calcite and dolomite were dominant at the back end, albite and anhydrite were dominant in the dissolution of the whole section. Furthermore, the clay minerals chlorite and kaolinite were hydrolysed. Therefore, in the process of GWHP recharge, water–rock reactions, such as dissolution and precipitation, were obvious. Daniele et al. (2013) used a combination of GIS software and geochemical modelling to describe and identify the physicochemical processes of carbonate aquifers in four different regions, explaining the variations in groundwater composition with water–rock interactions, such as calcite precipitation, dolomite and gypsum dissolution. Craw (2000) conducted a long-term hydrochemical monitoring of the water–rock interaction in a large schist debris dam in Macraes mine, Otago, New Zealand, and found that the concentration of sulphate, bicarbonate, calcium and magnesium in the water gradually increased; meanwhile, the combined decomposition rate of calcite and chlorite was controlled by the oxidation rate of pyrite. However, relevant studies had shown that the weathering of carbonate rocks is not only controlled by chemical dissolution, but that the particle separation in the water–rock interaction process might also be caused by the repulsion between calcite particles, although this needs to be confirmed in further studies (Levenson & Emmanuel 2017).

As shown in Figure 6, there were certain differences in each spectrum in the wavelength range of 600–2,500 cm⁻¹, in which multiple characteristic peaks appeared: the peaks at 2,335 and 835 cm⁻¹ are the characteristic absorption peaks of minerals, which may be related to the stretching vibration and bending vibration of the groups in calcite, dolomite and other carbonate minerals; the peaks at 2,160 cm⁻¹ and 1,410 cm⁻¹ are the characteristic absorption peaks of Al–OH minerals, which may be related to the stretching vibration and bending vibration of the Al–OH groups in aluminate silicate minerals, such as kaolinite and muscovite; the peak at 1,004 cm⁻¹ is the characteristic absorption peak of minerals, which may be related to the stretching vibration of the groups in aluminosilicate minerals such as albite; the 1,160 cm⁻¹ peak is the characteristic absorption peak of minerals, which may be related to the stretching vibration of the groups in sulphate minerals such as anhydrite; the peaks at 778 and 693 cm⁻¹ are the characteristic absorption peaks of Si–OH minerals, which may be related to the bending vibration of the Si–OH groups in minerals such as quartz. The above analysis confirmed the existence of various minerals in the sand samples and provided multiple active sites for the water–rock reaction.

Nevertheless, compared with the original sand samples, it was found that the characteristic absorption peaks of the P1–P10 FI-IR spectra curves at 693, 778, 835, 1,004, 1,160, 1,410, 2,160 and 2,335 cm⁻¹ were weakened, and some of them were significantly decreased (Senthil Kumar & Rajkumar 2014). This indicates that during the recharge process, the minerals in the sand column suffered a dissolution reaction, and the density of the mineral adsorption layer was weakened, reducing the adsorption amount (Wu et al. 2021). Additionally, the FI-IR spectral curves of P1–P10 showed significant differences in the increase and decrease of multiple characteristic absorption peaks. The intensity of the characteristic absorption peak at 835 cm⁻¹ showed an increasing trend in the P6–P9 section, indicating that carbonate minerals such as calcite and dolomite might precipitate at the back end of the sand column, resulting in an increase in the adsorption capacity. The intensity of the characteristic absorption peak at 778 cm⁻¹ increased in the P2–P3 and P6–P9 sections, which means that the quartz content in these sections of the sand column increased. While the intensity of the characteristic absorption peak at 1,004 cm⁻¹ was weakened and the peak value also had a certain deviation in the multi-segment of the sand column, it was suggested that new chemical or chemisorption reactions might occur in the aluminosilicate minerals, such as albite, during recharge.

The statistics of the main water quality parameters of representative sampling points in the recharge cycle are shown in Table 4: the TDS concentration ranged from 11,740 to 15,621 mg·L⁻¹, with an average of 14,023 mg·L⁻¹, which belongs to a saline type, according to the ‘Standard for groundwater quality’ (GB/T 14848-2017). Therefore, it would be more conducive to the water–rock reaction in a recharge experiment. The content of TDS in raw water reached 12,678 mg·L⁻¹, and a high concentration of dominant ions played a positive role in promoting the incomplete dissolution of minerals, as well as influencing the direction of the chemical equilibrium in the process of cation exchange adsorption. In general, with the increase in TDS concentration, the carbonate of calcium and magnesium carbonate first saturated the solution and then precipitated. As its content continued to increase, the sulphate of calcium sulphate also saturated the solution and precipitated. The higher the concentration of TDS, the more soluble sodium chloride was dominant.

Furthermore, the order of the average cation concentration in the recharge water was Na⁺ > Ca²⁺ > Mg²⁺ > K⁺, and the average content of Na⁺ accounted for 90.51% of the total cation mass fraction. Meanwhile, the order of the average anion concentration was Cl⁻ > > ⁠, and the average content of Cl⁻ and accounted for 82.74 and 16.58% of the total amount, respectively. Therefore, Na⁺ cation and Cl⁻ and anions were the dominant ions, and the recharge water belonged to the Cl–Na chemistry type. Moreover, the pH value of the detected water was 6.05–8.54, which is neutral or weakly alkaline water.

As can be seen from Figure 7, the concentration of Na⁺ and Cl⁻ increased greatly and the changing trend was basically synchronous. In the first 3 days of reinjection, the concentration increased rapidly, then gradually decreased, and slowly increased until stabilization was achieved in the later stage of reinjection. Among them, the recharge period from 30 to 63 days at P1 showed a fluctuating upward trend, the concentration values at P4 increased from day 21, while those at P7 and P4 showed large fluctuations near day 51; the overall concentration change at P10 was basically null. This was mainly due to the dissolution of halite, albite and other minerals in the sand sample at the front end of P1–P3 in the initial stage of the reinjection, in which the concentrations of Na⁺ and Cl⁻ increased in the recharge water. However, in the P9–P10 segment at the end of the sand column and in the late stage of the reinjection, the change in concentration tended to be stable with the dissolution slowing down and with co-occurring hydrochemical reactions of minerals, such as ion exchange adsorption.

The Ca²⁺ and concentration changes showed an overall decreased shock curve, and increased or decreased synchronously at multiple recharge points, indicating that the dissolution and precipitation of calcite occurred during reinjection. Among them, the concentration of Ca²⁺ increased significantly at days 12 and 51 days at P4, and at days 15 and 30 at P7, which may be due to the dissolution of anhydrite in the sand samples. Meanwhile, the concentration of basically increased steadily during the recharge process. In addition, the concentration of Mg²⁺ increased slightly, and that of K⁺ showed a basically horizontal curve.

The TDS parameter comprehensively reflects the concentration of a combination of main ions in groundwater, which usually changes with the ion composition. In the pumping reinjection process of the GWHP system, due to the water–rock reaction between the reinjection water and minerals, the concentration of main ions in water changes regularly with the TDS (Vinson et al. 2013; Cabassi et al. 2019; Missi & Atekwana 2020).

Figure 8 shows that with the increase in the concentration of TDS, the concentration of Na⁺ increased significantly and the concentration of the main cation components, Mg²⁺ and Ca²⁺, increased slightly, and that of K⁺ remained basically unchanged. Among them, the linear correlation between Na⁺ and TDS was the best, and the correlation coefficients R² at P1, P4, P7 and P10 were 0.9312, 0.8351, 0.9266 and 0.9182, respectively. Therefore, Na⁺ had the greatest influence on the increase in TDS in the reinjection process, meaning that the dissolution of Na minerals in the sand samples played a major role in the increase in TDS concentration. In addition, the Ca²⁺ and Mg²⁺ mass concentration changes reflected the small correlation with TDS, and the correlation between the K⁺ concentration and TDS was weak.

It can be seen from Figure 9 that in the main anion components, the concentration of Cl⁻ and showed a good linear relationship with the increase in TDS, while that of was basically unchanged. At P1, P4, P7 and P10, the correlation coefficients R² between the Cl⁻ concentration and TDS were 0.9929, 0.9872, 0.9826 and 0.9874, respectively, and the correlation coefficient R² between the concentration and TDS was 0.9206, 0.9179, 0.8853 and 0.9017, respectively. Therefore, Cl⁻ and contributed to the main TDS concentration in the whole recharge process, and the correlation between the Cl⁻ concentration and TDS was the best. It was also shown that the dissolution of minerals containing Cl and S is likely to occur in the water–rock reaction.

Based on the above analysis, the evolution of the ion concentration with the TDS indicated the dissolution and precipitation of minerals under the water–rock interaction. In combination with the change in mineral composition analysed using XRD at each sampling point in the sand column, the dissolution of halite, albite, anhydrite and chlorite, as well as calcite precipitation, occur in the process of heat pump recharge. The chemical reaction equation is as follows:

Liu et al. (2018) analysed the salinization process of underground fresh water in the Shuhe River Basin, coastal Jiangsu Province, China, and concluded that under the action of ion exchange, clay minerals continuously release Na⁺ and adsorb Ca²⁺ and Mg²⁺, resulting in a gradual increase in the relative content of Na⁺ and K⁺, and in a decreased content of Ca²⁺ and Mg²⁺ in regional groundwater. Relevant studies had also verified the exchange and adsorption of cations such as Na⁺ and Ca²⁺ under the water–rock reaction in an underground aquifer with a high TDS concentration (Brikowski et al. 2014; Li et al. 2018). In addition, the concentration of [Na⁺ + K⁺] and [Mg²⁺ + Ca²⁺] at P4 increased synchronously with that of TDS, indicating that the cation exchange adsorption at this point was not obvious, and that other chemical reactions might occur. Meanwhile, it also indicated that the water–rock interaction was a complex hydrogeographic reaction process.

• (1)

  The variation in the mineral composition at each point reflected the possible water–rock reaction, especially at the front of the sand column. The content of albite decreased at all points, except for P4 and P6; the anhydrite and chlorite contents decreased throughout the interval, while the quartz content increased at all points, except for P4. The calcite content increased in the P3–P4 and P7–P10 segments, and the dolomite content increased at P4, P7, and in P9–P10. In addition, the stretching vibration and bending vibration of the ⁠, Al–OH, ⁠, and Si–OH groups, as observed using infrared spectroscopy, provided multiple active sites for the water–rock reaction of minerals.

• (2)

  The water–rock interaction induced by heat pump recharge was a mineral dissolution and precipitation process. Firstly, Na⁺, Cl⁻ and in the water samples were the dominant ions, and the changing trend was basically synchronous. In general, they increased first and then stabilized; consequently, the dissolution of halite and albite might occur. Secondly, the Ca²⁺ and contents increased or decreased simultaneously at multiple recharge points, and the concentration of Mg²⁺ also changed to some extent, which might have led to the dissolution and precipitation of calcite and dolomite. Moreover, the dissolution of anhydrite gradually increased the concentration of and supplemented the recharge water with Ca²⁺. Furthermore, chlorite and other aluminosilicate minerals might be dissolved in the reinjection process.

• (3)

  Regarding the chemical composition of the water samples, the linear correlation between the concentration of cation Na⁺ and TDS was the best, and the concentration of anion Cl⁻ and increased linearly with the TDS, indicating that the dissolution of the Na and S minerals played a decisive role in the TDS concentration. Therefore, the increase in TDS in recharge water was mainly due to the dissolution of halite, albite and anhydrite. Additionally, during heat pump recirculation, [Na⁺ + K⁺] increased linearly with the TDS, while [Mg²⁺ + Ca²⁺] decreased linearly with the TDS, indicating that the increase in Na⁺ and K⁺ and the relative loss of Mg²⁺ and Ca²⁺ under the water–rock reaction correspond to a cation exchange adsorption process.

• (4)

  The chemical blockage of aquifers during recharge is the most complex and urgent problem faced when operating a GWHP system. This study revealed the water–rock interaction process of aquifer chemical blockage induced by GWHP recharge. For this, we used micro-scale analysis combined with geochemical principles especially suitable for the heat pump recharge blockage of porous saline aquifers with a high concentration of TDS. In the next step, different types of water and sand samples were collected in multiple regions, and parameters such as the temperature and pressure were changed to conduct an indoor test of the heat pump recharge. Moreover, the appropriate heat pump project was selected to carry out the recharge blockage field test, and the technical measures and engineering feasibility of slowing down the heat pump recharge blockage caused by the water–rock interaction were further studied.

This research work was supported by National Natural Science Foundation of China (Grant No. 42107162), Natural Science Foundation of Anhui Province (Grant No. 1908085QD168) and the Fundamental Research Funds for the Central Universities of China (Grant No. PA2021 KCPY0055).

The authors declare no conflict of interest.

All relevant data are included in the paper or its Supplementary Information.