Remediation of zinc-contaminated groundwater by iron oxide in situ adsorption barriers - From lab to the field.

Heavy metals such as zinc cannot be degraded by microorganisms and form long contaminant plumes in groundwater. Conventional methods for remediating heavy metal-contaminated sites are for example excavation and pump-and-treat, which is expensive and requires very long operation times. This induced interest in new technologies such as in situ adsorption barriers for immobilization of heavy metal contamination. In this study, we present steps and criteria from laboratory tests to field studies, which are necessary for a successful implementation of an in situ adsorption barrier for immobilizing zinc. Groundwater and sediment samples from a contaminated site were brought to the lab, where the adsorption of zinc to Goethite nanoparticles was studied in batch and in flow-through systems mimicking field conditions. The Goethite nanoparticles revealed an in situ adsorption capacity of approximately 23 mg Zn per g Goethite. Transport experiments in sediment columns indicated an expected radius of influence of at least 2.8 m for the injection of Goethite nanoparticles. These findings were validated in a pilot-scale field study, where an in situ adsorption barrier of ca. 11 m × 6 m × 4 m was implemented in a zinc-contaminated aquifer. The injected nanoparticles were irreversibly deposited at the desired location within <24 h, and were not dislocated with the groundwater flow. Despite a constantly increasing inflow of zinc to the barrier and the short contact time between Goethite and zinc in the barrier, the dissolved zinc was effectively immobilized for ca. 90 days. Then, the zinc concentrations increased slowly downstream of the barrier, but the barrier still retained most of the zinc from the inflowing groundwater. The study demonstrated the applicability of Goethite nanoparticles to immobilize heavy metals in situ and highlights the criteria for upscaling laboratory-based determinants to field-scale.

Introduction

Heavy metals are the most frequent point source pollutants in the western world accounting for about 30% of all contaminated sites in Europe (Panagos et al., 2013). Since heavy metals cannot be degraded by microorganisms, they constitute persistent hazards to groundwater resources and are frequently transported over long distances as much as several kilometres endangering possible recipients, e.g. drinking water production. However, it is very difficult to sustainably remove heavy metals from contaminated aquifers. The most frequent applied techniques are excavation with subsequent deposition or pump-and-treat (Brusseau, 2019). Both technologies have advantages and disadvantages, but –in sum– they are very expensive and can rarely remove all heavy metals to the desired extent even despite extremely long pumping times up to 20 years or longer. Consequently, many sites are simply not or only insufficiently remediated due to the lack of appropriate and cost-effective remediation technologies and there is a need for new, affordable remediation technologies preventing the spreading of heavy metals.

In recent years, in situ technologies became more and more attractive for groundwater remediation since they promise to be relatively cost-effective and easier to apply than the abovementioned approaches. In the case of heavy metals such as zinc, belowground immobilization is considered as an appropriate remediation measure to stop further spreading. The mobility of zinc in aquifers is determined mostly by pH, but also by the clay content, phosphorus availability, organic matter content and redox conditions (Kumpiene et al., 2008; Meers et al., 2006). In acidic water, the transport is controlled by cation exchange, while chemisorption and the formation of insoluble hydroxides determine the transport of zinc under alkaline conditions (Kabata-Pendias, 2010). Since zinc precipitates at alkaline pH as Zn(OH)2 it has been attempted to remediate zinc contamination by raising the pH. However, maintaining the increased pH of the groundwater over prolonged time periods is not practical (Seibert et al., 2016). It is also well known that heavy metal ions adsorb to iron (oxyhydr)oxides such as Goethite or ferrihydrite (Jenne, 1968). Due to the omnipresent iron (oxyhydr)oxides content in sediments, heavy metals are thus naturally retarded in aquifers but the limited adsorption capacities are mostly exceeded by the high amounts of pollutants that are released upon spills, leading to continuous spreading of plumes. Goethite is advantageous over other adsorbents such as activated carbon since it is an abundant mineral in soils and sediments. In addition, standardized ecotoxicological tests revealed that the exposure to Goethite does not harm living organisms to a large extent (González-Andrés et al., 2017; Höss et al., 2015b; Mollenkopf et al., 2021). It thus appears as an obvious solution to inject synthetic iron (oxyhydr)oxides into an aquifer to produce adsorption barriers for heavy metals. However, regular iron (oxyhydr)oxides produce slurries of aggregates >1 μm, which can hardly be injected into a porous matrix. Such large aggregates would immediately clog the pores preventing their spreading through the sediment to achieve a reasonable radius of influence around the injection well (>2 m).

Here, we demonstrate the eligibility of humic acid-coated Goethite nanoparticles for the immobilization of heavy metals in groundwater from lab- to field-scale. Such particles must fulfil several criteria in order to apply for site remediation: the injected iron (oxyhydr)oxide nanoparticles must be colloidal, finitely mobile in sediment, meta-stable, adsorptive, and non-hazardous to the environment.

Colloidal state and stability against aggregation

The nanoparticles must be perfectly colloidal during production, transport to the site and injection, which can last several weeks altogether. If the colloidal stability of the nanoparticles is insufficient, they form large aggregates, which reduces their mobility in the aquifer and their surface area for adsorption. Once the nanoparticles are injected and reached the desired radius of influence (ROI), they must precipitate and stay immobile at the intended location. They should not migrate further with the groundwater flow. This property must be adjusted in a way that precipitation occurs with a time delay and does not negatively affect the mobility of the nanoparticles in the aquifer during injection. The particles must remain colloidal during the injection time (several hours) to maintain their mobility, but start to adhere to the sediment matrix only shortly after the injection due to aggregation, gravity, or colloid-surface interaction with the sediment (Molnar et al., 2015).

Mobility

While traveling through the aquifer, the nanoparticles interact with the sediment and may be retained due to physico-chemical interactions (Tosco et al., 2014). For in situ remediation purposes, these interactions must be minimized and the nanoparticles must be able to travel several meters within the aquifer from the injection point. Otherwise, the radius of influence (ROI) is too short and the injected nanoparticles accumulate near the borehole.

Adsorption

The nanoparticles deposited in the aquifer must adsorb the heavy metal contaminants. It is also important that the settled/deposited nanoparticles form only a thin layer covering the sediment, which does not clog the pore space allowing the contaminated water to flow through the adsorption zone (permeable barrier). In this way, a highly adsorptive surface area will be generated on the surface of sand grains for eliminating heavy metal contaminations from water.

Environmental-friendly

Finally, the particles must be non-toxic to prevent any harm to the environment or to the potential users of the groundwater.

We synthesized Goethite nanoparticles and tested the adsorption properties for zinc as heavy metal representative. Previous studies have investigated the colloidal stability (Tiraferri et al., 2017), mobility (Bianco et al., 2017; Velimirovic et al., 2020), and ecotoxicity (Cabellos et al., 2018; González-Andrés et al., 2017) of Goethite nanoparticles and have identified their potential and applicability for groundwater remediation in accordance with the above-mentioned criteria. The current study presents the full route, from laboratory testing to field implementation of in situ nanoparticle-based remediation for a contaminated aquifer according to the abovementioned criteria. We evaluated the metastability and adsorption isotherms of the nanoparticles for immobilization of zinc from groundwater. Laboratory batch and column experiments were carried out in order to evaluate the performance of the Goethite nanoparticles with field materials. The results were used to design and implement a pilot injection of an iron oxide adsorption barrier for attenuating a zinc contamination plume at an industrial site.

Materials and methods

Bulk Goethite and Goethite nanoparticles

Bulk non-colloidal Goethite powder (30–63% iron) was purchased from Sigma Aldrich. Goethite nanoparticles were synthesized according to the patent EP2334608B1/US8921091B2 (Meckenstock and Bosch, 2014). Briefly, humic acids were added to stabilize Goethite crystals against aggregation. The suspension contained ca. 70 ± 7 g Fe L−1 and consisted of Goethite as exclusive iron (oxyhydr)oxide phase, and were exposed to atmosphere in various stages of the experiments. The hydrodynamic diameter of 600–800 nm and the Zeta potential of −37 ± 2 mV was determined with the particle-sizing system Nicomp 380 ZLS (Entegris, Billerica, USA). Point of zero charge (PZC) of these Goethite nanoparticles has been reported to be <3 (Tiraferri et al., 2017), which is lower than reported PZC of bulk Goethite, which is at pH ~8.0–8.5 (Cristiano et al., 2011; Kosmulski, 2016). For more information on the characteristics of the iron oxide nanoparticle used in this study, see Montalvo and Smolders (2019); Montalvo et al. (2018); Velimirovic et al. (2020).

In all experiments, iron was determined by the ferrozine assay according to Braunschweig et al. (2012) at a wavelength of 560 nm. Zinc was measured according to the protocol of Ghasemi et al. (2003), using the azo dye Zincon (2-carboxy-2′-hydroxy-5′-sulfoformazyl benzene) at a wavelength of 620 nm. Both were detected using a Tecan Infinite M200 photometer (Tecan, Männedorf, Switzerland). Element concentrations in the groundwater samples from the field were analysed with ICP-MS (X-Series II; ThermoFisher Scientific) (Mohammadian et al., 2021).

Batch adsorption experiments and determination of adsorption isotherms

Adsorption isotherms were determined in MilliQ-purified water (MilliQ, 18.2 μS/cm, Millipore, Eschborn) and in synthetic groundwater at pH values of 6.5 and 7.0 with Goethite nanoparticles and bulk non-colloidal Goethite. The pH values were buffered with 10 mM MES buffer (Sigma, MES hydrate, minimum 99.5% titration). The synthetic groundwater contained: K2SO4: 11.7 mg L−1, CaCl2: 343.3 mg L−1, MgSO4: 166.6 mg L−1, NaSO4: 89.1 mg L−1, KNO3: 12.7 mg L−1, Na2CO3: 151.4 mg L−1, NaH2PO4 · H2O: 200.8 mg L−1. Isotherms were determined in 2 L Erlenmeyer-flasks containing 1.25 L MilliQ-water and Goethite nanoparticles or bulk Goethite (dry powder, Sigma-Aldrich) at an end concentration of 1.78 g L−1 Goethite. A zinc chloride solution was added to final concentrations of 100, 50, 20, 10, 5, 1, 0.1 and 0.01 mg L−1 and isotherms were determined in duplicates per zinc concentration and two controls without Goethite addition. The flasks were incubated at room temperature and constantly agitated at 150 rpm on an open orbital shaker for 15 days (artificial groundwater incubations) or 36 days (MilliQ-water incubations) to reach equilibrium. For sampling, 1900 μL of the solution were sampled and then centrifuged for 10 min at 10,000 relative centrifugal force (RCF). Subsequently, 500 μL of the supernatant were carefully sampled without stirring up and then transferred to 15 mL Falcon tubes for preparation for ICP-MS analysis. Adsorption rates were studied by analysing samples taken from the groundwater samples after 1, 3, 6, 10, 15, 20, and 30 days after incubation. In separate experiments, the influence of pH on adsorption of zinc was studied via further batch experiments in triplicates at pH-values between 4 and 8. For these experiments, 200 mL beakers were filled with 160 mL of a suspension containing 1.78 g L−1 Goethite, 5 mg zinc L−1 solution and 10 mM MOPS (3-(N-morpholino)propanesulfonic acid) buffer in MQ water. The suspensions were treated in the same way afterwards as before.

Adsorption tests in sand columns

In order to investigate the adsorption capacity of Goethite nanoparticles in presence of sand grains, adsorption tests were carried out in columns (Poly methyl methacrylate (PMMA), 3.2 cm diameter x 7.0 cm length). The columns were filled with sand obtained from the site (see below) and flushed with artificial groundwater from bottom to top. The Goethite nanoparticle suspension was diluted in H2O (final concentration: 10 g Goethite L−1) and one pore volume of Goethite nanoparticles (ca. 409 mg Goethite) suspension was injected into the column using a peristaltic pump (Ismatec ICP8, Cole-Parmer, Wertheim). Then, the pump was turned off to allow the Goethite nanoparticles to cover the sediment surface. After 24 h, artificial groundwater containing 10 mg zinc L−1 was injected into the column at a flowrate of 0.5 mL min−1 equivalent to a seepage velocity of 2.1 m day−1. The effluent was collected with a fraction collector (Omnicoll, Lambda, Brno, CZ) and the Zn and Fe concentrations were determined using the methods explained before.

Metastability

Colloidal metastability and the sedimentation time of the Goethite nanoparticles were studied by addition of calcium (0 mM, 5 mM, 10 mM, 20 mM, and 30 mM) to Goethite nanoparticles suspensions. A 50 mM calcium standard solution was prepared from calcium chloride dihydrate (CaCl2 x 2 H2O, Carl Roth GmbH & Co. KG, Karlsruhe). Different volumes of the calcium standard solution were added to aliquots of 0.4 mL Goethite nanoparticles, and the mixtures were diluted to 4 mL in MiliQ Water (final Goethite nanoparticle concentration: ca. 10 g L−1). During the period of 24 h, aliquots (0.01 mL) from the supernatant were collected. The aliquots were transferred into 0.99 mL 6 M HCl (1:100 dilution). Afterwards, the iron concentration was determined with the ferrozine assay. The half-life time was defined as the time during which the concentration of still dispersed Goethite nanoparticles in the supernatant was reduced to 50% of the original value.

Transport experiments

Transport tests were carried out in a similar PMMA columns filled with sand taken from well bores of the site. The columns were wet-packed with the sand and for each experiment the average porosity was calculated based on the bulk density. An average dispersivity of 0.27 ± 0.03 cm was estimated with tracer tests prior to each experiment.

The transport tests were performed at four flow rates of 0.5, 1.0, 4.0, and 8.0 mL min−1, equivalent to Darcy velocities of ca. 0.06, 0.13, 0.50 and 1.0 cm min−1. The Goethite nanoparticles suspension (10 g L−1 in 10 mM CaCl2) was injected into the column with a peristaltic pump. Each experiment was carried out in duplicates, and consisted of (i) injection of at least 5 pore volumes (PV) of groundwater, (ii) pre-conditioning with 1 PV of tap water in order to reduce the direct contact between high saline groundwater and the Goethite nanoparticles suspension, (iii) injection of 3 PV of Goethite nanoparticles, and (iv) post-flushing with 2 PV of tap water. The effluent was collected in 12 mL vials and iron concentration was measured with the Ferrozine assay.

Modeling and prediction of Goethite mobility

In order to transfer the results of 1D transport experiments in columns (where the flow velocity is constant) to the radial geometry under field conditions (where the velocity decreases hyperbolically with distance), we used the methodology proposed by Tosco et al. (2018). Briefly, the transport of Goethite nanoparticles in linear and radial geometries can be described using a modified advection-dispersion equation, which takes into account the mechanisms for retention of Goethite nanoparticles. In this work, two retention mechanisms were considered: (i) a linear reversible deposition (site i = 1) and (ii) a reversible blocking mechanism (site i = 2). Under linear 1D conditions, the Goethite nanoparticle transport can be modelled by the following differential equation (Tiraferri et al., 2011):where ε is the porosity of the medium [−], q is the Darcy velocity [L T−1], c is the colloid concentration in the mobile phase [M L−3], s is the colloid concentration in the solid phase [−], D the dispersion coefficient [L2 T−1], ρ is the bulk density of the solid matrix [M L−3], k and k are respectively the attachment and detachment coefficients [T−1] and Smax is maximum particle concentration for the site with blocking. Under radial geometry the above equations above become:

Attachment and detachment coefficients (kai and kdi) change with velocity as described by Tosco et al. (2014):where v is the effective velocity, defined as the Darcy velocity divided by the effective porous medium porosity, μ is the suspension viscosity, and η0 is the single collector contact efficiency, here calculated using the formulation given by Yao et al. [54]. Cai and Cdi are empirical coefficient that describe the velocity-dependent kinetics of kai and kdi for adsorption site i.

For prediction of field-scale ROI, the experimental breakthrough results of the transport experiments were modelled using Eqs. (1), (2), (3) and fitted using the software MNMs (http://areeweb.polito.it/ricerca/groundwater/software/MNMs.php). From inverse data fitting, kinetic parameters including attachment and detachment coefficients (kai and kdi) and maximum blocking concentration Smax for each velocity were estimated, and were used to determine the velocity-dependent kinetic coefficients (Ca and Cd, Eqs. (7), (8)). Forward modeling of MNMs based on Eqs. (4), (5), (6) was then used to predict the Goethite nanoparticles distribution around injection wells and hence to estimate the effective radius of influence (ROI). In this study, a ROI was defined as the radius where the Goethite nanoparticles concentration was higher than >7 g L−1. The operating conditions (e.g., injection flowrate and concentration) were adopted to the real pilot-scale field study (see below).

Pilot-scale field application

We worked on an industrial site near Cologne, Germany, with a zinc contamination in the groundwater, which has been monitored for more than 15 years. Over the years, a long plume with an average Zn concentration of approximately 10 mg L−1 developed in the groundwater, despite sealing of the source zone against rain water seepage. The sediment of the aquifer is dominated by middle sand with minor amounts of fine sand with an average hydraulic conductivity of 1.06 × 10−3 m s−1 and a porosity of 15–20%. The groundwater velocity in the test area was estimated to be ca. 0.73 m day−1. The zinc concentrations in the planned range of the in situ barrier were between 5 and 20 mg L−1. Previous monitoring, revealed that the O2 concentrations were higher than 1.7 mg L−1, with measured the redox potential >165 mV in all the wells. Active operation of the industrial site and the depth of the aquifer (groundwater level approx. 14 m below ground level) prevented an excavation for economic reasons, so far.

The pilot study consisted of two injection wells, one upstream, and one downstream monitoring well (Fig. S1, Supplementary Materials). All wells were located in the estimated center of the plume. The distance between the injection wells was 5 m, allowing an overlap considering a planned radius of influence of >2.5 m for each well.

Concentrated Goethite nanoparticles suspensions (ca. 100 g L−1) were delivered to the site in six IBCs (intermediate bulk containers; 1 m3). A total of ca. 21 m3 of diluted Goethite nanoparticles (measured concentrations 9.5 g L−1, diluted in the municipal water) were injected into the two wells at a flowrate of ca. 30 L min−1, each. Prior to injection, ca. 500 L of the local municipal water was injected into each well to reduce the direct contact between the groundwater and the Goethite nanoparticles. The injection was carried out simultaneously in both wells using packers mounted above the screen, which was located directly below the groundwater table. A pressure transducer was mounted above the packer inside the tubing in order to record the pressure inside the wells. Pressures and flowrates were also monitored continuously at the surface to detect possible blocking, which, however, did not occur. A level logger was placed into the downstream well to estimate the rise of water table in IF01 and IF02 during injection, and the response behavior of the aquifer (Fig. S2, Supplementary Materials).

Analysis of groundwater samples

Groundwater samples were taken from all four wells during 11 sampling campaigns over the course of approximately 9 months. Three sampling campaigns were conducted before the injection of HA-Goe. The groundwater was analysed for zinc and other metals with ICP-MS (X-Series II, ThermoFisher Scientific) in unfiltered and filtered samples (0.22 μm; polyethersulfone, Restek). Prior to analysis, the samples were treated with HCl (37%, AnalaR NORMAPUR; VWR Prolabo) and H2O2 (30%, Rotipuran; Carl Roth GmbH + Co KG) to completely dissolve potentially occurring Goethite particles. pH (Sentix 41; WTW) and electric conductivity (TetraCon 325; WTW, Weilheim) were determined in untreated groundwater samples.

Results

Adsorption isotherms for bulk Goethite and Goethite nanoparticles

Adsorption isotherms of Zn were performed with Goethite nanoparticles and bulk Goethite to define the adsorption capacities at pH 6.5 and 7.0. The data were fitted with Freundlich, Hill, and Langmuir isotherms, while the Langmuir equation revealed the best fit (others are not shown here). The Langmuir isotherm is indicated by:where Γ is the amount of Zn adsorbed per gram of Goethite (mg g−1), Ceq is the equilibrium concentration of the Zn in solution (mg L−1), KL is the Langmuir constant (L mg−1) that relates to the affinity of binding sites, and Γmax is the maximum adsorption capacity (mg g−1).

As expected, both maximum adsorption capacity and affinity constants are higher at pH = 7.0 compared to pH = 6.5 (Fig. S3, Table 1). The maximum adsorption capacities are slightly lower in groundwater compared to those in ultrapure water, most likely due to the presence of ions competing with Zn for adsorption sites. The bulk non-colloidal Goethite adsorbed significantly less Zn (Γmax) exhibiting lower affinities for Zn (KL) even in ultrapure water, indicating the superior adsorption properties of the Goethite nanoparticles compared to bulk material (Table 1).

Table 1

Adsorption properties of Zn to Goethite nanoparticles determined by fitting the experimental data from batch incubations with Langmuir isotherms after equilibrium was reached (36 days). MQ water stands for ultra pure water (Millipore, Eschborn) with a conductivity below 18.2 μS/cm.

Goethite	pH	Solution mixture	Γmax (mg Zn g−1 Goethite)	KL (L mg−1)
Nanoparticles	6.5	MQ Water	41.53 ± 2.63	0.25 ± 0.01
	6.5	Groundwater	32.00 ± 0.11	0.25 ± 0.01
	7.0	MQ Water	53.44 ± 0.88	0.73 ± 0.03
	7.0	Groundwater	48.89 ± 1.13	0.77 ± 0.03
Bulk Goethite	7.0	MQ Water	6.36 ± 0.48	0.31 ± 0.12

For both bulk Goethite and Goethite nanoparticles, higher adsorption was observed at higher pH-values (Fig. 1a). At pH 5 and below, less than 50% Zn was adsorbed to Goethite compared to pH = 7. This pH dependency was more pronounced for Goethite nanoparticles compared to bulk Goethite, but the nanoparticles also adsorbed more Zn than bulk Goethite at all tested pH-values. Furthermore, the adsorption kinetics of Zn to the Goethite nanoparticles was faster than for bulk Goethite. During the first day of incubation ca. 60% of Zn adsorbed to the Goethite nanoparticles and near-complete removal of Zn took over 15 days, while bulk Goethite adsorbed less than 40% in more than 6 days (Fig. 1b).

Fig. 1

pH dependency (a) and kinetics (b) of zinc adsorption to Goethite nanoparticles (closed diamonds) and bulk Goethite (closed squares) at pH 7, with fitted pseudo-first (PFO) and pseudo-second order (PSO) rate kinetics laws for the Goethite nanoparticles (see the supplementary materials). Symbols and error bars show arithmetic means and standard deviations of three replicate incubations.

Adsorption tests in column with sediment from the field

To mimic field conditions, adsorption tests were performed with sediment columns with or without amendment of Goethite nanoparticles (Fig. 2). The iron concentration in the effluent after addition of Goethite nanoparticles was found to be negligible, proving that the introduced nanoparticles were indeed irreversibly immobilized in the sediment. Without addition of Goethite nanoparticles, Zn was adsorbed for ca. 7 PVs to the sand before breakthrough appeared at the column outlet. This retardation was probably due to the natural iron (oxyhydr)oxide content of the sediment (XRF analysis: 26.6 g Fe2O3 kg−1 Sand). In the Goethite-loaded column (added Goethite ca. 2 g Goethite kg−1 Sand), almost complete Zn removal was observed for ca. 29 pore volumes, four times more than the natural adsorption capacity of the sand. For both the control and the Goethite-loaded column, the slow increase of Zn concentrations at the outlet continued until a sudden sharp breakthrough appeared when the maximum capacity of the adsorption was reached. Based on the injected zinc concentration of 10 mg L−1 and the applied influx, we calculated the in situ adsorption capacity of Goethite nanoparticles as 23.35 mg Zn per g Goethite. The natural adsorption capacity of the sand was calculated as 0.02 mg Zn per g sand.

Fig. 2

Breakthrough curves for Zn at pore velocity of 2.1 m day−1 with and without addition of Goethite nanoparticles. Inlet concentration was 10 mg Zn L −1.

Metastability and half life time of the precipitation

In order to induce aggregation of the Goethite nanoparticles, we tested different concentrations of calcium amendment to reduce the colloidal stability of the nanoparticles. Mere dilution of the Goethite nanoparticles to a concentration of 10 g Goethite L−1 induced their sedimentation within 24 h whereas the undiluted nanoparticles remained stable for periods of several weeks. Increased calcium concentrations were found to reduce the stability of the Goethite nanoparticles. Almost immediate sedimentation was observed by adding 30 mM calcium to 10 g L−1 Goethite nanoparticles (Fig. 3). The aggregation half life time (the time during which 50% of the particles moved by gravitation to the bottom of the reaction vessel) for 10 g L−1 Goethite nanoparticles was found to be less than 2 h in presence of 20 mM and 30 mM calcium, and ca. 6 h for 10 mM calcium.

Fig. 3

Aggregation in 10 g L−1 Goethite nanoparticle suspensions illustrated as relative Fe concentrations remaining dispersed (non-settled) 24 h after addition of calcium chloride.

Transport experiments and 1D-modeling

Determination of kinetic parameters governing transport of Goethite nanoparticles

Fig. 4 shows the breakthrough curves of Goethite nanoparticles reported as normalized effluent concentration Cout/Cin as a function of the number of pore volumes injected for Goethite nanoparticles suspensions for different injection velocities. A very good fit between the experimental data and the mathematical model was observed for all four tested flow velocities. At injection velocities lower than 0.13 cm min−1, the deposition rate of the Goethite nanoparticles decreased with time (increasing trend of Cout/Cin during the nanoparticle injection), which indicates the typical behavior of a blocking mechanism due to saturation of sediment surface with Goethite nanoparticles (Mondal et al., 2021). As expected, the mobility of the Goethite nanoparticles increased with injection velocity, but the high recovery (i.e. Cout/Cin > 0.85) in all experiments indicated that the nanoparticles were highly mobile in the sediment even at lower injection velocities.

Fig. 4

Experimental breakthrough curves of Goethite nanoparticles in sediment columns and mathematical fits modelled with the program MNMs based on two-site reversible attachment and blocking.

By fitting the breakthrough curves to the modified advection-dispersion equation (Eqs. (1), (2), (3)) using MNMs, the kinetic parameters describing the transport of the Goethite nanoparticles were derived (Table 2). No dependency of Smax and ka2 on velocity was found. On the other hand, ka1, kd2, and kd1 were found to change with velocity, with the empirical velocity-dependent coefficients of Ca1 = 0.015, Cd1 = 9.658E+3, and Cd2 = 1.031E+3 (Eqs. (7), (8)). These coefficients were used to simulate the injection and distribution of the Goethite nanoparticles at larger scales in radial geometry around an injection point.

Table 2

Kinetic parameters obtained through the fitting of the column transport test at different velocities.

Darcy velocity (cm min−1)	ka1 (s−1)	kd1 (s−1)	Smax (g Goethite nanoparticles/g sand)	ka2 (s−1)	kd2 (s−1)
0.06	1.15E-05 (±0.15E-5)	1.08E-04 (±0.12E-4)	1.00E-02	1.17E-04 (±0.16E-4)	1.07E-04 (±0.12E-4)
0.13	1.04E-05 (±0.04E-5)	2.12E-04 (±0.23E-4)	1.00E-02	2.05E-04 (±0.20E-4)	2.11E-04 (±0.140E-4)
0.05	3.37E-04 (±0.47E-4)	3.55E-03 (±1.55E-3)	1.00E-02	1.00E-05 (±0.00E-5)	1.00E-03 (±0.00E-3)
1	4.30E-04 (±0.23E-4)	8.70E-03 (±0.15E-4)	1.00E-02	1.00E-05 (±0.00E-5)	1.00E-03 (±0.00E-3)

Pilot injection modeling and determination of ROI

A predictive simulation of injecting Goethite nanoparticle suspension into a well in radial geometry (mimicking field conditions) was carried out using Eqs. (4), (5), (6) and the kinetic parameters obtained from column tests. The operating parameters were adopted to the field study (2.6, 3.5). The following consecutive injection steps were modelled: i) injection of 500 L of Goethite-free water at a flowrate of 30 L min−1, ii) injection of Goethite nanoparticles at the same flowrate for 9.5 h, and iii) post-flushing with 500 L of Goethite-free water at the same flowrate. The aquifer properties were set based on the field data with an effective porosity of 15% and a field-scale dispersivity of 25 cm. The simulation results are reported in Fig. 5, where the profile obtained from simulation of an inert tracer is also shown for comparison.

Fig. 5

Expected concentration profiles of Goethite nanoparticles (NP) and of a conservative tracer as a function of the radial distance from the injection well, simulated using Eqs. (4), (5), (6).

The above figure shows that transport of the injected Goethite nanoparticles is similar to that of the tracer, and in both cases an effective radius of influence (i.e. concentration > 7 g L−1) of at least 2.75 m is reached. This allows for an overlap of 50 cm between the ROIs of the two injection points (5 m apart), ensuring that a fully continuous barrier can be implemented. The increased iron oxide concentrations in radius 1–2 m is most likely because of re-mobilizing the accumulated nanoparticles in the close vicinity of the wellbore due to post-flushing.

Implementation of a permeable in situ adsorption barrier in the field with Goethite nanoparticles

The pilot implementation of an in situ adsorption barrier based on Goethite nanoparticles was performed at an industrial site by injecting the nanoparticles into two wells (Fig. S1). According to prior stability tests, 10 mM calcium was added to the Goethite nanoparticles suspension (9.5 g L−1) to ensure the deposition of the particles within 24 h. The flowrate of the injection fluid in each well was kept constant at 30 ± 6 L min−1. The measured pressure in the injection wells showed a first peak, when particle-free water was injected for pre-conditioning during injection of Goethite nanoparticles. During the following injection of the Goethite nanoparticle suspension, the pressure remained constant at both wells and in the same range as observed during injection of water (Fig. 6). No indication of pore clogging (increase of pressure) or fracturing (sudden pressure drop) was observed. The injection was only interrupted for changing the containers with the Goethite nanoparticle stock suspension (drops in the pressure). During the injection (in 2 h intervals) and also on the day after the injection, samples were taken from the downstream well to visually check whether the injected Goethite nanoparticles migrated further than calculated but no indication of breakthrough was observed.

Fig. 6

Observed downhole pressures in the two injection wells IF01 and IF02.

The iron concentrations in unfiltered samples of all four wells were analysed during the monitoring period as an indicator of potentially mobile Goethite nanoparticles (Fig. 7). The total iron concentrations remained always in the range of the natural background concentrations indicating that the injected Goethite nanoparticles were stably deposited and were not re-mobilized with groundwater. Only 90 days after the injection, an increase of the iron concentrations was observed in the injection wells IF01 and IF02. However, the sampling protocols for this day indicated that the pumps were lowered deeper than usual into the well and most likely bottom sediment was suspended and pumped together with the groundwater. Metal concentrations measured at this sampling time also differed strongly from other days (Figs. S4-S8, Supplementary Materials), which indicates these data can be treated as outlier.

Fig. 7

Fe concentrations over time as indicator for mobile Goethite nanoparticles in the upstream and downstream monitoring wells as well as in the two injection wells IF01 and IF02.

The Zn concentrations in the upstream well increased constantly during the monitoring period, almost doubling the initial value of 4 mg L−1 to 8 mg L−1 (Fig. 8). The increase was especially strong in IF02 where the zinc concentrations reached 20 mg L−1 before injection of the barrier. Right after the injection at day 0, the Zn concentrations decreased drastically in the two injection wells and the downstream monitoring well to less than 0.5 mg L−1 for ca. 90 days despite continuously increasing inflow concentrations. Again, elevated values were determined at day 60, most likely because of lowering the pump too deep into the well and pumping of sludge from the well. From day 90, the Zn concentrations increased also in the injection wells of the barrier, starting from ca. 2.2 ± 0.6 mg L−1 at day 90 and increasing with almost at the same rate as the inflow concentrations of the upstream well.

Fig. 8

Observed Zn concentrations in pre- and post-injection field samples.

A total mass of 400 kg of Goethite nanoparticles was injected into the two injection wells, which, according to the findings of the column tests (23.35 mg Zn (g Goethite)−1), is expected to adsorb ~9340 g Zn.

Considering the groundwater flow velocity of 0.7 m day−1, the porosity of 0.15, the barrier area of 11 × 4 m2, and the background Zn concentration of 10 mg L−1, we could calculate the flux of Zn and the absolute removal by the in situ barrier. The in situ barrier received almost constantly 46.2 g Zn day−1and was therefore expected to immobilize the Zn up to 202 days. However, the breakthrough of Zn happened ca. 90 days after injection and continuously increased over the monitoring period of 188 days (Fig. 8), which indicated that additional processes influenced the Zn adsorption of the installed barrier.

Discussion

Heavy metal pollution is one of the biggest problems in groundwater remediation. Taking zinc as a model compound, we demonstrate here the development of in situ adsorption barriers based on iron (oxyhydr)oxides. Since iron oxides such as Goethite are natural minerals present in aquifers they are not of environmental concern for injection into groundwater.

Several studies indicated the great potential of ferric iron nanoparticles as adsorbents for heavy metals (Braunschweig et al., 2013; Xu et al., 2012). Specifically, retention of zinc on Goethite around neutral pH values has been reported by several authors. For example, Gräfe and Sparks (2005) found that large amounts of Zn2+ can be adsorbed to Goethite in form of outer-sphere complexes. Others reported that the adsorption of zinc to Goethite as an endothermic chemical type of reaction resulting in the formation of inner-sphere complexes (chemisorption) (Nachtegaal and Sparks, 2004; Shi et al., 2021; Trivedi and Axe, 2001; Trivedi et al., 2001). In the present study, both bulk Goethite and Goethite nanoparticles adsorbed zinc around neutral pH, where the Goethite nanoparticles showed 6–9 times higher adsorption of zinc compared to bulk Goethite. Our data are supported by Rahimi et al. (2015), who showed that nanoparticles exhibit higher adsorption capacities than larger particles and bulk materials. At pH 6.5–7.0 zinc is largely present as dissolved species rather than as insoluble Zn(OH)2 indicating a real sorption process. This is supported by the significant concentrations of dissolved Zn (several mg L−1) in both milliQ water and groundwater in the adsorption experiments. The presence of humic acid in the Goethite nanoparticles suspension, and especially their negatively charged –COO−-moieties at the pH-conditions prevalent in our study, could also enhance the electrostatic attraction of cations such as zinc. This, coupled with increased reactive surface area due to their minute size lead to increased adsorption capacity of the Goethite nanoparticles. Nevertheless, sorption of zinc was also observed to bulk Goethite at neutral pH, where no humic acid was present. Hence, the adsorption of zinc to Goethite is clearly not dependent on humic acids for electrostatic interaction.

The maximum adsorption capacity of our nanoparticles of 53 mg zinc per g Goethite determined in Langmuir isotherms is quite remarkable and allows for calculating the amount of Goethite nanoparticles that needs to be injected into the aquifer. The adsorption of Zn to the Goethite nanoparticles was strongly pH-dependent (Fig. 1a) reaching its maximum capacity of 50 mg g−1 at pH 7.0, whereas the maximum capacity for bulk Goethite was reached at a higher pH of 8.0. This indicates that the Goethite nanoparticles can expose their full adsorption capacity at typical groundwater pH conditions. Even at pH 6.0, the adsorption capacity was still approximately 75% of the maximum, extending the range of possible applications into acidic conditions. The effect of pH on adsorption was also observed in the field studies: with decreasing pH more mobile zinc was observed in groundwater samples (see below).

As stated, an iron oxide must fulfil certain key criteria for producing in situ adsorption barriers for contaminated groundwater. A first criterion for injectable adsorption barriers is the metastability of the colloidal suspension of Goethite nanoparticles. To be injectable at all, the iron oxides must be perfectly colloidal and stable over several months because they must not settle during production, transport to the site, and injection. Once the particles coagulate to larger aggregates, they will clog the pores of the sediment during injection, which prevents the desired spreading over several meters in the aquifer. Coagulation must occur only shortly (max. 24 h) after the injection is finished in order to ensure that the Goethite is stably deposited with no danger of dislocating from the barrier. Overcoming this apparent dilemma (mobility during injection but deposition shortly after) is only possible via fine-tuning the kinetics of particle aggregation inside the aquifer via controlling the metastability. We therefore developed Goethite nanoparticles that were totally stable over several months during the production and transport, and mobile during injection, but they deposited shortly after the injection, covering the sediment with an iron oxide layer. This colloidal behavior and the mobility of Goethite nanoparticles have been studied in detail earlier (Tiraferri et al., 2017) and showed a strong dependency on the salinity of the solution. Especially divalent cations such as calcium and magnesium lead to a destabilisation of the particles (Illés and Tombácz, 2006). However, this allowed us to produce a metastable solution leading to deposition of the particles shortly after the injection into both sediment columns as model systems or the aquifer during the pilot. This controlled, time-dependent destabilisation is essential for the in situ barrier since the particles must stay at the site of injection and may not be transported with the groundwater flow. The amount of destabilizing agent was determined in the laboratory based on the composition of both the suspension and the aquifer constituents to ensure irreversible deposition of the Goethite nanoparticles within 24 h. This method is advantageous to other proposed methods such as by Bianco et al. (2017), where alternative pulses of the destabilizing agent and the nanoparticles with different retardation factors are injected. In the method of Bianco et al. (2017), local heterogeneity in the aquifer might result in uncontrolled mixing of the destabilizer and the nanoparticles and, hence, lead to early aggregation and deposition. Furthermore, the mixing and forced sedimentation only occurs in the boundaries of the pulses and the rest of the nanoparticles may stay mobile. Furthermore, the present study showed that addition of calcium did not negatively affect the quality of the Goethite nanoparticles, i.e. did not reduce the particle mobility by early aggregation. The constant pressure of injection indicated that no pore-clogging occurred during the injection. Similar to previous studies by Velimirovic et al. (2020) and Mohammadian et al. (2021), these Goethite nanoparticles can create a stable, homogenous and permeable adsorptive barrier with a large radius of influence of 3 m or more. We could also not observe a nanoparticle-facilitated co-transport of zinc, which would be a major concern for groundwater applications.

Most importantly, the adsorbing material must fulfil its function under in situ conditions, i.e., the adsorption of heavy metals. Several authors emphasized the sorption of zinc to Goethite, most likely caused by formation of surface complexes (Bolland et al., 1977; Forbes et al., 1976; Tiller et al., 1984). (Grimme, 1968) reported the formation of surface complexes with Goethite's 16 hydroxyl-groups by exchange of protons with the divalent metal ions. This indicated the formation of inner-sphere complexes under acidic conditions, which was confirmed by (Schlegel et al., 1997) using extended X-ray absorption fine structure (EXAFS) methods. Zn was found to form distorted oxygen octahedral complexes that share edges and/or corners with Goethite surface Fe-octahedra.

Besides adsorption, diffusion of zinc into the Goethite crystals is important, which significantly decreases putative desorption (Barrow et al., 1989; Gerth and Brümmer, 1983). Gerth and Brümmer (1983) found that 14 days treatment of Goethite needles with 0.7 N HNO3 released only 61% of the sorbed zinc. This inner-sphere sorption probably dominates between pH 4.5 and 7. Below pH 4.5 the solubility of Zn2+ is too high while above pH 7, Zn(OH)2-complexes adsorb to the Goethite surface. The diffusion into the Goethite crystals is a very interesting process for the in situ adsorption barriers since it will decrease a potential dissolution and thus increase the sustainability of the barrier.

The excellent adsorption properties of the Goethite nanoparticles could also be shown in sediment columns. However, the adsorption capacity was lower compared to the batch incubations. The observations are in agreement with those of (Montalvo and Smolders, 2019; Montalvo et al., 2018), who also observed that the adsorption capacities in flow-through systems are lower than in batch systems. A likely explanation for this phenomenon is the contact time of the fluid passing through the column as compared to the reaction time of the adsorption process. The reaction time to yield 50% of the maximal adsorption of zinc to our Goethite nanoparticles was around one day, while the maximum adsorption was observed after 15 days. In contrast, the contact time between Zn-containing solution and Goethite nanoparticles in both the sediment columns and at the field site was much lower with only 1.2 h and ca. 8.5 days, respectively. This effect of contact time has to be considered for the design of in situ barriers, which dimensions have to be adopted to groundwater velocities and the consequent reaction times.

Another possible explanation for the lower adsorption in the columns and in the field is a reduced number of available adsorption sites on the surface of the Goethite nanoparticles. The Goethite nanoparticles aggregated and deposited on the sediment and, therefore, only a portion of their surface area was accessible to adsorb heavy metals because the other portion was in contact with other particles or with the sediment. To our knowledge, no quantification of the surface area of deposited nanoparticles has been reported.

After successful injection and implementation of the barrier at field-scale, we observed an instantaneous decrease of the zinc concentrations in all affected wells. During the first 3–5 days, this was certainly caused by the displacement of groundwater due to injection of ca. 42 m3 of Goethite nanoparticles-bearing suspension. However, the decrease of the zinc concentrations was maintained for at least 90 days, when almost no zinc was observed in either injection wells or the downstream well. The pH of the groundwater in the injection zone and downstream was increased to 9.1 during the injection, and decreased with time until the end of the monitoring period. This increase in pH is assumed to significantly enhance adsorption of zinc onto Goethite particles as reported previously (Mohammadian et al., 2021; Montalvo and Smolders, 2019; Montalvo et al., 2018). The amount of Goethite nanoparticles injected in the pilot was calculated for an active adsorption barrier over ca. 200 days, yet the breakthrough was observed to begin at around 90 days and continued to the end of monitoring period. Several reasons can be argued for the reduction of the expected life time of the barrier: first, the contact time of the flowing groundwater inside the barrier (ca. 8.5 days) was lower than the expected time for complete adsorption (Fig. 1, >15 days). Hence, the residence time required for reaching equilibrium was not reached leading to an earlier breakthrough. Therefore, the contact time should be taken into account when designing a full-scale implementation. The contact time can be increased, e.g., by inserting the adsorptive barrier in two or more overlapping rows and hence increasing barrier length. Second, we recorded a constantly increasing influx of zinc over time. The zinc concentration in the upstream well almost doubled during the monitoring period, and the pre-injection concentrations in IF02 were as high as 20 mg L−1. Indeed, the background concentration of all elements in IF02 were significantly higher from that of the upstream well (see supplementary materials, Figs. S5-S7), which suggests another inflow pathway into the barrier in addition to the groundwater from the upstream well. The increased concentrations, along with potential changes in groundwater flow velocity and direction, caused the increased zinc inflow into the barrier. Third, the early breakthrough could be caused by the presence of co-contaminants (e.g., other metals) in the groundwater, which compete for adsorption with zinc. Detailed analysis of groundwater samples showed that the non-treated groundwater contained, among other elements, copper (ca. 300 μg L−1), manganese (ca. 3500 μg L−1), and lead (ca. 1000 μg L−1), all of which were completely removed from the groundwater during the whole monitoring period. Competitive sorption of (divalent) cations to Goethite and hence reducing effective adsorption capacities have been reported in the literature (Goldberg et al., 2007; Kratochvil and Volesky, 2000). In columns filled with Goethite-coated sand, Montalvo and Smolders (2019) observed highest adsorption affinity for copper followed by zinc, nickel, and cadmium. Continuous inflow of higher affine cations (e.g., copper) will result in a reduced adsorption of cations with lower affinity for Goethite surface sites (e.g., zinc) and hence in an earlier breakthrough. This mechanism could be significant considering the decreasing pH of the groundwater inside the barrier from 9.1 to 6.5, which also reduces the affinity of zinc for Goethite adsorption and hence result in early breakthrough.

In order to examine the discrepancies between results of column experiments and the field data, transport of Zn was simulated using a modified advection-dispersion equation, which assumed Langmuir adsorption under flow conditions. The details are presented in the supplementary materials (S9). First, the results of the column experiments (Fig. 2) were simulated assuming both chemical equilibrium and non-equilibrium conditions to investigate the effect of contact time. As expected, the results showed that the kinetic model allowed a better match between the predicted breakthrough curves and the results of the column experiments (Fig. S10 and Table S11). This confirms the presence of non-equilibrium conditions and hence supports our hypotheses that the contact times were insufficient for complete Zn adsorption onto iron oxide nanoparticles.

Then, a simple one-dimensional linear model with the length of 10 m was built to simulate the field conditions (Fig. S12). The fitted kinetic parameters –obtained from the column tests– were applied to the first 6 m of the model (i.e. the adsorptive barrier), while no adsorption was assumed for the last 4 m. Breakthrough of Zn in the middle of the barrier (i.e. the injection point) and outside the barrier (downstream well) were compared to field data. The hypothesis of increased inlet Zn concentration was also examined by changing boundary condition at the inlet of the model from 10 to 12 and 15 mg Zn L−1. The simulation results were compared to the actual field data (Fig. 9), where simulations with inlet Zn concentration of 10 mg Zn L−1 matched the data from IF01, while an elevated Zn concentration to 12 mg Zn L−1 explained the data from IF02. Therefore, and in addition to the contact time, the increased inflow of zinc into the barrier resulted in an earlier breakthrough compared to the expected. On the other hand, it was observed that none of the models could predict the observed Zn concentrations in the downstream well. Note that the breakthrough time assuming inlet Zn concentrations of 10 and 12 mg Zn L−1 matches the predicted values based on the mass balance (202 days). However, the observed data in the downstream well clearly deviate from this prediction. We propose that changes in groundwater direction have caused a bypassing of the reactive barrier by Zn-containing groundwater towards the downstream well, because the installed barrier did not cover the whole cross section of the Zn plume. It is very likely that not all the groundwater that arrived at the downstream well had flown through the barrier, which explains the early breakthrough compared to the expected time of 200 days.

Fig. 9

Comparison between observed Zn concentrations at the site and the simulation results for the injection wells (left) and downstream well (right) based on kinetic parameters from the column data. Note that Zn concentrations at the inlet are 10, 12, and 15 mg L−1 in top, middle, and bottom row, respectively.

A major concern about in situ adsorption barriers is their long-term stability and their long-term capability to immobilize contaminants. The current study showed that Goethite adsorption barriers remained stable over long times as the injected Goethite nanoparticles remained bound to the aquifer sediment matrix. This is in agreement with a previous field study, where no re-mobilization or dissolution of deposited Goethite nanoparticles was observed even when pH dropped to 4 (Mohammadian et al., 2021). On the other hand, the immobilized zinc remained bound to the Goethite over the monitoring period, as the elevated zinc concentrations were due to increased inflow and saturation of adsorption sites. A recent study has proved that such humic acid-coated Goethite nanoparticles can retain their mineral properties and high adsorption affinity towards heavy metals over long times even when exposed to groundwater (Mollenkopf et al., 2021). The authors observed no release of zinc or other metals previously bound to the Goethite nanoparticles. A release of adsorbed zinc is only expected under highly acidic conditions or in presence of other metal ions that compete with Zn2+ for adsorption to HA-coated Goethite. The first scenario is very unlikely at this site, as the pH values remained permanently above 6.5 throughout the monitoring period. On the other hand, zinc possesses a high affinity with Goethite nanoparticles in single- and multi-element systems (Mollenkopf et al., 2021; Montalvo and Smolders, 2019). Montalvo and Smolders (2019). Only copper is reported to outcompete zinc for adsorption to Goethite nanoparticles. The copper concentrations in this study were > 30 fold smaller than zinc concentrations and hence the impact is expected to be small. However, in aquifers polluted with several heavy metals this competitive adsorption and possible release of adsorbed heavy metals must be taken into account.

Finally, in order to be applicable for large-scale in situ remediation applications, the particles must be non-toxic to prevent any harm to the environment or potential users of the groundwater. Goethite has a low toxicity to humans and ecosystems, and is a common pigment in food industry (Cooper et al., 2003). The toxicity of Goethite nanoparticles, similar to those used in this study, has been analysed for the nematode Caenorhabditis elegans. No synthetic Goethite nanoparticles were more toxic than colloids found naturally in the groundwater. Coated Goethite nanoparticles showed a low effect concentration of 129 ± 3.1 mg Fe L−1 after 96 h exposure. Toxicity decreased with increasing aggregate size, a common effect for NP-toxicity (Cabellos et al., 2018; González-Andrés et al., 2017; Höss et al., 2015a; Höss et al., 2015b).

Concluding remarks

The present study presented a full route from laboratory experimentation to pilot-scale field implementation of using Goethite nanoparticles for groundwater remediation. Important criteria for a successful field implementation, including colloidal stability of the nanoparticles, their mobility in aquifer materials and in situ adsorption potential were measured in laboratory, and considerations for upscaling to field implementation were taken into account. A field study was performed with injection of approximately 42 m3 of Goethite nanoparticle suspension into a contaminated aquifer to establish a permeable adsorptive barrier of ca. 11 m × 6 m for the in situ immobilization of zinc. Monitoring the performance of the barrier showed that the zinc contamination broke through sooner than expected, most likely due to elevated inflow concentrations of contaminants and too short contact time between the injected nanoparticles and dissolved Zn2+ to allow for equilibrium. The discrepancies between the expected and observed field data indicated that even though it is important to determine the site-specific sorption behavior using soil and groundwater from the respective field, using mere site-specific adsorption capacities from batch or column experiments might lead to overestimating the adsorption capacity and longevity of the adsorption barrier. This is especially important for non-linear and/or non-equilibrium adsorption systems, where a single parameter such as maximum adsorption capacity cannot explain the sorption behavior. Hence, such potential limitations should be taken into account when designing the remediation of metal-contaminated aquifers with in situ permeable barriers. For example, the contact time can be increased via increasing the length of the barrier in direction of groundwater flow, e.g. two or three injection rows.

CRediT authorship contribution statement

Beate Krok: Conceptualization, Synthesis of Goethite particles, Adsorption experiments, Design of field work, Writing – Original Draft, Writing – Review & Editing. Sadjad Mohammadian: Conceptualization, Methodology, Mobility experiments, Column experiments, Writing – Original Draft, Writing – Review & Editing. Hendrick M. Noll: Methodology, Adsorption Experiments, Design of field work, Writing – Review & Editing. Carina Surau: Methodology, Mobility Experiments, Writing – Review & Editing. Stefan Markwort: Methodology, Design of field work, Writing – Review & Editing.

Andreas Fritzsche: Methodology, chemical analysis, Writing – Review & Editing.

Rainer U. Meckenstock: Conceptualization, Methodology, Supervision, Writing – Original Draft, Writing – Review & Editing.

Declaration of competing interest

BK, SM, and RM own a company which sells colloidal particles for various applications including groundwater remediation.