Modular treatment of arsenic-laden brackish groundwater using solar-powered subsurface arsenic removal (SAR) and membrane capacitive deionization (MCDI) in Vietnam

To evaluate energy ef ﬁ cient concepts for the modular treatment of brackish water, pilot trials for groundwater desalination and arsenic (As) removal were carried out in the Mekong Delta, Vietnam. Groundwater here is affected by naturally occurring high iron (Fe 2 þ ) and As concentrations, while, in coastal regions, groundwater is additionally contaminated by high salinity mostly due to seawater intrusion. Desalination was conducted by membrane capacitive deionization (MCDI), which shows low speci ﬁ c energy consumption (SEC). Anoxic groundwater with As(III) and Fe 2 þ was treated using a pre-oxidation step called subsurface arsenic removal (SAR) with the main advantage that no As-laden waste is produced. The pilot plant was operated using a photovoltaic system (3 kW p ) and a small wind turbine (2 kW p ). The SEC of drinking water produced was 3.97 kWh/m 3 . Concentrations of 1,560 mg/L were lowered to 188 mg/L, while Fe 2 þ was reduced from 1.8 mg/L to the below detection limit and As from 2.3 to 0.18 μ g/L. The results show that SAR is a feasible remediation technique for Fe 2 þ and As removal in remote areas, and demonstrate the potential of MCDI for brackish water desalination coupled with renewable energies. However, improvements in energy demand of the MCDI module can still be achieved.


GRAPHICAL ABSTRACT INTRODUCTION
Currently, about 600 million people worldwide live in lowelevation coastal zones affected by progressive salinization (Wheeler ; Dasgupta et al. a). Groundwater in coastal line aquifers is often affected from seawater intrusion due to overextraction and rising sea levels, which will be aggravated by climate change in the future. The salinity level is subject to seasonal variations during dry and wet periods, and tidal channels are an important potential source of high salinity water (Brecht et

Goal and scope
This study was carried out within the German joint research project WaKap as part of one of three plants installed around the Mekong Delta for subsurface arsenic removal (SAR) and/or desalination of brackish water (WaKap ). This study focused on the implementation of a combined desalination and As remediation technology using membrane capacitive deionization (MCDI) and SAR, respectively.
Combining the two technologies aimed at a robust, sustainable and easy-to-use approach for high removal and water recovery efficiencies as well as low energy demand.
Additionally, an important goal was to evaluate the supply of the system with renewable power sources such as solar (photovoltaic, PV) and wind (wind turbine, WT) power. A sustainability assessment and evaluation of the system was carried out taking environmental and socio-economic aspects of the region into consideration (WaKap ; Fritz et al. ).

State of the art
Capacitive deionization (CDI) is a relatively new electrochemical desalination process with the general advantage of low specific energy consumption (SEC) since no high pressure pumps or heat sources are required. In this process, saline water flows between two porous electrodes made of activated carbon, on which a voltage is applied. In contrast, every membrane-based technology for removing As from water creates an As-concentrated volume as a co-product, and safe disposal methods need to be evaluated. Examples are reverse osmosis (RO) or nano-, ultra-and microfiltration (Shih ; Figoli et al. 

Pilot plant
The modular pilot plant for the desalination of brackish water including the subsurface removal of Fe 2þ , Mn 2þ and As was installed in June 2018 on a household rooftop in Trà Vinh city, in a coastal region in the Mekong Delta, Vietnam. The process scheme of the system consisting of an MCDI (CapDI, Voltea ® ) plant with upfront SAR (FERMA-NOX ® -Wasseraufbereitung type BV 104) for iron and As removal is shown in Figure 1. Two different product streams are produced from the modular system: As and Fe 2þ free treated water from SAR is stored in T1 (e.g. for washing purposes), while drinking water from the MCDI is stored in T2.
(1) SAR: A submersible pump (PENTAX ® 4ST 14-8) is used to pump the groundwater (water table at 2 m below the surface) through a sand filter (1354 FRP Vessel, sand 0.5-1 mm as filtration media) and to prevent particles from passing into a storage tank on the rooftop (T1, 12 m above the surface). The treatment process is divided into the aeration and infiltration cycles: firstly, water from T1 is aerated with an air injector using a booster pump (feed pump 1; GRUNDFOS ® SCALA2) and stored in the SAR tank (900 L). Then, the oxygenenriched water is infiltrated back into the aquifer by gravity (SAR backflow). Iron oxidation and As adsorption take place underground, and water with very low, nontoxic concentrations of As is obtained around the well. Subsequently, all extracted water from the well, which is stored in T1, can be used by the household or further treated with MCDI.
Aeration and infiltration cycles must be repeated constantly to maintain the oxygenated zone around the well.
When water is extracted from the well, oxidationadsorption takes place in the oxidation zone and the infiltrated oxygen is gradually consumed. In order to avoid leakage of As or any other contaminant, a daily water extraction limit V E is calculated as a design parameter based on the previous experience of the manufacturer and fixed for using SAR (Luong et al. , ; Cañas Kurz et al. ). With a daily operation of two infiltration cycles (each cycle V I ¼ 900 L), the maximum designed capacity of the SAR plant is V E ¼ 8.3 m 3 /d. This leads to an extraction/infiltration ratio at the pilot site of Q E ¼ V I /V E ¼ 0.22 m 3 / m 3 . Since the oxygenated water is always infiltrated back into the aquifer, another advantage of the process is that no water is lost and no waste stream is produced.
(2) MCDI desalination: Secondly, a home-made MCDI unit with a commercial electrode module (Voltea B.V ® Model C3) is used for the desalination of the groundwater and to obtain drinking water (see Figure 2). The information provided by the manufacturer is the total surface area of the electrodes at 3.7 m 2 as well as sulfonyl and quaternary ammonium groups as active functionality groups for the cation and anion exchange membranes, respectively. By using a diaphragm pump (feed pump 2; SHURflo ® 8000-243-155), SAR-treated water from T1 is passed through the MCDI module in alternate cycles for desalination (charge) and regeneration of the electrodes (discharge). The MCDI module is operated with constant current (CC) during both phases with reversed voltage in the discharge phase using an XP power supply (XP Power HDS 1500 PS12, with a maximum electrical current at the outlet   The diluate of the MCDI is stored in T2, which is connected to the water supply, whereas the concentrate effluent is discharged directly into the sewer. The pilot plant is operated by renewable energy (see is controlled by an on-grid inverter, and a 3 kW p solar PV system (SUNERGY, SUN72M, monocrystalline) is connected with a hybrid inverter (Goodwee, GW-3648) that includes two lithium batteries (Scared sun, 2,400 Wh 50 Ah) for a total battery storage capacity of 4.8 kWh. The pilot plant is supplied directly by PV power. Excess energy is fed into the regional power grid; if energy is required, for example at night, the plant is powered additionally by the battery storage. If the battery state of charge (SOC) is lower than 20%, the plant can be supplied through the grid. A separate net-metering system controls the in-going and out-going energy of the latter. Additional power generation from the WT is directly fed into the grid.
Furthermore, a home-made weather station including anemometer (Davies, 6410) and pyranometer (Apogee Instr., SP-215) is used for monitoring wind direction, wind speed, ambience temperature and solar radiation at the pilot site.

Laboratory tests
Laboratory tests were carried out in order to compare the performance of two MCDI modules with different setups. The focus in evaluating the performance of each setup is its SEC, which is defined by the used energy divided by the produced volume of the diluate (V d ). Two different types are therefore defined. One SEC for desalinating the water in the MCDI module to separate the salt ions from the water, taking into account the electrical power input measured at the outlet of the according power supply (P CDI ) and (t), and another SEC CDI,sys for the system, which includes the performance of the diaphragm pump (P p ). They are calculated as stated in the following equations: To compare the performance of the MCDI modules of desalinating water, the charge efficiency η ch can be used.
Hereby, the electrical current I ch during charge time is compared to the amount of adsorbed ions in the electrodes as stated in the following equation: where Δm is the removed amount of total dissolved solids (TDS) in g, M the molar mass, Q the electrical charge and F the Faraday constant. For all calculations, the molar mass of NaCl (M ¼ 58.5 g/mol) is assumed.

Sample analysis
The raw water samples were taken directly from the pipe behind the submersible pump after water was abstracted for some time.

RESULTS AND DISCUSSION
Raw water quality at the pilot site The chemical and physical properties of the raw groundwater in comparison to the Vietnamese standard for drinking water quality are summarized in Table 1. Analysis was carried out at different times before commissioning.
In addition to initial elevated concentrations of Fe 2þ (c max ¼ 3.1 mg/L) and t-As (c max ¼ 11 μg/L), the groundwater at the pilot site has an increased salinization with TDS concentrations of c > 1.5 g/L. In order to assess the groundwater quality at the beginning of the pilot trials, relevant water parameters were monitored over a period of 1 month directly before commissioning. During this period, an average of 15.5 m 3 of water was extracted on a daily basis, while no infiltration was carried out. This volume of water is greater than the extraction limit of 8.3 m 3 /d.
Measurements showed a slow increase in Fe 2þ and t-As concentrations during this month of sampling until the start of the SAR operation, which indicates that the extraction of large amounts of groundwater without treatment has an influence on the concentrations given hydrogeologically.
However, the averaged concentrations of Fe 2þ and Mn 2þ were similar compared with the values from March 2018, and the averaged t-As level decreased slightly, remaining below the WHO guideline value of c ¼ 10 μg/L, which also indicates that the rise in the contaminant levels during the measurement period was within the naturally occurring fluctuations. The value for the t-As concentration from the precommissioning phase differed from the initial measured value of 11 μg/L (see Table 1) 1 year before and can be also attributed to naturally occurring fluctuations.

Iron, As and Mn removal (SAR)
Results for the first 181 days of operation of the SAR plant for t-As, Fe 2þ and Mn 2þ removal are shown in Figure 4.
The averaged background concentrations measured during 34 days before commissioning (see Table 1) are considered as the initial concentrations for Fe 2þ , Mn 2þ and t-As, which are indicated at the y-axes as c 0,Fe , c 0,Mn and c 0,As , respectively. During this time, the t-As and Fe 2þ concentrations increased to c As > 4.8 μg/L and c Fe > 2.1 mg/L due to naturally fluctuations and the extraction of unusual high amounts of water without infiltration cycles. The Mn concentration remained at c Mn > 0.24 mg/L. Thus, the t-As concentration at the first day of SAR operation is c As > c 0,As , indicating as well the slower removal of As over Fe. Measurements in Table 1 show also that the t-As concentrations can be above 10 μg/L on this site.
After the start of operation, the concentration of As was continuously lowered following the removal of Fe 2þ until the value 0.55 μg/L was reached on day 78, corresponding to elimination rates of 76%. The highest increase in t-As measured from day 91 to 120 from 1.0 to 1.6 μg/L follows the rise in Fe due to the technical failure of the pump outlined above. The t-As mobilization decreases immediately when normal operation continues from day 120. The last measured value of t-As indicated a concentration of  1.1 μg/L and thus a removal of 54%. These two abovementioned removal rates for t-As are acceptable for an already significantly low background concentration of c 0,As ¼ In general, the results show a slow but continuous removal for all three species (Fe 2þ , As and Mn 2þ ), and after 181 days of operation, no breakthrough above the limits for drinking water standard was observed.
In addition, ammonium (NH 4 þ ) was lowered from initial concentrations of 3.1 mg/L to an average of 1.1 mg/L corresponding to a removal of 65%.

Desalination results (MCDI)
To determine the ideal operational parameters for the MCDI plant, laboratory tests with synthetic water (setup 2) have been carried out. Subsequently, these parameters were adjusted for the operation at the pilot site with setup 1. The comparison between lab and pilot scale is discussed below.
The optimum process parameters for the different MCDI setups employed are shown in  (1)). After the transition from the laboratory to the pilot site, a higher SEC CDI,1 ¼ 1.75 kWh/m 3 has been achieved, treating a complex salt matrix with setup 1 and thus bivalent cations as well, which can reduce Na þ and Cl À ions adsorption (Suss ). setup 2) is not considered in the following (see Equation (2)).
The SEC CDI,sys can be calculated using Equation (2) to SEC CDI,sys,1 ¼ 2.20 kWh/m 3 for setup 1 and SEC CDI,sys,2 ¼ 1.08 kWh/m 3 for setup 2. The performance of the diaphragm pump (P p,1 ¼ 8.9 W and P p,2 ¼ 1.8 W, respectively), which is proportional to the pressure loss over the MCDI module, is then added to the power consumption for the desalination. It should be noted that MCDI can be also operated by gravitational flow in combination with a flow regulating valve, and consequently, the diaphragm pump can be omitted. The comparatively low water recovery in setup 1 of 33% shows that the size of the module should be increased to handle the water under pilot conditions.
Thus, a higher water recovery and hence a lower SEC value could be achieved.
The desalination with MCDI setup 1 on the pilot site with operational parameters of Table 2 started after the pre-treatment with SAR was running stably on day 112 (see Figure 4). The late application of MCDI also ensures that the Fe 2þ concentration in groundwater is in compliance with the manufacturer's feed water recommendations of Fe 2þ < 0.5 mg/L (Voltea ). Although the desalination efficiency of MCDI can be mainly accessed by the overall salt removal (see Table 2), the adsorption efficiency for selected ions was taken within 12 weeks to evaluate the overall ion removal especially for Fe 2þ and t-As. The results also showed specific ion removals of 75%    To use the solar power production more efficiently, the operation of SAR is shifted during day time as shown in Figure 7(d), demonstrating that an autonomous (off-grid) operation is feasible. By doing this, the consumption of energy from the grid can be significantly reduced by 3.65 kWh, which is the average energy of one full infiltration cycle, and thus the degree of autonomy is increased. A high degree in autonomy or even of grid solutions is worthwhile, especially for a possible operation on the numerous small Vietnamese islands.

Renewable energy supply
In total, the average energy consumption for the operation of the SAR and MCDI module was 11.6 kWh/d. A degree of autonomy for the overall pilot plant of 97% could be achieved with a total energy supply of 11.2 kWh with PV. The energy deficit was supplied by the power grid. If the wind power of 3.4 ± 0.3 kWh/d at an average wind speed of 4.0 ± 1.3 m/s can be used, the degree of autonomy increases to 126% and thus 3.0 kWh/d of electrical energy can be fed into the grid, which provides governmental financial benefit (see the section State of the art), or used for a different purpose.  It should be noted that the SEC of the total SAR process consists of an SEC for the subsurface treatment (SEC SAR ) and an SEC for supply of the product water (SEC SAR,supply ) (see Table 3). On the pilot site, the water is supplied by the PENTAX well pump to the rooftop of the house via an elevation of H ¼ 14 m, with an average performance of

CONCLUSION AND OUTLOOK
• SAR has proved a feasible in situ remediation technique for Fe, Mn and As in brackish anoxic aquifers in combination with MCDI desalination, resulting in a total SEC SEC pilot, tot ¼ 3.97 kWh/m 3 of drinking water. Fe, Mn and As concentrations could be lowered below drinking regulation standards. The TDS removal by the MCDI of 1,372 mg/L to a concentration of c ¼ 188 mg/L could be achieved.
• Finally, two water streams were produced, which can be modularly adjusted according to the local requirements.
One treated by SAR with low Fe, Mn and As but high TDS, for washing purposes and another additionally treated by MCDI, for drinking purposes.
• The very low efficiency of the power supply for the MCDI module shows the improvement potential of the homemade desalination plant on the pilot site.
• Based on the previous experience, it is expected that the efficiency of the system can be increased by upscaling the capacity of the MCDI module.
• The supply with 3 kW p solar PV energy was feasible and a degree of autonomy of 97% for the water treatment was achieved, producing 0.48 m 3 drinking water and supplying 3.16 m 3 of washing water per day (out of the potential 8.3 m 3 ). Using a 2 kW p WT, a degree of autonomy of 126% can be realized. However, the use of the WT was difficult because the high noise level was not accepted by those living in the neighborhood.