Adsorption and (induced) desorption of Cd(II) from the corrosion scales of water distribution pipes, following a deliberate contamination event

Intrusion of toxic heavy-metal cations into water-distribution systems (WDS) may cause severe adverse health-effects on large populations, along with an undesirable psychological impact. The corrosion (scale) layer, that invariably develops on the pipes’ inner walls, is capable of adsorbing a significant mass of metal-cations and releasing them thereafter via diffusion to the water once operation is resumed, thereby causing a secondary contamination event. To overcome this, the contaminant should be completely removed, in a controlled fashion, from both the aqueous and scale phases, with minimum damage to the pipe’s physical stature. This study determined the range of the Cd(II) adsorption capacity of corrosion-scales and quantified alternative treatments for desorbing it, using an assortment of metal water-pipes, extracted from the WDS. Batch, waterrecirculation and flow-through experiments were conducted to determine the extent of Cd(II) adsorption and the best way to desorb it. Corrosion-scales showed substantial Cd(II)-absorption capacity (up to 0.75 mg Cd(II)/g scale) with an approximately linear relation between the aqueous Cd(II) concentration and the adsorbed mass. Desorption experiments included dosages of various acids. Sequential rinsing (eight pipe-volumes) by pH3 solution was found to be the best approach, releasing close to ∼100% of the adsorbed Cd(II), with only a minor effect on the pipes’ integrity.


GRAPHICAL ABSTRACT INTRODUCTION
Water distribution systems (WDS) may be contaminated intentionally as an act of terror, or unintentionally, by human error or system failure. To name a few incidents, in 2002, an attempt to poison drinking water by cyanide was averted in Italy (Carroll ); in 2012, 140 Afghan students and their teachers were admitted to a local hospital after drinking contaminated water from their school's water tank (Popalzai ); in 2018, an attempted ricin/anthrax water-poisoning terror attack was reported in Italy (Xuequan ); and recently in Israel, a cyber-attack was recorded in an attempt to manipulate drinking-water chlorine-dosing pumps (Kerstein ).
Once a contamination event is identified, the contaminated section of the WDS should be immediately isolated and then cleaned, before it is returned to service (Qiu et al. ). Potential water contaminants, as related to terrorism activity, may be divided into four groups, based on their source and chemical properties, as follows: (1) organic micro-pollutants; (2) fuel products; (3) microbial agents; and (4) toxic metal ions (Israel Ministry of Health ). Depending on the nature of the contamination, the cleaning process may include flushing, absorption of the contaminants by surfactants, complexing agents or activated carbon, and application of oxidation/ reduction and/or disinfection measures. This paper focuses on contamination by heavy metal ions, and specifically on Cd(II) salts. The cadmium ion was chosen as a representative metal contaminant because of its high toxicity to humansthe World Health Organization (WHO) recommends a maximal concentration of 3 μg/l in drinking water (WHO ). Additionally, Cd(II) has a relatively high solubility in water, as demonstrated by its K sp values: 10 À13.7 , 10 À12.1 and 10 À2.6 for Cd(OH) 2(s) , CdCO 3(s) and CdCl 2(s) , respectively, which makes it an easy chemical to apply to a WDS without the contamination being discovered immediately (Charlton & Parkhurst ; Powell et al. ; Tchounwou et al. ). Assuming a fairly 'common' drinking water composition (total inorganic carbon concentration (C T ) of 10 À3 M, pH ¼ 7.5, [Cl À ] ¼ 2·10 À3 M; and temperature of 25 C), these K sp values translate into a maximal dissolved value of 0.1 mg-Cd(II)/l, i.e. 33 times higher than the allowed drinking-water threshold.
In addition to being relatively soluble, the cadmium ion is stable only at an oxidation state of þ2 (i.e. Cd(II)). Thus, theoretically, its intrusion to the WDS cannot be treated by dosing an oxidizing or reducing agent, but rather, by flushing the WDS, as previously suggested by Ohar et al. (). In other contamination events, the toxicity can be minimized by reduction (e.g. Cr(VI)) or oxidation (e.g. As(III)), however in most cases toxic metal ions are unaffected by the redox value (e.g. Cd(II), Be(II), Ni(II)) (Bruce & Odin ; Fawell ; Tchounwou et al. ). Generally speaking, since applying a vigorous treatment procedure can also cause severe damage to the pipes, in-situ treatment is not necessarily the best solution, and it might make sense to first remove the contamination from the pipe and then (if required) to treat, or otherwise discard the solution, in a safe fashion. However, if a fraction of Cd(II) (or any other heavy metal), gets adsorbed on the pipe's corrosion scale, treating only the aqueous phase will not completely remove the contaminant and a secondary contamination event may unfold by a subsequent gradual release of the adsorbed Cd(II) once the system is put back into operation. Such a phenomenon was observed before by Munk (Peng et al. ). The deposition of corrosion products, along with other minerals (e.g. CaCO 3(s) ), on the pipe's wall forms a layer on the inner side of the pipe, which is commonly denoted corrosion scale. The composition, structure and depth of the scale vary significantly from system to system and are mainly a function of the age of the pipes and the water quality that flowed in them during their lifespan, along with the hydraulic conditions (Sarin et al. ). The quantity, the reactivity and the fractionation between different iron corrosion products (e.g. magnetite, lepidocrocite, siderite, hematite, goethite, etc.) may vary significantly between different pipes, depending on the corrosion scale structure and age, and the source water quality (Yang et al. ).
Previous studies have shown that while inorganic con-  (Yao et al. ), yet the adsorption of relatively high Cd(II) concentrations (i.e. concentration higher than the drinking-water threshold) on pipeline scales has not been investigated thus far. That said, the main factors which were found to affect Cd(II) adsorption/desorption on/from synthetic iron compounds were solution pH, Cd(II) concentration, ionic strength, competing ions in solution, temperature, and the adsorption period (Mustafa et al. , ). It was also shown that Cd(II) adsorption was more pronounced in waters with a high carbonate concentration, likely as a result of precipitation and subsequent adsorption of CdCO 3(s) . Previous studies (Peng et al. ; Gao et al. ) also reported on the potential adsorption capacity of pipeline scale when it was exposed to extremely low Cd(II) concentrations, emphasizing the importance of a controlled Cd(II) desorption in order to prevent secondary contamination. It thus seems important to develop a desorption procedure that does not damage the pipe's integrity nor dissolve its corrosion layer considerably, taking into consideration that the same factors that promote Cd(II) desorption can also act to decompose the corrosion scales and may cause a lingering red water event (Lahav & Birnhack ).
In the current paper, we aimed at investigating Cd(II) adsorption and desorption from pipeline corrosion scales in order to identify a method that would maximize the release of the adsorbed Cd(II) while not jeopardizing the pipe's integrity. Due to the high heterogeneity of pipeline scales and the difficulty in working with undisturbed corroded pipes, the work started by performing batch experiments on representative corrosion layers that were peeled from a variety of old iron pipes before executing experiments on 'real' old and corroded pipes. Because of the large heterogeneity between the corrosion scales found on live pipes, the experiments in this work may serve for establishing a reasonable working range for the possible adsorption capacity value of the scales for Cd(II) and particularly for determining the best method to be used for removing the adsorbed Cd(II) from the scale following the contamination event.

MATERIALS AND METHODS
Collection and stabilization of solids extracted from pipeline corrosion scales Peeled scales from old corrugated iron (CI) pipes (previously used for cold water transport in northern Israel) were collected, sieved (104 μm) and rinsed with desalinated water for 27 h. The aqueous phase was replaced five times during the rinsing period and the discharged solutions were analyzed for Fe, turbidity, TDS and pH. After the stabilization period, the wet scales (∼75% solids) were kept refrigerated. The surface area of the stabilized scale was measured (BET) at 45.5 ± 6.4 m 2 /g.

Batch adsorption experiments
A 500 ml beaker comprising the examined water quality combined with the desired Cd(II) concentration and 1 g of stabilized scales was placed in a rotating shaker (temperature range 24-26 C). In each set of experiments, the water content of the scale and the total suspended solids (TSS) concentration were analyzed (Method 2540D in Standard Methods ). The results indicated that no significant mass loss occurred during the adsorption experiments.

Batch desorption experiments
Desorption experiments were carried out in closed beakers placed inside a Julabo SW22 water bath shaker (temp ¼ 25 ± 0.2 C). Each desorption experiment was preceded by an adsorption stage with an initial Cd(II) concentration of 0.64 mg/l. After 24 hours of adsorption, 450-470 ml of the Cd(II)-contaminated solution was separated from the scale using a magnet (neodymium 80 × 80 × 20 mm) and 200 ml of fresh solution (free of Cd(II)) was added to the beaker (that contained the scale and the remaining solution from the adsorption step). Based on the resulting water volume, the required dosage of the desorption agent was calculated and added.

Water-recirculation batch experiments on live CI pipes (pipes #1-#3)
A corroded CI pipe (one metre in length, 0.0254 m diameter), extracted from the water distribution system just prior to the experiment, was connected to a flow rotameter, a Masterflex peristaltic pump (pump head model 77601-00), a ten-litre feeding tank and a sampling point. Between the extraction from the distribution system and the experiment, the pipes were kept filled with local water and sealed. First, a stabilization stage was performed, in which local groundwater (the same water that the pipe was exposed to during its lifespan) was circulated through the pipe at a flow velocity of 0.2 m/s for four hours (inline pressure of 2.7-3.0 bar) followed by 48 h in which the groundwater was retained stagnant in the pipe. Next, two adsorption experiments were performed with the same water, spiked with initial Cd ( (2) desorption by a pH5 solution (groundwater to which HCl was added); (3) desorption by pH3 acidified tap water; (4) one or two additional rinses by pH3 water.

Single-pass flow-through experiments on live pipes (pipes #4-#7)
Three ∼ten-metre (diameter ¼ 0.019 m) live CI pipes were extracted from the WDS and connected to a peristaltic pump that pumped water from a 500 l feed container at a pressure of 2.7-3.0 bar. The water flowing out of the pipe was collected in a second 500 l container (after being sampled). A stabilization stage was performed before the experiments, similar to the one applied in the recirculatingflow experiments: tap water was introduced in the pipe at a flow velocity of 0.2 m/s for 4 h followed by a stagnation period of 48 h. Table 1 summarizes the experimental steps and flow regimes that were applied in each step. The steps were performed sequentially following the stabilization phase.
Distinguishing between adsorption of dissolved Cd(II) and of CdCO 3(s) (pipes #8 and #9) The first stage of this phase comprised circulating tap water at ∼2.7 bar and 55 C through eight new black iron pipes (length ∼24 m, diameter 1 0 ) for enhancing scale formation for two months under a daily flow regime of a solution circulating for eight hours followed by retaining the water in the pipe until the next morning. During this time, air was added to the solution via a diffuser and the Ca 2þ and alkalinity concentration were maintained such that the CaCO 3(s) precipitation potential was always slightly positive. For the experiment report here two pipes were separated from the system for Cd(II) adsorption with tap water to which 10 mg Cd(II)/l was added. The experimental flow regime was similar to the one described in Table 1 for the adsorption phase.

Approximating the average scale mass in the pipes
The average mass of scale in the pipes used in the recirculated and live-pipe flowthrough experiments was estimated by carefully peeling 20 cm length of pipe scales and weighing the obtained dry (three hours at 60 C) sample. and (4) deionized water with EC < 1 μS/cm. The characterization of the first three water sources is shown in Table 2.

Analyses
The cation and sulfate concentrations were analyzed by  Three 20-min flowrate phases at 0.2 m/s, followed by 2 h stagnation. Then an additional overnight stagnation phase (the solution in the pipe was replaced between the two stagnation periods).
Flush with raw tap water Similar to the Cd(II) adsorption flow regime.
Intentional desorption of Cd(II) by a pH3 solution (tap water acidified by HCl) The contaminated pipe was rinsed three times in a row, each lasting 2 h at batch conditions, followed by an overnight stagnant phase. The procedure was conducted twice, a total of eight rinses (six 2 h rinses and two 18 h rinses).
predominantly from iron. Much lower fractions of zinc, calcium, sodium, and potassium were also observed. The results are summarized in Table 3. No significant difference was found in the composition of scales before and after the applied stabilization procedure.   Original scale 5.1 ±0.4 82.9 ± 4.1 1.9 ± 0.2 2.9 ± 3.0 7.2 ± 1.7
Temp ¼ 25 C. adsorption. It is emphasized that the scales used in this experiment were peeled from many different pipes and thus represent various conditions with respect to the corrosion layers that developed on the original pipes. Clearly, the layered structure of the original scales was different, as was also probably their age and composition, due to both redox reactions (e.g. oxidation of ferrous-containing iron oxides to ferric oxides) and the fact that the scales were intensely rinsed during their 'stabilization' phase, which invariably led to a change in the solids composition and layer structure.
It is also noted that 'old' corrosion scales are expected to be less reactive than 'fresher' precipitates (Sarathy et al. ).
Nevertheless, the results indicate that the applied procedure is capable of supplying information (albeit semi-qualitative) on the ability of pipe corrosion scales to adsorb Cd(II). The results listed in Table 4 seem to indicate that 'fresh' scales have a higher capacity for Cd(II) adsorption than the older (and mostly oxidized) mixed scales. They also indicate that when the Cd(II) concentration is relatively low, and sufficient time is allowed for adsorption, practically all of the metal ion can be expected to be adsorbed by the scales, a fact that would render it almost non-detectable in the aqueous phase. However, once Cd(II)-devoid freshwater flows in the pipe the adsorbed Cd(II) mass can be expected to diffuse gradually to the aqueous phase over time and contaminate the water. Such an occurrence may even be Next, in order to test more realistic contamination conditions, flow-through experiments were conducted with three live pipes (denoted pipes #4, #5 and #6). Figure 3 depicts the adsorption results obtained over time with these pipes, which were disconnected from the WDS just before the start of the experiment. The pipes were left at their original position, and were thus almost entirely undisturbed.
Horizontal cross-sections from the three pipes, showing their developed corrosion scales, appear underneath the graph. Water contaminated with 0.64 mg/l was pumped through the pipes using the flow pattern described in  To test the difference between adsorption of dissolved Cd(II) and adsorption of CdCO 3(s) , a side-experiment was conducted on two pipes that were exposed to similar conditions and that were assumed to have an almost identical corrosion layer. Tap water was dosed with CdCl 2 to result in 10 mg-Cd(II)/l. In the experiment performed on pipe #8, the solution pH was reduced to pH6, to attain negative precipitation potential with respect to CdCO 3(s) . On the other, in the experiment performed on pipe #9, the water was allowed to stand for 1 h after the CdCl 2 addition to allow for full precipitation of CdCO 3(s) , before the water was pumped through the pipe. Turbidity-based side-experiments showed that 1 h was sufficient for the CdCO 3 precipitation to attain its full extent. The results, shown in Rather, the percentage of Cd(II) that was desorbed from the mass that was previously adsorbed is reported, with the aim of identifying the conditions that would allow for extracting and rinsing out close to 100% of the previously adsorbed mass. Since all the methods that were assessed to release the Cd(II) from the scales could also have a potential adverse impact on the integrity of the pipes, the mass of iron and zinc that dissolved into the water in the course of desorbing the Cd(II) is also reported.
The first set of experiments was conducted on mixed scales loaded with Cd(II) and its purpose was to identify the best desorption treatment. Desorption experiments were attempted by acidifying the background water with a strong acid (HCl), leading to an initial pH5, pH4 and pH3. The acidified solution was brought into contact with the scale and the solution was allowed to stabilize for 24 hours. The pH value increased to around pH6 after about two hours in all of these tests and remained there throughout the experiment, indicating that the dissolution effect wears off after ∼2 h.
Weak acids (acetate and citrate) were also tested in the same fashion at pH4 and pH5. These organic acids have a relatively high buffering capacity around the target pH values tested and citrate also forms stable soluble complexes with divalent metal cations. Finally, the effect of dissolving a sulfide salt (Na 2 S) was also tested, with the underlying hypothesis that the formation of CdS (s) , which has a very low K sp , would desorb the Cd(II) from the scales and maintain it as a CdS colloid in the water, which could be readily washed out.
The results of these experiments, which were conducted at a water volume to scale-mass ratio of ∼0.357 L/g, are summarized in Table S1 in the Supplementary Information file. Generally speaking, Table S1 shows that the most efficient Cd(II) recovery (>80%) was attained when the mixed scales were flushed with a pH3 solution. Right after the acidification, the pH of the solution went up, specifically due to dissolution of carbonate solids, and generally stabilized at ∼pH6 after two hours. The redox value, measured in one experiment, stabilized around 225 mV, predominantly due to the low pH value. The dissolved oxygen concentration was close to saturation throughout the run. To test more realistic conditions, a second set of experiments was performed on the single-pass flow-through live pipes (pipes #4-#7). The water volume to scale-mass ratio (single rinse) applied in these experiments was 0.015-0.02 l/g, which is around one order of magnitude lower than the ratio applied in the desorption experiments on pipes #1-#3. The duration of each rinse was also lower (2 h in the first six rinses and then 18 hours for rinses 7 mass), the attainment of close to 100% Cd(II) recovery required at least eight full pipe-volume rinses, and a cumulative desorption period of at least 48 h. It therefore appears that both the volume of the water that is required in this procedure (that needs to be eventually discarded to a hazardous waste site) and the period of time that is required for the desorption can be considered realistic.
The zinc and iron concentrations recorded in the rinsing water are listed in Table S2 in the SI file. The low iron mass that was released is an indication that the dissolution/breakdown of the iron oxides in the corrosion layer was minimal.
Zn(II), on the other hand, was released from the scales at a higher mass, indicating that it had adsorbed on them over time, after being released from the ZnO coating of the inner (galvanized) pipe surface. Either phenomenon is not alarming from the pipe-integrity point of view and, although these experiments should clearly be repeated with largerdiameter pipes, the conclusion thus far is that the suggested procedure is not overly damaging to the pipe's physical condition. The calcium, iron, and zinc percentages out of the dry scale mass in pipes #4-#6 and the mass per surface area of the total inorganic and organic carbon in the scale of the same pipe are shown in Table S3 and Table S4 (SI file), respectively, before and after the adsorption/ Since according to our study old pipes may develop a corrosion scale layer of up to ∼10 g/cm 2 inner pipe surface area, the Cd(II) mass that can be adsorbed on the pipes' walls is substantial, as indeed shown in the adsorption experiments conducted in this work.
Desorption experiments showed that the most efficient method for desorbing the Cd(II) without overly harming the pipes is to consecutively rinse it with tap water acidified to pH3 by strong acid dosage. Eight sequential rinsing periods (six 2 h periods followed by two 18 h periods) that was used to fill the pipe, and then drained and analyzed. Once full, the pipe was sealed for two hours (in the first six rinses) and then for 18 hours (in the last two rinses). The final pH value of the water at the end of all the (acidified) rinsing steps tended towards pH6. Rinse #0 represents the desorption caused by raw (non-acidified) water.
were found to extract close to 100% of the cadmium without dissolving a large mass of iron from the scales. Zinc, on the other hand, was released in larger quantities, but the overall integrity of the pipes was not (at least visually) compromised. The experiments described in this work should be repeated with other toxic heavy metals (Ag þ , Co 2þ , Hg 2þ , Cu 2þ , etc.) and also larger and more diverse types of metal pipes installed in water distribution systems, before a formal procedure could be recommended for application in heavy-metal-based contamination events. A further challenge would be to upscale the problem from a local contamination-event treatment to dealing with rehabilitation of a contaminated portion of a real water distribution system. This will require constructing an optimal decision support model on locations where treatment chemicals should be injected into the network, as well as their concentrations, injection times, flow rates and drainage outlet locations, which is based on the topology and topography of the system.