Cr(VI) and Cr(III) Release From Different Layers of Cast Iron Corrosion Scales in Drinking Water Distribution Systems: The Impact of pH, Temperature, Sulfate, and Chloride

Chromium accumulated from source water and pipeline lining materials in corrosion scales could potentially be released into bulk water in drinking water distribution systems (DWDS). Chromium behaviors between corrosion scale phase and the surrounding water phase vary spatially in different layers and temporally in different DWDS running periods. In this study, corrosion scales sampled from actual DWDS were rst characterized by SEM, XRD, XRF, and the modied BCR three-step sequential extraction procedure. Then scales were divided into the outer and inner layers for subsequent analysis. Static accumulation and release experiments were performed with Cr(VI) and Cr(III) on two distinct scale layers to systematically assess the inuence of pH, temperature, sulfate, and chloride. The release behaviors of Cr(VI) under the co-effect of multiple factors were investigated in orthogonal experiments. Results showed that in the outer and inner layers of corrosion scales, chromium exhibited differences in accumulation and release behaviors, with the outer layer accumulating less and releasing more. The mechanisms of chromium retention based on different iron (oxyhydr)oxides were discussed.


Introduction
Drinking water safety and quality are fundamental to human health and social development. It is estimated that in China the drinking water quality classi cation in metropolitan areas reduces by about 10% throughout drinking water distribution systems (DWDS). Corrosion scales in pipelines are one of the major factors responsible for this reduction in quality and are a key concern, serving as a source and a sink for various contaminants. Heavy metals can accumulate in corrosion scales and under certain conditions be released into drinking water, signi cantly reducing water quality (Gerke et  contamination in DWDS has attracted increasing attention due to its mobility and toxicity. Source water and leaching from cement mortar linings are the two of the main sources of accumulated chromium in DWDS (Estokova et al., 2018;Veschetti et al., 2010). The toxicity of chromium is largely related to its valence state (Moreira et al., 2018). Long-term exposure to high concentrations of Cr(VI) will pose a severe health risk by adversely affecting the immune system. In contrast, Cr(III) is an essential nutrient at trace levels (Liu and Yu., 2020). According to World Health Organization (WHO) recommendations, the maximum permitted concentration of Cr(VI) in drinking water is 0.05 mg/L (WHO, 2017). However, Cr (III) has the potential to be oxidized into Cr(VI) by strong oxidants such as chlorine, ozone and permanganate, which are commonly used as disinfectants in the water treatment process (Chebeir et al., 2016;Lindsay et al., 2012). As a result of these toxicological risks, environmental conditions that in uence the accumulation or release of both Cr(VI) and Cr(III) are of concern.
Water quality parameters such as temperature, pH, alkalinity, sulfate and chloride concentrations, affect interactions between the corrosion scale phase and water phase. Furthermore, Zachara et al. (1988) found that excess sulfate was likely to enhance the sorption of chromium. However, few studies have systematically investigated the in uencing factors and mechanisms of chromium accumulation and release from iron corrosion scales, or the behavior of chromium under the co-effect of multifactors.
The composition and structure of corrosion scales can also in uence heavy metal accumulation and release behaviors. To date, the characteristics of corrosion scales on different pipe materials, under different conditions, have been analyzed most comprehensively Peng et al., 2010;Niu et al., 2006;Sarin et al., 2004). Typical iron corrosion scales are composed of three distinct layers, a surface layer, a shell-like layer, and a porous core layer (Sarin et al., 2001). Different structures and compositions have been observed in samples from different layers of the same corrosion scale tubercle. The formation of corrosion scales is a dynamic process and with ongoing of DWDS, the scale layer thickens and the original outer layer gradually becomes an inner layer. The signi cant physiochemical difference between layers may affect heavy metal release. Therefore, the behaviors of heavy metals in corrosion scales may vary spatially, in different corrosion scale layers and temporally, in different DWDS running periods. However, few studies have comprehensively analyzed these differences. Therefore, in this study, chromium behaviors between corrosion scale phase and the surrounding water phase were investigated spatially in different scale layers. The aim of this work were to: 1) identify the physicochemical distributions of heavy metals in different layers of grey cast iron corrosion scales; 2) reveal the impact of pH, temperature, sulfate and chloride on chromium retention by different scale layers; 3) analyze chromium retention mechanisms based on different iron (oxyhydr)oxides; 4) investigate release behaviors of Cr(VI) under co-effect of multiple factors and provide reference for Cr(VI) control strategies in DWDS.

Samplecollection and layering
Corrosion scales were sampled from a 200 mm-diameter gray cast iron pipe that had been in service for more than 20 years in the drinking water distribution system of Tianjin, China ( Fig. S1 (a)). Samples were sealed in bottles lled with nitrogen before use. Certain layering characteristics were typically observed in most of iron corrosion scales ( Fig. S1(b)). Layering was done with a spatula. The surface layer was a yellowish-brown thin sliding layer of about 0.5 mm, which was directly in contact with bulk water in DWDS. Lightly scraped off the surface layer with a spatula, the middle layer was found of shell-like layer, exhibiting a black, metallic luster, with a relatively hard and dense structure. After scraping off the middle layer, the rest was the porous core layer, which was yellow and black with a relatively loose structure. Prior to all experiments, the layered samples were separately crushed and passed through 100-mesh nylon sieves to ensure samples were fully homogenized. The collected powder samples were then dried in a freeze dryer and stored at 4 ℃ for analysis. The layered samples were partly used for characterization and BCR procedure, and part for the subsequent accumulation and release experiments. For the latter, the surface layer and shell-like layer were combined as the outer layer, while the porous core layer was considered as the inner layer. While it was worth noting that to ensure the uniformity of scales for the accumulation and release experiments, heavy metals contained in the original corrosion scales should be released rst. Thus that part of scale blocks were soaked with ultrapure water for 72 h, with ultrapure water replaced every 24 h. The soaked scales were then dried and crushed as mentioned above.

Characterization of iron corrosion scales
The morphology of three scale layers was observed by scanning electron microscopy (SEM, HITACHI S-4800, Japan). X-ray uorescence spectrometer (XRF, S4 Pioneer, German) were used to analyze the elemental content of corrosion scale samples. The X-ray diffraction (XRD) spectra were obtained by X-ray diffractometer (XRD, D/MAX-2500, Japan) to identify crystalline solid phases. The concentration of heavy metals in the original corrosion scales were analyzed by inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7700X, USA).

Modi ed BCR three-step sequential extraction procedure
The modi ed BCR three-step sequential extraction procedure was performed to establish the physicochemical speciation distributions of heavy metals in different layers of corrosion scale samples.
The experimental procedure is outlined in detail in Table S1 (Rauret et al., 1999), with all extracts then ltered through 0.45 μm membrane lters and stored at 4 ℃ for analysis by ICP-MS.
The modi ed BCR three-step sequential extraction process partitions elements into four fractions, classi ed as exchangeable and acid-soluble, reducible, oxidizable, and residual. The exchangeable and acid-soluble fraction is composed of both the exchangeable fraction and carbonate-bound fraction, containing heavy metals associated with exchangeable binding sites and carbonates. Heavy metals in the exchangeable fraction are relatively easily displaced by competing ions. Acid-soluble fraction metals are coprecipitated with calcite and therefore, can be easily dissolved by acid. The reducible fraction was dominated by heavy metals bound to Fe-Mn oxides. pH and oxidation-reduction potential (ORP) in DWDS have a signi cant in uence on the migration and transformation of the reducible fraction. When the water is reducible or anoxic, Fe-Mn oxides tend to dissolve and heavy metals are released. The oxidizable fraction refers to metals bound with complex organics or sul des, which are easily decomposed and released under oxidation conditions. The residual fraction consists of metals in the form of crystallised oxides within the crystal lattice of minerals, resulting in these metals being stable and non-bioavailable (Linge, 2008). Therefore, in contrast to the residual fraction, the other extractable fractions are relatively highly mobility and active, presenting a signi cant risk of pollution due to their environmental availability (Barcelos et al., 2020). In this study, the three non-residual fractions were considered together, to assess the mobility and release potential of heavy metals in each layer of corrosion scales under varying conditions.

Static accumulation and release experiments
Visual MINTEQ 3.1 was used to calculate chromium speciation under different pH conditions. All experiments were conducted with ultrapure water as the background water to eliminate the interference of other ions. For data quality control, each sample was measured in duplicate with results showing an error of less than ± 5.0%.

Accumulation of Cr(III) and Cr(VI)
According to the water quality conditions reported in real DWDS pipelines, the initial pH value and temperature were selected as variable factors in accumulation experiments. Initial concentration of chromium was set as 2.5 mg/L on the basis of relevant water quality standard of WHO, to simulate the rapid accumulation of chromium in pipe network accidents. The accumulation experiments were conducted independently for Cr(III) and Cr(VI), as shown in Table S2. For each pH condition (pH 4, pH 5.5, pH 7, pH 8.5, pH 10), 25 mL of chromium solution was added to a 50mL centrifuge tube that had been soaked in 10% nitric acid solution and rinsed prior to use. The solutions were then placed in a constant temperature shaker at 250 rpm for 1440 min, with 0.00250 g of layered scale powder then added after the required temperature (5 ℃, 15 ℃, 25 ℃) was reached. In addition, another experimental group was established at 25 ℃ and pH 7 for kinetic analysis, with solution samples collected at 0, 5, 10, 30, 60, 120, 240, 360, 480, 720, 1440, and 2880 min. Following collection, all solution samples were ltered through 0.45 μm membrane lters and acidi ed with nitric acid to obtain pH < 2 solutions. Inductively coupled plasma emission spectrometry (ICP-OES, THERMO iCAP-7400, USA) was then used to analyze chromium concentration in the solutions.

Release of Cr(III) and Cr(VI)
Release experiments were conducted to assess the in uence of variations in key factors of pH, temperature, sulfate, and chloride. The experimental parameters are described in detail in Table S3. The scale powder samples were pretreated by enrichment with 10 mg/L of chromium standard reserve uid for 24 h. More details can be found in the supplementary material. Centrifuge tubes were removed at 0, 5, 10, 30, 60, 120, 240, 360, 600, 720, 1440, 2880, and 4320 min, for analysis. The remaining experimental steps were the same as accumulation experiments, with ICP-MS used to analyze chromium concentrations.

Orthogonal experiments on Cr(VI) release
Orthogonal experiments were performed to identify the signi cance of factors affecting Cr(VI) release from different layers of iron corrosion scales. pH, temperature, sulfate, and chloride were applied as experimental factors, with each factor having three variation levels. The orthogonal design table L 9 (4 3 ) is shown in Table S4. All of the 9 experimental trials were carried out in parallel duplicate experiments. The Cr(VI) concentration in solution was measured at 2880 min. Range analysis was conducted to assess and compare the in uencing degree of all four factors. The corresponding average k i value and R value were calculated. The k i value represents the impact of the three variation levels of each factor, which was calculated as the mean concentration of Cr(VI) released. The R value indicates the difference between the maximum and minimum k i value for the three variation levels for each factor, with the factor exhibiting the highest R value having a stronger impact on Cr(VI) release. Variance analysis was also performed using SPSS v.22.0 to assess the signi cance of each factor.

Results And Discussion
3.1 Characterization and speciation distribution of each scale layer SEM images (Fig. 1) have demonstrated the differences in morphology of each scale layers. More loose and porous structure could be observed in the porous core layer (Fig. 1 (c)), which may be bene cial for metal adsorption and accumulation. More signi cant characteristics can be seen from images magni ed by 50,000 times ( Fig. 1 (d), (e), and (f)). Combined with the XRD spectra result (Fig. S2), the main components de ned in the shell-like layer of the corrosion scales were magnetite (Fe 3 O 4 ) and hematite (Fe 2 O 3 ), considered to be relatively stable. Goethite (a-FeOOH) and lepidocrocite (γ-FeOOH) were the main crystalline iron minerals in surface layer and porous core layer. The speciation distribution of metal elements in each scale layer was investigated to analyze their mobility and release potential in corrosion scales ( Figure 2). The fraction distribution of Fe showed little difference in each of the three layers. But beyond that, the residual fraction of all the other metal elements was signi cantly much more in the porous core layer, especially for Al (83.46%) and Cr (72.05%). This result demonstrates that metal elements in the porous core layer mainly present in stable matrices and were relatively immobile. Whereas metal elements in the surface layer and shell-like layer have stronger mobility and release potential. Therefore, the surface layer and shell-like layer of corrosion scale samples were combined to represent the outer layer in subsequent studies, in comparison to the inner porous core layer. In the surface layer and shell-like layer, the majority of Cr (67.34% and 41.78%) existed in reducible fraction. This corroborated that the reducible fraction dominated by heavy metals bound to Fe-Mn oxides exhibited a strong a nity for Cr (Rai et al., 1989).

Effect of pH
The effects of pH on chromium accumulation are shown in Fig. 3 (a). Data was measured at 1440 min when the accumulation process was approaching chemical equilibrium. Results demonstrated that the accumulated amount of Cr(VI) slightly increased under higher pH conditions, and the average value was about 1.6 mg/L. But it is of note that Cr(III) accumulation was much smaller than Cr(VI) and changed dramatically when pH increased to 10. Previous studies have shown that under acidic conditions, H + ions and metal ions compete with each other for the available surface sites (Gu et al., 2018), resulting in less heavy metal accumulation in corrosion scales. In addition, pH affects chromium speciation, especially Cr(III). The simulation results by Visual MINTEQ 3.1 (Fig. S3 (b)) show that Cr 3+ and Cr(OH) 2+  Therefore, although the initial concentration of Cr(III) was 2.5 mg/L, the actual concentrations present in solution were 1.83, 1.7, 0.23, 0.32, 1.66 mg/L at pH 4, pH 5.5, pH 7, pH 8.5, pH 10, respectively. Thus the accumulation ratio was calculated and illustrated in Fig.S4 (a). Similarly, Cr(VI) accumulation ratio increased slightly as pH rose. The accumulation ratio of Cr(III) showed a rising trend in general but increased rapidly around pH 7. This was mainly attributed to the high level of Cr(OH) 3 precipitation under near-neutral pH conditions. The accumulation ratio of Cr(VI) was about 90 % under all experimental pH conditions, signi cantly higher than that of Cr(III).
However, the pH of water in DWDS is universally neutral and at pH 7 the average amount of Cr(VI) accumulated was 1.56 mg/g. While only 0.16 mg/g Cr(III) was found to accumulate on average, approximately ten-fold less than Cr(VI) under the same pH conditions. There was also a difference of about 20% in accumulation ratio. This indicated that in DWDS Cr(VI) release should be considered as a major concern, due to its high toxicity and potential for accumulation.

Effect of temperature
The effect of temperature on chromium accumulation was investigated at different temperature conditions of 5 ℃, 15 ℃, and 25 ℃. Temperature mainly exerts an in uence by affecting the reaction rate, chemical equilibrium and microbiological processes. Fig. 3 (b) and Fig. S4  On the other hand, chromium accumulated in the inner layer was always more than that in the outer layer in general. Moreover, the accumulation of Cr(VI) was signi cantly greater than that of Cr(III). Compared with Cr(III), results indicate that it is easier for Cr(VI) to migrate from the water phase to the corrosion scale phase and, thus Cr(VI) was more likely to be absorbed and accumulate in corrosion scales.

Cr (VI) accumulation kinetics
The effects of reaction time on Cr(VI) accumulation are shown in Fig. S5. Due to su cient availability of surface sites, the accumulation rate of Cr(VI) increased rapidly within the initial 60 min. However, equilibrium was not reached until 2880 min. To further investigate the mechanisms of Cr(VI) accumulation, the pseudo-rst-order kinetic model, pseudo-second-order kinetic model, Elovich model, and Webber Morris intra-particle diffusion model were used to analyze Cr(VI) accumulation kinetics data. The kinetic parameters of these four models are summarized in Table S7. The correlation coe cients (R 2 ) of the pseudo-second-order kinetic model were greater than that of the pseudo-rst-order kinetic model, indicating that the rate-limiting step was the chemisorption process. The Elovich model R 2 values for outer layer and inner layer Cr (VI) accumulation were 0.889 and 0.910, respectively. The validity of the Elovich model implies that the initial Cr (VI) accumulation process was rapid, while the desorption process was slow and occasional (Labied et al., 2018). Webber-Morris intra-particle diffusion model data exhibited two unique linear plots, demonstrating that the accumulation procedure was not only controlled by intraparticle diffusion. Under all pH conditions tested, Cr(VI) released from the outer layer of corrosion scales was consistently greater than from the inner layer, which could be explained by different physicochemical speciation distributions of chromium, as shown in Fig. 2. The results of sequential extraction show that chromium in the outer layer of corrosion scales possessed greater mobility than the inner layer. In addition, chromium mainly existed in a bound form with carbonate minerals and Fe-Mn oxides in the outer layer, exhibiting to be more sensitive to pH variations.
Compared with Cr(VI), the release of Cr(III) uctuated considerably, as shown in Fig.4 (c) and (d). This was largely due to the release amount of Cr(III) being much smaller and therefore, more affected by instrument error. Results appear to show that Cr(III) has weaker mobility and was less easily released from the corrosion scale phase to the water phase.

Effect of temperature and release kinetics
In general, DWDS temperature conditions exhibit distinct seasonal and diurnal changes. Therefore, the effects of temperature on chromium release were studied and the results are shown in Fig. 5. The amount of Cr(VI) released at 25 ℃ (28.10 μg/L from the outer layer, 23.82 μg/L from the inner layer) was signi cantly greater than at 5 ℃ (10.35 μg/L from the outer layer, 10.22 μg/L from the inner layer) or 15 ℃ (13.81 μg/L from the outer layer, 12.76 μg/L from the inner layer). This may be attributed to the mobility of metal ions and the ion exchange rate being increased at higher temperatures, thereby chromium migrated more easily from the corrosion scale phase to the water phase. Moreover, the increase in temperature could accelerate the dissolution of precipitants such as carbonate bound fractions, resulting in an increase in the release of chromium.
Cr(VI) release kinetics tests in the temperature range of 5 ℃ to 25 ℃ were performed to further illustrate the effect of temperature, with the Elovich model, Double constant model, and pseudo-second-order kinetic model applied. Detailed information and parameters for these models are provided in Table S8.
The kinetic models selected all tted data well ( Fig. 5 (a)), while the Elovich model tted best on the whole, indicating that chromium release may be controlled by a variety of complex reaction mechanisms.
The values of the parameters for the two-constant model at different reaction temperatures (Table S8) showed that the values of the initial release rate constant 'a 2 ' varied relatively widely. This is consistent with previous results on the increase of Cr(VI) release as temperature went higher. In Elovich model, value 'a 1 ' represents the initial desorption rate and value 'b 1 ' is the desorption constant (Rezaei Rashti et al., 2014). In the present study, the Elovich model rate constant 'a 1 ' values increased widely in both the outer layer and the inner layer when temperature increased (Table S7). Moreover, the values of constant 'b 1 ' were compared in the outer layer and the inner layer, showing that 'b 1 ' values in the outer layer were consistently larger at all assessed temperatures. This result is can also be explained by the different physicochemical speciation distributions of chromium in the two iron corrosion scale layers.
The release of Cr(III) under all conditions was less than 8 μg/L, with data showing no obvious trend with varying temperature conditions (Fig. S6). However, combined with the data on the effects of temperature on Cr(VI) release, it can be concluded that the release potential of chromium is theoretically smaller at lower temperatures. Therefore, more attention should be paid to chromium release in high temperature environments and seasons.

Effect of sulfate and chloride
The concentration of chromium released into solution under three sulfate and chloride concentrations (50,150, 250 mg/L) was determined over time. The effect of chloride on Cr(VI) release is presented in Fig. 6 (c) and (d). When the concentration of chloride increased from 50 mg/L to 150 mg/L, the equilibrium concentration of released Cr(VI) increased from 8.94 to 13.20 μg/L, respectively. Similarly to sulfate, chloride also competes with heavy metals for adsorption sites and electron transfer reactions (Peng et al., 2013), increasing Cr(VI) release. The presence of sulfate and chloride could also increase the acidity of the solution (Sarin et al., 2004), promoting the dissolution of iron corrosion scales. However, when the concentration of chloride was increased to 250 mg/L, chromium release decreased by about 4.24 μg/L compared with that at 150 mg/L chloride, which was in agreement with previously reported results (Lytle et al., 2020). This phenomenon may occur due to the existence of CrO 3 Cl − under excessive chloride concentrations (Han et al., 2008), indicating that chloride may complex with Cr(VI) and therefore inhibit Cr(VI) release.  Table S9 and Table S10. Results demonstrate that Cr(VI) released from the outer and inner layer exhibited similar trends under the co-effect of multifactors. Whereas Cr(VI) released from the inner layer was always less than that from the outer layer, which also illustrates the signi cant difference in heavy metal release from different layers of corrosion scales. As for the outer layer, temperature, sulfate, pH, and chloride exerted an inhibitory effect on Cr(VI) release (in descending order), as shown by the range analysis calculation results. According to the results of variance analysis (Table S11) )). Whereas surface layer and porous core layer were dominated by iron oxyhydroxides (goethite (a-FeOOH) and lepidocrocite (γ-FeOOH)), relatively instable. The differences in crystallographic face and bulk structure of iron (oxyhydr)oxides make the retention mechanism and metal capacity vary. The mechanism analysis diagram of chromium retention by iron (oxyhydr)oxides is shown in Fig.S8.

Conclusions
The mobility and toxicity of chromium is largely dependent on their physicochemical speciation. For the rst time, the accumulation and release behaviors of chromium were assessed in two distinct layers of iron corrosion scales based on the speciation distributions of heavy metals. The following conclusions can be reached.
1. The results of modi ed BCR three-step sequential extraction showed that most of the heavy metals in the surface layer and shell-like layer of iron corrosion scales existed in mobile and bioavailable states.
2. In general, the accumulation of chromium in iron corrosion scales increased under conditions of higher pH and higher temperature. Cr(III) was more sensitive to the variation of pH and temperature, while the accumulated amount of Cr(VI) was signi cantly greater than that of Cr(III). Neutral or weakly alkaline pH conditions, lower temperatures, and lower concentrations of sulfate and chloride, were found to reduce chromium release. Overall, chromium was found to accumulate more easily in the inner layer, with greater potential to release from the outer layer of iron corrosion scales.
Thereinto, Cr(VI) has much more stronger potential of accumulation and release in corrosion scales thus exhibits more health risk.
3. The release behaviors of Cr(VI) under the co-effect of multifactors were demonstrated by orthogonal experiments. Results illustrate that temperature, sulfate, and pH signi cantly affected Cr(VI) release from the outer layer (1% signi cance level), followed by chloride (5% signi cance level). The degree of in uence of the four factors on Cr(VI) release from the inner layer was ranked in the descending order of pH temperature chloride sulfate, with chloride and sulfate exerting no signi cant effect, while pH and temperature had signi cant effects at the 5% and 10% signi cance levels, respectively.
4. The surface layer and shell-like layer have similar speciation distribution of metal elements but signi cantly different crystal structure. The difference of chromium behaviors between the outer and inner layer implies that chromium retention is effected by both metal migration potential and scale crystal structure. Surface complexation, surface precipitation are the principal mechanisms of chromium retention by iron (oxyhydr)oxides.
These ndings contribute to both scienti c and practical signi cance in the comprehensive and systematic understanding of chromium behaviors between the corrosion scale phase of two distinct layers and water phase However, this research only considered the accumulation and release behaviors of Cr(VI) and Cr(III) independently, while the in uence of co-existing Cr(VI) and Cr(III) and the potential for mutual conversion still requires further investigation. Declarations