Solving Copper Corrosion Problems while Maintaining ... - CiteSeerX

5 downloads 0 Views 3MB Size Report
Aug 30, 2001 - carried out using zinc orthophosphate chemicals of different Zn:PO4 ..... consistently with a solubility and stoichiometry of Cu3(PO4)2•2H2O.
Presented at the Ohio AWWA Annual Conference, Cleveland, OH, August 30, 2001.

Solving Copper Corrosion Problems while Maintaining Lead Control in a High Alkalinity Water Using Orthophosphate Michael R. Schock* Chemist Water Supply & Water Resources Division, USEPA Cincinnati, OH 45220 James C. Fox Superintendent Indian Hill Water Works Cincinnati, OH, 45243

Abstract Lead and Copper Rule sampling in 1992 uncovered high copper levels in many homes in the Indian Hill Water Works (Ohio) water system. The 90th percentile copper and lead levels were 1.63 mg/L and 0.012 mg/L, respectively. Indian Hill Water Works (IHWW) supplies water to several suburban communities to the east of Cincinnati. Finished water hardness is approximately 150 mg/L as CaCO3, total alkalinity approximately 250 mg/L as CaCO3, DIC approximately 60-65 mg C/L, and pH ranges from about 7.1 to 7.5, mostly near 7.3. Final treatment consists of chlorination, fluoridation, and at different times, addition of caustic and/or corrosion inhibitor. Historically, water heater failures and high copper levels resulting in blue water were reported in the 1950’s, and caustic addition was used to elevate the pH to as high as 8 at different times throughout the next 3 decades. In the late 1980’s and early 1990’s, zinc orthophosphate treatment was employed, which was mostly successful for lead and inconsistently successful for copper. In 1997 with the 90th percentile still at 1.54 mg/L for copper, the zinc orthophosphate treatment was withdrawn. In 1998, new pilot tests were carried out in collaboration with the Water Supply and Water Resources Division of USEPA in Cincinnati to do some new pilot testing and a more systematic analysis of copper corrosion control chemistry. Tests compared pH adjustment, partial DIC removal through anion-exchange and increasing orthophosphate dosing. The most economical alternative identified was an increased orthophosphate residual of 3 mg/L as PO4. The recommended treatment was implemented beginning in 1999. Lead levels were below 0.005 mg/l for the 90th percentile, and the 90th percentile copper levels were reduced to 1.04 mg/L, for the fall 1999 and Spring 2000 sampling rounds. Copper levels remain consistently below the Action Level, and IHWW has now successfully met simultaneous lead and copper control requirements of the regulations and is now in “reduced monitoring” status.

*

This is a work of the Federal Government and is not subject to copyright. E-mail address: [email protected]

Introduction Nine years after promulgation of the Lead and Copper Rule, many small and medium-sized water systems continue to have difficulties in complying with the copper Action Level, sometimes concurrently with problems meeting the lead Action level. The problem is especially troubling and widespread with ground water supplies having high alkalinities as are common in the Midwest. Orthophosphate and blended phosphates have been used by many of these systems, but success has not been consistent. Different utilities frequently receive many conflicting suggestions about courses of treatment for this problem. Lead and Copper Rule sampling in 1992 uncovered high copper levels in many homes in the Indian Hill Water Works (Ohio) water system. The 90th percentile copper and lead levels were 1.63 mg/L and 0.012 mg/L, respectively. Historically, water heater failures and high copper levels resulting in blue water were reported in the 1950’s, and caustic addition was used to elevate the pH to as high as 8 at different times throughout the next 3 decades. Scaling was a considerable problem and reports of blue staining continued. In the late 1980’s and early 1990’s, various coupon and field tests were carried out using zinc orthophosphate chemicals of different Zn:PO4 ratios and dosages. The zinc orthophosphate treatment was mostly successful for lead and inconsistently successful for copper throughout the early and mid-1990’s, yet the 90th percentile was still at 1.54 mg/L for copper in 1997. This prompted withdrawal of the zinc orthophosphate treatment, and a new evaluation of corrosion control approaches. Relatively little is known about the systematic impacts of different treatments on mitigating copper solubility (“cuprosolvency”) and copper release from waters containing high concentrations of dissolved inorganic carbon (DIC). These high-DIC waters were predicted to be corrosive on theoretical chemistry grounds (Schock et al., 1995a, 1995b; Edwards et al., 1996). Consistent with these predictions, the Indian Hill Water system had exceeded the regulatory action level for copper in 1992 (Indian Hill Water Works, 1994). Previous experience with dosing of zinc orthophosphate by IH Water Works was relatively unsuccessful in meeting regulatory metal concentration requirements. In 1998, IHWW began collaboration with the Treatment Technology Evaluation Branch of the Water Supply and Water Resources Division of USEPA in Cincinnati, OH, which was interested in gaining a better understanding of some of the problems associated with copper control in high alkalinity ground waters. The results of the research investigation could then be used to form the basis for revised treatment recommendations to be made by IHWW to the State of Ohio EPA, based on consideration of relevant resource, operational, and regulatory factors.

Indian Hill Water System

2

Indian Hill Water Works (IHWW) supplies water to approximately 15,000 residents and 250 businesses within several suburban communities to the east of Cincinnati. As of 2000, the average daily production was 1.89 MGD, with a maximum of 3.46 MGD (approximately 60% of maximum capacity). The system is fed by 10 wells in the Little Miami Buried Aquifer ranging from approximately 350 to 800 GPM, which is largely composed of limestone and limestone till. The treatment plant currently softens the water using cation exchange that removes approximately 50% of the hardness from approximately 300 to 150 mg/L as CaCO3. Alkalinity is approximately 250 mg/L as CaCO3, DIC is approximately 60-65 mg C/L, and pH ranges from about 7.1 to 7.5, mostly near 7.3. Final treatment consists of chlorination, fluoridation, and at different times, additions of caustic and/or corrosion inhibitor have been employed. A schematic representation of the Indian Hill Water Works Treatment plant is shown in Figure 1. Other than chemical stabilization, the method of treatment has not changed since the plant was built in 1949. The treatment plant is automated, and treats and pumps water to the distribution system around the clock. The distribution system consists largely of cast or ductile iron water mains, with a small percentage of asbestos-cement mains. The system covers approximately 20 square miles. Although some of the homes are older and have galvanized iron services, the vast majority of homes has copper piping from the main to the house, and of course, copper piping throughout the house.

Theoretical Treatment Alternatives for High-DIC Systems Water systems having high carbonate (DIC) contents and oxic conditions (persistent dissolved oxygen and chlorine residual concentrations) face numerous problems in controlling cuprosolvency, resulting from the high amounts of complexation of Cu2+ ion by carbonate and bicarbonate ligands (Schock et al., 1994, 1995a, 1995b; Edwards et al., 1996; Ferguson et al., 1996; Schock, 1999). Ideally, the high DIC concentration would translate into rapid formation of Cu2CO3(OH)2 (malachite), which would reduce copper solubility and byproduct release. All of the factors governing the rate of this aging phenomenon are not yet understood (Schock et al., 1994, 1995a; Hidmi et al., 1999; Edwards et al., 2000). Unfortunately, this beneficial malachite scale development may not take place rapidly enough. High DIC water systems across the United States have frequently failed to meet the regulatory action levels, in spite of the fact that Lead and Copper Rule targeting is biased towards lead and consequently favors plumbing that often has had now more than 15-20 years for protective copper scales to develop (USEPA, 1991, 2000). From a theoretical standpoint, orthophosphate dosing has been suggested by several research groups as a means to substantially reduce cuprosolvency and copper release under certain pH conditions (Schock et al., 1995a, 1995b; Lytle & Schock, 1996; Schock & Clement, 1996; Lytle et al., 1996; Lytle & Schock, 1997b; Schock & Clement, 1998; Schock, 1999). Several field studies, however, have had mixed results when employing various blended phosphate chemicals. One of the suggestions from the theoretical research is that high orthophosphate residuals may be required, and this

3

will ultimately suspend the operation of the normal scale mineralogy evolution and aging process, whereby the high surface-area cupric hydroxide and cupric oxide initially formed recrystallizes and structurally orders into much less soluble and more crystalline CuO (tenorite) or Cu2CO3(OH)2 (malachite). Thus, long-term use of high dosages of orthophosphate may not reflect optimal treatment from the copper control standpoint if wastewater plant loading of copper, zinc or phosphate becomes an issue with tighter regulations in the future. Another driving force might be tightened filtration requirements for ground waters. There are several alternative treatment schemes that would warrant future investigation, if an incentive develops to further optimize overall treatment. These include: •

Split-stream sequential cation exchange (zeolite softening) and anion exchange, both in chloride form, with blending and/or other post-treatment.



Split-stream sequential cation exchange (hydrogen ion form) and anion exchange (chloride or hydroxyl ion form), with aeration post-treatment for pH adjustment and enhanced TIC removal.



Split-stream mixed bed anion/cation exchange TIC removal with pH adjustment post-treatment by chemical addition or aeration.



Replacement of zeolite softeners with nanofiltration membranes having reduced rejection efficiencies, to produce partially demineralized water with acceptable hardness and TIC levels.

All of these alternative treatment strategies need additional research to determine the feasibility for smaller water systems such as Indian Hill. The nanofiltration technology has been demonstrated to be viable in some studies in Florida and France to provide corrosion control and good filtration, providing that careful membrane selection and proper post-treatment are done.

Pipe Analysis Information A 4-foot section of 0.5-in (12 cm) ID domestic copper piping was removed for X-ray diffraction analysis when the short period of orthophosphate dosing was suspended in 1998. The exact age of the pipe was unknown, but from indirect information, the minimum age was determined to be at least 10 years old. The house had experienced blue staining in porcelain sinks where faucets dripped, and first draw sampling of 60 mL, 60 mL and 500 mL sequentially yielded copper concentrations of 1.80, 1.02 and 1.74 respectively. Thus, it is considered representative of the types of plumbing situation causing the exceedence of the copper action level. Figure 2 shows a photograph of a location on the pipe interior, with a magnification of 20X. The copper metal pipe surface was covered with a bluish green layer, upon which was a greenish tan layer. Upon optical microscopic examination, the bluish green layer was composed of very small spherical mounds of malachite, approximately 0.2 to 0.5 mm in

4

diameter. When broken open, many revealed small pits underneath, or attachment directly to the shiny copper surface. Overall, the scale layer thickness was much less than 1 mm. In areas of the pipe covered more extensively with the tan layer, the malachite spheres were much smaller and less numerous, more typically 0.02-0.05 mm in diameter. The scale material showed no obvious layering features as has been described in wellperforming duplex films for copper (Ferguson et al., 1996; Feng et al., 1997), but was sampled in two ways, because there had been some exposure to orthophosphate. The dosage, however, was believed to be too low to develop any sort of a significant presence of a crystalline cupric orthophosphate mineral in the scale. First, the scale material was brushed off with a moderately stiff plastic-bristled lab brush, but not much material could be removed. The scale material remaining was then scraped from the pipe with a stainless-steel spatula down to the roughened surface. This revealed a very non-uniformly corroded surface, when the malachite mounds were removed. Both portions of scale material were ground by hand with in an agate mortar, until the material passed through a 200-mesh stainless steel sieve. The scale volume was sufficient to enable the use of a conventional packed-powder mount on using a special zero-background quartz holder*. Samples were analyzed using a Scintag† XDS 2000 powder diffractometer with Copper Kalpha (K I ) radiation, controlled by manufacturer-provided software. Analytical X-ray conditions were 35 kV and 40 mA for the X-rays, over a scan range of 5 to 90 degrees 2-theta (2P), step-scan mode, step length 0.02 degrees, hold time 3 seconds per step. Pattern analysis was done using Jade+ 3.1 software‡ on an IBM-compatible PC, with reference to the 1995 release of the Powder Diffraction File § along the guidelines given by ASTM (ASTM, 1996). Calibration of the 2θ angles was checked using the NIST** 640b Si standard, and was found to produce d-spacings within 0.01 D at 10° and within 0.0002 D at 90°. For the aged copper pipe, a subsample was also taken for coulometric TIC analysis. Figure 3 shows the results of the X-ray diffraction analysis, indicating that the crystalline components of the scale deposit consist almost entirely of Cu2CO3(OH)2 (malachite) and Cu2O (cuprite). On sharp peak at 24.73° 2θ could not be identified, but it does not correspond to any of the common copper solids such as tenorite, langite, posnjakite, brochantite, or atacamite, among others checked. Though the scale appears to be looks uniform and passivating, the film did not control copper dissolution under oxidizing conditions consistent with any of the widely-accepted Gibbs free energy or solubility constant values for malachite reported in the literature (Schock et al., 1994, 1995a). Thus, the sample gave a perfect example of the high alkalinity cuprosolvency phenomenon observed by many water systems across the United States.

*

The Gem Dugout, State College, PA. Thermo -Optek Corporation, Franklin, NJ. ‡ Materials Data Incorporated, Livermore, CA. § International Centre for Diffraction Data, Swarthmore, PA ** National Institute of Standards and Technology, Gaithersburg, MD. †

5

Experimental Plan Four 55-gallon recirculating tank systems with floating lids were used to investigate the solubility behavior of copper ("cuprosolvency") in oxic conditions. These systems were constructed and actually used in published USEPA in-house corrosion experiments (Schock et al., 1995a). All corrosion control test system approaches have pros and cons, and while not exact simulations of household plumbing use patterns, the recirculation systems such as these have considerable value for screening corrosion control treatment approaches and impacts on 90th percentile tap metal levels. Advantages of these systems compared to pipe rigs include: •

Concentration of copper represents “worst case” scenario for cuprosolvency; hence, levels measured under Lead and Copper Rule sampling scenarios of 6-16 hours of stagnation will mostly be below those predicted from these systems, making them a conservative estimator of treatment performance.



The systems are mechanically simpler than pipe loops with timers, flow adjustments, feed manifolds, etc.



Control of critical parameters such as pH, orthophosphate concentration, and DIC can be extremely precise, much more so than with pipe rigs.



The systems use a minimal amount of water, and minimal amounts of labor for sampling and analysis.

One principal disadvantage to these systems is that the development of pipe scales may be slower than in real systems, because they do not cycle through stagnation and flowing conditions. Additionally, these experimental systems do not mimic hydraulic conditions and nonequilibrium corrosion reactions, which might affect copper release of both dissolved, and particulate forms in household systems. There is also an opportunity to accumulate unwanted increased concentrations of stable chlorination and pH adjustment byproducts, such as sodium, chloride and chlorate ions. The systems have recirculating pumps that send water through a 4-foot (122 cm) length of 0.5in (1.3 cm) ID type-L copper pipe with a surface velocity of about 1 ft/sec (0.3 m/sec), representative of actual household use conditions. The tanks and lids are constructed of high-density polyethylene. All plumbing aside from the copper is PVC, nylon, or Tygon™. Sampling ports and chemical feed ports are located on each system so that sampling can be done without exposing the water to the air. Floating o-rings of Tygon3 tubing are used beneath the lids to assure minimal air contact and gas transfer. Figure 4 is a photograph of the systems as installed at the Indian Hill water treatment plant.

6

Unlike the systematic solubility studies conducted in the EPA laboratory, the background water for these tests was the current Indian Hill finished water, rather than carefully controlled synthetic water. Based on a review of historical data with IH staff, a pH of 7.3 was selected as being most representative of the normal treated water pH, and it would form the basis for comparison with three possible treatment changes. To maintain the control of key experimental variables (pH, TIC, chlorine residual) chemical composition would be adjusted when necessary by analytical reagent grade chemicals or the same proprietary zinc orthophosphate chemical contemplated for use by IH Water Works* (CalciQuest Zink-1037, Zn: PO4 ratio = 1:3). The adjustment of pH was accomplished using 6N hydrochloric acid solution or 6 N sodium hydroxide solutions. Sodium bicarbonate was used to adjust TIC, and AR-grade sodium hypochlorite (nominally 5%) solution. All chemical dosing was done with adjustable macropipettes, and exact volumes added were recorded in the lab notebook. Even though the free chlorine residual may affect copper corrosion rates, it is necessary to maintain disinfection, and to approximate distribution system conditions. All tanks were targeted to be the same, starting each week with a free chlorine residual of 1.0 mg/L as Cl2. Dissolved oxygen was monitored approximately semi-weekly to assure that oxic conditions were maintained if the chlorine residual was lost. DO levels rose because of atmospheric exchange, and were somewhat above normal background levels observed in prior studies with IH water. However, in the presence of free chlorine residual, the effect of dissolved oxygen on copper levels in this type of experimental system is likely to be unimportant (Schock et al., 1995a, 1995b; Lytle & Schock, 1996, 1997a). Three possible treatment approaches were tested in parallel with a tank representing the “control” condition, current Indian Hill finished water at pH 7.30. These were: 1. pH adjustment to 7.80 2. Addition of 3 mg/L as PO4 orthophosphate using CalciQuest Zink-1037, at pH 7.30. 3. Removal of approximately 40% of TIC through anion exchange bicarbonate removal, at first pH 7.30, then pH 7.80. Because Indian Hill Water Works already uses zeolite softening of part of the water production for hardness reduction, anion exchange was considered a possible approach to reduce bicarbonate (hence, DIC) that could be integrated into the existing treatment plant without substantial and costly modifications. In order to simulate a full-scale ion-exchange removal process in a realistic manner, a large column was constructed as is shown to the right side of the systems shown in Figure 4. Delays in acquiring the resin and setting up the column caused this experiment to start 34 days after the other three experiments. The column was a plexiglass cylinder, 4-in (10.2 cm) ID, left over from U. S. Environmental Protection Agency filtration projects. Bulkheads and fittings were fabricated for feeding with tap water through a conventional hose, and feeding into the experimental tanks. A small stainless*

CalciQuest, Inc., Belmont, NC.

7

steel screen on a platform was placed in the column over the exit. On top of this, the column was packed with an approximately 1-inch (2.5 cm) thick bed of glass wool at the exit, which was covered with a bed of approximately 2-in (5.1 cm) thickness of 6 mm glass beads. Preliminary sizing calculations were performed using data on the raw water quality, to assure adequate capacity of the anion exchange resin. A 22-inch (56 cm) deep bed of resin was used, consisting of an equal volume mix of Amberlite IRA-410 and IRA-910 strongly basic resins. This corresponded to a volume of 0.16 ft 3. Specifications provided by the resin manufacturer indicated a maximum loading rate for IRA-410 of 2-4 gpm/ft3 (0.22-0.44 L.min-1/m3), and a maximum loading rate of 1-3 gpm/ft3 (0.11-0.33 L.min-1/m3) for IRA-910. At 3 gpm/ft 3, the maximum tap water flow should be 0.48 gpm (1.8 L/min). The resins were prepared in the laboratory by soaking overnight in a 15% solution of analytical reagent grade sodium chloride. The resins were rinsed until the wash water was color and odor free with laboratory deionized water, and then poured into the column. The column was further rinsed with 12 gallons (45 L) of laboratory deionized water, before relocation to the test site. To achieve the target TIC removal finished IH tap water was fed through the column at a rate of 0.4 gpm (1.5 L/min), which was well under the design target of 0.48 gpm (1.8 L/min). The resin treated a total volume of 20 gallons (76 L). The balance was then made up with IH finished water. At the beginning and end of the passage through the resin, effluent water samples were collected for pH, TIC, and complete water chemistries. The initial effluent pH was observed to be 6.05; rising to 6.38 after the 20 gallons had been treated. This contrasts with the influent tap water pH, which held at 7.33 during the ion exchange treatment.

Operational and Analytical Protocols New 0.5-in (1.3 cm) ID type-L copper tubing was installed in each tank. The pipe sections were cleaned similarly to several recommended procedures for preparing copper specimens for weight loss experiments. The pipe interior was first cleaned with a 30-minute soak with 5% Contrad 70 detergent, followed by a thorough deionized water rinse. After that, the pipe section was filled for a short time (approximately 5 minutes) with 10% HCl. After another deionized water rinse, the pipe was finally rinsed with acetone, followed by large amounts of laboratory ultra-pure deionized water. Finally, and the pipe interior was air dried to be free of droplets and staining from oxidation. Following moving the experimental systems to the site, the first three test systems were started on March 19, 1998, and were operated through June 24, 1998. The remaining system, testing the removal of about 40% of the DIC by anion exchange, could not be started until April 23, 1998, and it was finally shut down on July 30, 1998. Upon initial filling, and after adjustment of chlorine residual and pH, background water chemistry analyses were done for the parameters listed in Table 1. Minimal volumes for samples were used: 60 mL high density polyethylene (HDPE) for metals, 250 mL HDPE for anions, and duplicate 30

8

or 40 mL amber glass vials with no headspace for TIC measurements. Sample volumes as necessary were taken for other analytes, such as dissolved oxygen and chlorine residuals. The analytes in Table 1 were monitored weekly using standard analytical procedures employed by other U. S. Environmental Protection Agency corrosion studies (Schock et al., 1995a). Free chlorine residual, temperature, dissolved oxygen and pH was analyzed immediately upon sample collection. The samples for pH were collected in at least triplicate in 40 mL Erlenmeyer flasks with plastic stoppers to exclude air, until analysis. Measurements were done following the “multiple equilibration” procedure with stability monitored using a personal computer interfaced to the pH meter, described in detail elsewhere (Schock et al., 1980; Schock & Schock, 1982; Schock et al., 1994). Following sample collection and analysis of pH and free chlorine residual, adjustments were made to the pH, chlorine residual, or other parameters as necessary (e.g. TIC, orthophosphate). The target for adjustment of pH was ± 0.05 pH unit of the target values, and any times the adjustment was much closer. For free chlorine residual, the target value was 1.0 mg/L ± 0.1. In most cases, follow-up analyses of pH and chlorine residual were done to assure that the final concentrations were within acceptable ranges. No attempt was made to do filtrations of samples to determine the concentration of dissolved copper or other analytes. Periodically, duplicate samples were collected to assess sampling and analytical precision for metals and anions. Major discrepancies in copper values of duplicates would indicate the presence of significant particulate release; though consistent levels of finely dispersed colloidal, copper could not be detected using duplicate samples alone. Duplicate samples revealed insignificant differences in the levels of different constituents. This indicates no problems with large inhomogeneous particulate material. Following stabilization of the single test system testing TIC reduction, pH was adjusted to 7.80 using NaOH, and the experiment was continued for another month, to see if there was a substantial pH effect after the ion exchange bicarbonate removal. Stability of the experiments was judged by consistency of copper levels given precisely controlled pH and TIC values. At the end of study, the copper pipe sections were removed and cut longitudinally with a band saw. The pieces were examined visually and by optical microscopy (Zeiss Model Stemi SV-11, magnifications of 6X- 66X). The amount of surficial scale material was miniscule. Attempts were made to scrape the material that was present, and the removed scale was mounted as a slurry with amyl acetate and evaporated onto a zero-background quartz plate and analyzed XRD using Cu Kα X-radiation at 45 kV and 33 mA. The diffractometer scan was done in step-scan mode, with steps of 0.03° 2θ, and a hold time of 3 seconds per step. Cuprite (Cu2O) was the only mineral observed in the XRD analysis. This was an interesting finding, because while little crystalline scale deposit was expected in the phosphate-treated system, some crystalline scale was expected from the other treatments. Given a similar length of test time, significant scales were always developed on similarly-configured phosphate-free systems tested in the U. S. Environmental Protection Agency laboratory as part of ongoing cuprosolvency research. Therefore, some natural substances appear to 9

retard the formation of malachite or other oxidized crystalline solids during the timeframe of these tests. Interestingly, there is a small amount of orthophosphate present in the treated water (see Table 1), even though none was added.

Experimental Results and Discussion Control of the primary variables, pH, TIC and orthophosphate concentration, was very tight. Graphs of pH and TIC for the different experiments are shown in Figures 5 and Figure 6. After 8 weeks, the copper concentration was still higher than desired for the experiment with ion exchange removal of bicarbonate. Then, the pH was raised to 7.8 to see if it would result in sufficient reduction to be a promising full-scale treatment combination. The tank was not re-filled and re-started; therefore, it is possible that very fine and evenly dispersed colloidal copper corrosion products could remain present, resulting in the relatively poor performance of the supplementary pH increase. The response of copper levels over time for all of the experiments is shown in Figure 7. The initial low concentrations reflect the time it takes for the copper to corrode from the pipe, and fill the volume of water in the tank to the level of equilibrium saturation with the solid or solids passivating the pipe surface. Equilibrium was rapidly achieved in 2-3 weeks with all systems. This may represent a metastable situation, with slow changes taking place over the time frame of months to years, and thus undetectable with pipe rigs, pipe loops or experiments like these in the limited amount of time available for testing (Schock et al., 1994, 1995a, 1995b). However, it does give a good indication of whether or not there could be high concentrations of copper in high-risk copper sites, which are not targeted under the Federal regulation, and provides conservative treatment recommendations. Data for the control and three treatments tested not using orthophosphate are summarized in Figure 8. Data points are plotted in comparison to theoretical predictions using the cuprosolvency model developed previously by U. S. Environmental Protection Agency, computed for an assumed ionic strength of 0.01 M and for a temperature of 25°C (Schock et al., 1995a, 1995b; Schock & Clement, 1998). Considering the range in solubility of metastable copper solids under subtly different aging conditions and background water chemistries, the agreement with the levels and trends predicted is good. Data points represent conditions after the initial stabilization period of approximately 3 weeks for all runs. Raising the pH to 7.8 produced a significant reduction in copper release (about 40%), but the median copper level stayed near 1.8 mg/L. This would not be promising for residences with new construction or remodeling that included copper pipe installation. Reducing the carbonate concentration by ion exchange at the same pH as the control (7.3) succeeded in reducing the copper level (20%), as would be predicted, but not by enough to get below the 1.3 mg/L target. Further increasing the pH to 7.8 caused additional significant copper reduction, for

10

a total of approximately 40% reduction. This was not nearly as much as the theoretical prediction, and may be biased high by not replacing the water in the tank when the pH was increased, possibly allowing the persistence of extremely fine colloidal copper-containing material to bias the copper results. Physical examination of the pipe sections removed from all of the experimental systems showed extremely small amounts of scale present on the pipe surfaces. The films were dark reddish brown, and showed very little visual indication of Cu2CO3(OH)2 (malachite), CuO (tenorite), or Cu(OH)2, the normal passivating minerals on copper pipe exposed to disinfected drinking water. These are shown in the photograph in Figure 9. X-Ray diffraction analyses showed only the crystalline mineral Cu2O (cuprite) present on the surfaces of the specimens from the phosphate addition and pH adjustment experiments. This is consistent with the red color, but it does not explain the high concentrations of copper seen in the water, as cuprite is normally an under-surface material of very low solubility compared to malachite, cupric hydroxide, and tenorite. It is probable that an amorphous cupric hydroxide or cupric orthophosphate material is present on the surface of the pipe and which is controlling the copper level, or it may be crystalline to some degree but present in an undetectably small concentration. Interference with good passivating film growth and evolution of the scale material into malachite or tenorite has been observed for copper pipe exposed to several different natural or treated drinking water matrices in USEPA corrosion control experiments. The chemical mechanism for this has not been positively identified. Cuprite has been the only crystalline solid identified on copper pipe treated with orthophosphate in USEPA laboratory experiments, which helps support the mechanism of operation of the orthophosphate as at least partially functioning as an oxidation inhibitor (Schock et al., 1995a; Lytle & Schock, 1996; Schock & Clement, 1996). Additionally, rapid and drastic reductions in “chlorine demand” of new copper pipes have been observed in these experiments. The most successful of the four alternative treatments in reducing copper release was the dosage of the “zinc orthophosphate” formulation to produce an orthophosphate residual of 3.0-3.5 mg/L as PO4. Although the information provided by the manufacturer for the zinc orthophosphate formulation gaves a Zn:PO4 ratio of 1:10, the observed ratio was 1:8. This is a much higher orthophosphate dosage than was previously applied by Indian Hill and by many other water utilities with lower alkalinities. However, it is reasonably consistent with the predictions of the U. S. Environmental Protection Agency research studies cited earlier. This treatment reduced the median equilibrium copper level by approximately 60%, and the treated water copper concentration stayed below 1.2 mg/L. This should be considered to be approximately a “worst-case” scenario, as the standing times of most home water samples would be much shorter. Additionally, the inhibited oxidation rates of newer copper pipes that has been observed and noted previously should also be a driving force for lower copper levels. One set of published case studies for waters with similar high alkalinities described inconsistent results in terms of 90th percentile copper reductions with blended phosphates (Rezania & Anderl, 1997). In those tests, the orthophosphate residuals were significantly lower than employed in this study, which is also in general agreement with the prior unsuccessful use of a lower dosage of “zinc orthophosphate” by Indian Hill.

11

The pilot study indicated that pH adjustment, alone, would be unlikely to make a substantial improvement in the 90th percentile copper levels long-term, because of the high TIC in the raw water. The spectrum of ages of monitoring sites appeared to corroborate the idea that for whatever reason(s), the formation rate of malachite in the Indian Hill water system was insufficiently rapid to go away in a satisfactory timeframe. In order for the carbonate reduction approach to work, a greater degree of TIC removal than was tested (down to 37 mg/L as C) would be needed. Additional pilot study would be required to determine more precisely how much carbonate reduction would be required, and how to integrate the anion exchange into the overall treatment processes of the plant.

Conclusions Of the treatments tested in the study, the increase in orthophosphate concentration to at least 3 mg/L as PO4 without pH adjustment appeared to be the most likely approach to achieve the meeting of the 1.3 mg/L Action Level for the 90th percentile of monitoring samples. Previous experience, as discussed in the cited references, suggested that copper levels in shorter standing time samples, such as 6-9 hours, would likely be lower than the concentrations seen in the recirculation experiments, providing that chlorine residuals and some dissolved oxygen are maintained throughout the distribution system. The normal treated water pH for IHWW naturally fits the range expected for nearly optimal cuprosolvency reduction by orthophosphate based on calculations of chemical equilibria. The two principal advantages of the high orthophosphate dosing were thought to be that it should result in rapid decreases in copper levels for “overnight” standing samples, and that appropriate chemical feed equipment already existed at the IHWW treatment plant. Prior experience also suggested that red water or other aesthetic complaints had not occurred at IHWW before with prior periods of the introduction and use of the orthophosphate chemical. There were three potential disadvantages of the orthophosphate addition considered while deliberating on the selection of the best approach to Lead and Copper Rule compliance. First, unlike the combination of pH increase and TIC control, most published studies or studies presented at conferences have not indicated that the copper solids formed on the pipe surface undergo ‘aging’ affects resulting in substantially lower solubilities over several years of treatment. So, copper levels in houses and wastewater plant loadings of zinc and copper will likely stay relatively consistent, and this may be considered to be too high at a later time. Little optimization of this approach seems to be possible, especially when balancing the control of lead from solders and faucets, which needs a higher pH than copper does for the orthophosphate to work. Third, there may be scaling issues on heat-transfer surfaces of commercial customers, and possibly more fouling in hot water heaters. Those kinds of effects are very hard to predict, and even harder to estimate precisely. Figures 12 and 13 represent a conceptual model for how the dosage of orthophosphate affects the natural aging of scales in copper pipes under oxic conditions (Schock et al., 1995a; Schock &

12

Clement, 1998; Edwards et al., 2000; Schock et al., 2000). In Figure 12, the starting point for the equilibrium copper concentration is the initial formation of cupric hydroxide on new pipe. Because the pH is still below the transition point for CuO (tenorite) stability, Cu2CO3(OH)2 (malachite) would be the endproduct scale material. After years of aging, wherein the high surface area hydroxide can slowly react with carbonate in the water, a scale dominated by malachite slowly develops. If the scale thickness becomes extensive enough that a redox potential gradient can form to enable a uniform and finely-crystalline Cu2O (cuprite) layer to develop underneath the porous malachite surface layer, ultimately the pipe surface will become passivated. In high DIC waters, rapid oxidation of the copper metal and solubilization by carbonate and bicarbonate complexation interferes with the rapid formation of the cuprite layer. In spite of thick surficial deposits, corrosion continues, and the copper levels in the water are still controlled by cupric hydroxide equilibrium and precipitation kinetics. When orthophosphate is added at this pH (Figure 13), a thin film of a cupric orthophosphate compound is formed on the surface, which has been shown in laboratory and pilot tests to act rather consistently with a solubility and stoichiometry of Cu3(PO4)2•2H2O. Oxidation is retarded, and a thin film of cuprite develops under the phosphate scale. Evolution of the scale from cupric hydroxide control to malachite control is prevented, or at least is significantly prolonged. Because the solubility of this cupric orthophosphate scale is lower than cupric hydroxide (at this pH), there will be a beneficial rapid decrease in dissolved copper levels. However, over tens of years of water exposure the final copper concentration in equilibrium with the pipe scale would not be expected to be as low as would be the case if a duplex cuprite/malachite scale actually developed. Following the completion of this pilot study, IHWW began feeding orthophosphate into the system again at the start of 1999, with the concentration of the chemical increased incrementally by approximately 0.5 mg/L every one or two weeks until the target dosage of 3 mg/L as PO4 was achieved. No abnormal adverse complaints, such as red or turbid water, were received from consumers during the resumption of the orthophosphate dosing. In fact, fewer complaints of blue porcelain staining, blue water, and even bluish discoloration of hair were received after implementing the new orthophosphate treatment program. One minor difficulty observed was that the increased phosphate levels can cause additional algal growths in home aquaria or open decorative water systems, so other utilities should be aware of such effects and provide sufficient warnings to consumers. Lead and Copper Rule monitoring samples were taken in October of 1999 and in April and May of 2000, for which the results are shown in Figures 10 and 11. As expected from the pilot study results, the orthophosphate dosing continued to maintain excellent lead control, and also immediately brought the copper corrosion byproduct release below the 90th percentile Action Level. In September of 1999 the lead levels were below 0.005 mg/l for the 90th percentile, and the 90th percentile copper levels were reduced to 1.04 mg/L. The 90th percentile levels for lead and copper were virtually identical for the Spring 2000 sampling round. Interestingly, the sampling results seem to bear out the cuprosolvency model projections of consistency of the copper release and slowing of the “aging” process of copper scales. Ohio EPA had informed IHWW that they have now successfully met simultaneous lead and copper control requirements of the regulations, and they can move to “reduced monitoring” status. 13

Acknowledgements The authors wish to thank the staff of Indian Hill Water works who assisted with installation of the test system and who also allowed use of some of their lab facility supplies and space to support the study. Leo Fichter of USEPA, National Risk Management Research Laboratory, Technology Transfer Support Division fabricated, transported and installed the experimental systems. Keith Kelty and James Doerger of the Water Supply and Water Resources Division of USEPA did analyses of most major water constituents not done on location, except for chloride, alkalinity, and TIC, which were done by TN & Associates on contract. Stephen Harmon of TN & A did the optical microscopic and X-ray diffraction analyses of the experimental pipe sections. Tom Sorg of WSWRD also helped with sample collection.

Disclaimer Mention of specific trade names or products is for descriptive purposes and does not constitute official endorsement of Indian Hill Water Works or of the U. S. Environmental Protection Agency. Opinions expressed are those of the authors, and do not reflect official policy of the United States Government or of the State of Ohio.

14

References ASTM (American Society for Testing and Materials) 1996. Standard Practices for Identification of Crystalline Compounds in Water-Formed Deposits by X-Ray Diffraction, D 934-80, Conshohocken, PA. Edwards, M., Meyer, T.E. & Schock, M.R., 1996. Alkalinity, pH and Copper Corrosion By-Product Release. Jour. AWWA, 88:3:81. Edwards, M. et al, 2000. Role of Pipe Aging in Copper corrosion By-Product Release, Workshop on Pipe Material Selection for Drinking Water Systems, Sustainable Drinking Water Distribution Management, September, Goteborg, Sweden. Feng, Y. et al, 1997. Corrosion Mechanisms and Products of Copper in Aqueous Solutions at Various pH Values. NACE Corrosion, 53:5:389. Ferguson, J.L., von Franqué, O. & Schock, M.R., 1996 (Second ed.). Corrosion of Copper in Potable Water Systems. Internal Corrosion of Water Distribution Systems, pp. 231-268. AWWA Research Foundation/DVGW-TZW, Denver, Colorado. Hidmi, L., Edwards, M. & Schock, M.R., 1999. Impact of Anions on Copper Solubility and Corrosion By-Product Release, Proc. AWWA Annual Conference, June 20-24, Chicago, IL. Indian Hill Water Works, 1994. Desk Top Study, Treatment Recommendations, Village of Indian Hill, Ohio. Lytle, D.A. & Schock, M.R., 1996. Stagnation Time, Composition, pH and Orthophosphate Effects on Metal Leaching from Brass, EPA/600/R-96/103, Office of Research and Development, Washington, DC. Lytle, D.A. & Schock, M.R., 1997a. Impact of Stagnation Time on the Dissolution of Metal from Plumbing Materials, Proc. AWWA Annual Conference, June 15-19, Atlanta, GA. Lytle, D.A. & Schock, M.R., 1997b. An Investigation of the Impact of Alloy Composition and pH on the Corrosion of Brass in Drinking Water. Adv. Environ. Res., 1:2:1,. Lytle, D.A., Schock, M.R. & Sorg, T.J., 1996. Controlling Lead Corrosion in the Drinking Water of a Building by Orthophosphate and Silicate Treatment. Jour. NEWWA, 110:3:202.

15

Rezania, L.-i.W. & Anderl, W.H., 1997. Corrosion Control for High DIC Groundwater; Phosphate or Bust, Proc. AWWA Annual Conference, pp. 167-181, June 15-19, Atlanta, GA. Schock, M.R., 1999 (Fifth ed.). Internal Corrosion and Deposition Control. Water Quality and Treatment: A Handbook of Community Water Supplies, pp. 17.01-17.109. McGraw-Hill, Inc., New York. Schock, M.R., Buelow, R.W. & Mueller, W., 1980. Laboratory Techniques for Measurement of pH for Corrosion Control Studies and Water not in Equilibrium with the Atmosphere. Jour. AWWA, 72:5:304. Schock, M.R. & Clement, J.A., 1996. Lead and Copper Control with Non-Zinc Orthophosphate, Proc. National Conference on Integrating Corrosion Control and Other Water Quality Goals, New England Water Works Association, U. S. EPA-New England, and the AWWARF, May 19-21, Cambridge, MA. Schock, M.R. & Clement, J.A., 1998. Control of Lead and Copper with Non-zinc Orthophosphate. Jour. NEWWA, 112:1:20. Schock, M.R. et al, 2000. The Chemistry of New Copper Plumbing, Proc. AWWA Water Quality Technology Conference, November 5-9, Salt Lake City, UT. Schock, M.R., Lytle, D.A. & Clement, J.A., 1994. Modeling Issues of Copper Solubility in Drinking Water. In J.N. Ryan & M. Edwards (eds.), Proc. ASCE National Conference on Environmental Engineering, pp. 17-25, ASCE, July 11-13, Boulder, CO. Schock, M.R., Lytle, D.A. & Clement, J.A., 1995a. Effect of pH, DIC, Orthophosphate and Sulfate on Drinking Water Cuprosolvency, EPA/600/R-95/085, Office of Research and Development, Cincinnati, OH,. Schock, M.R., Lytle, D.A. & Clement, J.A., 1995b. Effects of pH, Carbonate, Orthophosphate and Redox Potential on Cuprosolvency, NACE Corrosion/95, March 26-31, Orlando, FL. Schock, M.R. & Schock, S.C., 1982. Effect of Container Type on pH and Alkalinity Stability. Water Res., 16:1455. USEPA, 1991. Maximum Contaminant Level Goals and National Primary Drinking Water Regulations for Lead and Copper, Final Rule, CFR 40 Parts 141 and 142, Federal Register, 56:110:26460. USEPA, 2000. National Primary Drinking Water Regulations for Lead and Copper, Final Rule, CFR 40 Parts 141 and 142, Federal Register, 65:8:1949.

16

Table 1. Chemical analyses performed during the pilot testing. Most analyses were performed weekly, except for dissolved oxygen.

Control

pH Adjusted to 7.8

Standard N Deviation

Standard N Deviation

Zinc Orthophosphate pH 7.3 Standard N Deviation

Partial Anion Exchange pH raised from 7.3 to 7.8

Partial Anion Exchange pH 7.3 Standard N Deviation

Standard N Deviation

Parameter (mg/L unless otherwise noted)

Mean

Ca

35.4

1.5

14

35.1

1.2

15

35.1

1.5

15

32.4

0.61

8

31.8

0.63

7

Cl Free Cl2 Cu F K Mg Na (NO3+ NO2)-N pH, units Ortho-PO4 SiO2 SO4 Total Alkalinity as CaCO3 Temperature °C TIC, as C Zn DO

139 0.53 2.79 0.86 2.07 10.6 155 2.13 7.33 0.13 7.2 46

13 0.17 0.34 0.04 0.13 0.24 8.7 0.05 0.03 0.05 0.28 1.4

14 14 14 14 12 14 14 14 14 14 14 14

126 0.56 1.7 0.86 2.08 10.5 159 2.13 7.82 0.21 7.2 46

9.8 0.19 0.31 0.04 0.12 0.21 8 0.05 0.04 0.02 0.27 1.3

15 15 15 15 13 15 15 15 15 15 15 15

127 0.58 1.13 0.86 2.06 10. 6 152 2.14 7.3 3.3 7.2 46

7.0 0.19 0.19 0.05 0.12 0.21 7.1 0.05 0.02 0.17 0.31 1.5

15 15 15 15 13 15 15 15 15 15 15 15

192 0.4 1.95 0.91 1.81 9.1 142 0.9 7.32 0.07 7.5 21.0

6.1 0.3 0.8 0.03 0.06 0.21 3.5 0.02 0.01 0.04 0.09 0.68

8 7 8 8 8 8 8 8 8 8 8 8

209 0.22 1.73 0.94 NA 9.2 161 0.91 7.80 0.07 7.47 20.8

5.2 0.1 0.06 0.02 -0.08 5.4 0.03 0.02 0.05 0.04 0.33

7 7 7 7 -7 7 7 7 7 7 7

236

7.6

14

258

4.5

15

240

4.9

14

143

0.75

8

152

0.77

7

18.4 61.2 0.01 7.9

1.2 1.5 0 0.32

14 14 14 6

18.2 63.3 0.01 8.5

1.4 1.3 0.01 0.19

15 15 15 6

18.2 63.1 0.42 7. 8

1.4 1.1 0.02 0.09

15 15 14 6

17.7 37.44 0.01 7.4

2.2 0.33 0 0.21

8 8 8 3

21.2 37.2 0.02 7.6

1.0 0.19 0 0.11

7 7 7 3

Mean

Mean

Mean

Mean

Captions Figure 1. Schematic diagram of the Indian Hill Water Works treatment processes. Figure 2. Stereomicroscope photo of surface of household copper pipe sample, 20X magnification. Malachite spherules are approximately 0.2-0.5 mm diameter in this photograph. Figure 3. XRD pattern of household copper pipe sample, showing scale is essentially entirely composed of Cu2 CO3 (OH)2 (malachite) mixed with a small amount of cuprous oxide (Cu2 O) and quartz (SiO 2 ). Figure 4. Experimental systems to compare cuprosolvency mitigation treatments, set up in the IHWW treatment plant. Figure 5. Control of TIC in the experimental systems during pilot testing. Figure 6. Control of pH in the experimental pilot test systems. Figure 7. Copper levels in experimental systems during pilot testing. Figure 8. Comparison of interior pipe surfaces after removal from pilot test systems. Figure 9. Comparison of experimental to cuprosolvency model data for control and three treatment tests. Figure 10. Comparison of lead monitoring data from before the treatment change (1997) to after the addition of 3 mg/L as (PO4 ) orthophosphate. Figure 11. Comparison of copper monitoring data from before the treatment change (1997) to after the addition of 3 mg/L as (PO4 ) orthophosphate. Figure 12. Conceptual application of the cupric hydroxide model for cuprosolvency in the absence of orthophosphate dosing. Figure 13. Conceptual application of the cupric hydroxide model for cuprosolvency in the presence of orthophosphate.

Evolution of Scale Model for High DIC, Low pH 100

mg Cu/L

10

Cu(OH)2 Fresh Scale CuO

1

0.1

0.01 6.0

Ideal

Cu ( O 2 H) CO 2 3 Film

7.0

pH

Aging Process (in theory): vRecrystallizing Recrystallizing vDecreasing Decreasing surface area vReacting Reacting with CO3 or HCO3vCan Can take 20, 30 or more years with high DIC

8.0

Orthophosphate Effect on Scale Evolution at High DIC 100

Cu(OH)2 Fresh Scale

mg Cu/L

10

1

Cu3(PO4)3ž2 2 H2O

0.1

0.01 6.0

Ideal

Aging Process is Impeded: vSlows Slows oxidation vPrevents Prevents or drastically slows reaction with CO3 or HCO3vImmediate Immediate benefit

Cu ( O 2 H) CO 2 3 Film

7.0

pH

8.0