Corrosion Science 42 (2000) 1801±1822 www.elsevier.com/locate/corsci Copper corrosion in distribution systems: evaluation of a homogeneous Cu2O ®lm and a natural corrosion scale as corrosion inhibitors Abhijit Palit a, Simo O. Pehkonen b a Environmental Engineer Analyst, The Cadmus Group Inc., 1901 N. Fort Myer Drive, Suite 1016, Arlington, VA 22201, USA b Department of Chemical and Environmental Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260 Received 4 August 1999; accepted 20 January 2000 Abstract The aim of this work is to assess the performance of di€erent corrosion scales as e€ective copper corrosion inhibitors using synthetic waters and distribution system waters. The major objective of the study was to evaluate the stability of an arti®cially synthesized Cu2O ®lm and naturally formed heterogeneous corrosion scales under the in¯uence of a number of synthetic and real waters, using Electrochemical Impedance Spectroscopy (EIS). Cu2O coatings were synthesized by using a special ¯ame aerosol deposition process. Based upon the EIS data, conclusions regarding the stability of the di€erent scales under typical water quality conditions, (which are commonly encountered in water distribution systems), are drawn. 7 2000 Elsevier Science Ltd. All rights reserved. Keywords: Copper; EIS 1. Introduction The promulgation of the ``Lead and Copper Rule'' by the USEPA in 1991 led to the reduction in the action level for copper to 1.3 mg/l in ®nished drinking waters. This forced nearly 80,000 water utilities nationwide to monitor for copper, which arises mostly from the uniform corrosion of copper plumbing materials [1], and initiated signi®cant research to control copper corrosion in distribution 0010-938X/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 1 0 - 9 3 8 X ( 0 0 ) 0 0 0 2 4 - X 1802 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 systems. Regulatory monitoring data have been of limited use in the understanding of copper chemistry in water conduits. Some of the major reasons for this are: . Concentration data of some important background chemical constituents that might in¯uence copper corrosion (e.g., chloride, sulfate, silica, orthophosphate etc.) may not have been necessarily collected. . Although copper dissolution rates are highly dependent upon the age of the passivating ®lms, adjustments for di€erent plumbing ages may not have been conducted, which is very important to correctly deduce di€erences in cuprosolvency resulting from chemical, rather than temporal factors. . The collected water quality data may not be representative of the entire distribution system. Much research has been carried out in the past several years in the area of copper corrosion. Both, weight loss methods and electrochemical techniques have been utilized to assess the rate of copper corrosion. Feng et al. [2,3] have recently conducted very thorough experiments in order to understand the rate limiting processes of copper corrosion. They have carried out these studies by utilizing a sophisticated spectroscopic tool of X-ray Photon Spectroscopy (XPS) for the ®lm characterization as well as several electrochemical techniques in order to gain insight into the important elementary processes controlling copper corrosion. The processes include charge transfer, di€usion of copper ions through copper oxide ®lm and di€usion of copper ions in solution. Feng et al. [2,3] concluded that under their experimental conditions, the rate limiting process (which controls the overall corrosion rate) was the di€usion of copper ions through the oxide ®lm. Feng et al. [2,3] conducted Electrochmical Impedence Spectroscopy (EIS) and XPS studies to understand the e€ect of pH, immersion time and rotation speed of the electrode on copper corrosion and ®lm growth. Although much insight into copper corrosion can be obtained from the detailed experiments and discussion by Feng et al. [2,3], their studies are limited in terms of the di€erent water quality conditions under which corrosion ®lm development can occur in real distribution systems. The only water quality parameter, which was varied in their studies, was the pH. However, it must be stated that the focus of the study by Feng et al. [2,3] was not on the e€ects of numerous water quality parameters on the ®lm development and the rate limiting processes of copper corrosion. Copper may exist in water in the monovalent (cuprous) or divalent (cupric) states. In potable waters, copper undergoes the following electrochemical transformations: a† Cu s† $ Cu‡ ‡ e ÿ ; ‡ b † Cu $ Cu 2‡ ‡ e ÿ ; E 0 ˆ ÿ0:52 V E 0 ˆ 0:16 V Due to the potentials for reactions (a) and (b) being much smaller than that of the oxygen reduction reaction (O2(g) + 4H+ + 4eÿ t 2H2O; E 0 = 1.23 V), A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1803 copper pipes carrying water with dissolved oxygen continue to corrode until all the oxygen is depleted or a precipitated copper oxide ®lm arrests the rate of corrosion. There is some evidence that the transformation from Cu+ to Cu2+ is the rate limiting factor during copper corrosion, with Cu+ existing in reversible equilibrium with Cu metal at the pipe surface [4]. In drinking waters, the predominant electron acceptors (i.e., oxidants for copper) are dissolved oxygen and aqueous chlorine or chloramine species [5±7]. Some progress has been made in understanding the copper chemistry as far as inorganic components in drinking waters are concerned [8]. It is now widely believed that inorganic carbon species (i.e., bicarbonate, carbonate and dissolved carbon dioxide), ammonia, sulfate and chloride are the main inorganic species a€ecting the chemistry of copper corrosion. Evidence exists to suggest that although chloride is very aggressive initially, it has long-term bene®cial e€ects and may also counteract the deleterious e€ects of bicarbonate under certain conditions [9]. Long-term copper levels have been observed to be governed by the precipitation of solid phases like cupric hydroxide Cu(OH)2, malachite (Cu2(OH)2CO3), tenorite (Cu2O) etc., each one being in the most thermodynamically stable form at speci®c pH and DIC (dissolved inorganic carbon) ranges [10]. However, a thorough understanding of the mechanisms of copper corrosion as a function of chloride and DIC dosages at di€erent pH and background water quality conditions is lacking. 1.1. Copper (I) chemistry Cuprous ion forms a weak and unstable hydrolysis species (CuOH0). On account of this and the diculties in detecting and quantifying it, this species is considered quite irrelevant in most reviews of copper chemistry [11,12]. The 0 formation of CuCOÿ 3 and CuHCO3 has been hypothesized to explain some trends in oxidation rates of Cu(I) by dissolved oxygen. However, their kinetic and thermodynamic data are not well documented [13]. Chloride can react with the cuprous ion to form CuCl, which subsequently hydrolyzes to form Cu2O(s) [14]. Studies utilizing X-ray di€raction and other surface analysis techniques support the formation of a Cu2O(s) scale in the presence of chloride under various water quality conditions [15,16]. The aqueous chemistry of cuprous ion may also be dominated by the formation of several stabilizing complexes with the NH3(aq) ligand. The cuprous ammine complexes can be formed directly or by a reduction of cupric ammine complexes by a reaction such as [17]:   2‡ ‡  ‡Cu s† 4 2 Cu NH3 †2 Cu NH3 †4 1.2. Copper (II) chemistry The most important Cu(II) species governing its solubility are the hydrolysis species: Cu(OH)+, Cu(OH)02, Cu(OH)ÿ 3 etc. The stability constants of these species 1804 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 are quite unreliable and large di€erences exist between the reported values. Baes and Mesmer [12] have recommended both upper and lower limits for these complexes, rather than precise values. Carbonate complexation dominates Cu(II) solubility in most drinking waters at pH values greater than 7.3 [1,15,18]. The hydroxy-carbonate complex, CuCO3(OH)2ÿ 2 , becomes important in enhancing Cu(OH)2(s) solubility at pH values greater than 9.0. The prediction of increased solubility of copper(II) by waters of increasing DIC is supported by the results of many laboratory and ®eld studies [18ÿ21]. However, according to Schock et al. [1], DIC regulates cuprosolvency in conjunction with pH and the dominant solid phase under the corresponding exposure conditions. For example, malachite formation would enable attainment of copper levels less than 1.3 mg/l (i.e., the USEPA action level) after stagnation, below a pH of 6.5 for all DIC levels. The quantitative prediction of copper levels in drinking waters relies heavily on the solubility and physical properties of cupric oxide, hydroxide and basic carbonate solids, which comprise most scales in drinking water supplies. Due to considerable uncertainty in the solubility constants of all these minerals, the tendency for prolonged existence of thermodynamically metastable phases like Cu(OH)2(s) and the possible slow rate of formation of Cu2(OH)2CO3(s) (malachite), quanti®cation becomes dicult. The di€erences among several solubility constants for CuO(s) and Cu(OH)2(s) have been discussed by de Zoubov et al. [11], who selected two constants, which they felt agreed best with actual copper solubility data. Adeloju and Hughes [10] suggest that the formation of Cu(OH)2(s) is kinetically favored over the formation of the CuO(s) solid. Many researchers have observed signi®cant Cu(OH)2(s) being produced by corrosion of copper at anodic locations in electrochemical experiments [22,23]. These experiments, along with the copper levels observed in many sampling programs and pipe rig systems, support the argument that for relatively new plumbing systems, the use of Cu(OH)2(s) in the solubility models is more realistic than the use of CuO(s). Thus, although Cu2O(s) was the oxide of choice in the current study, additional studies with some of the aforementioned Cu(II) solid phases can be carried out. This allows a better understanding of the stability of the various dominant copper oxide species under di€erent water quality conditions. The current study attempts to build on the earlier ®ndings and relate the stability of the copper corrosion ®lms to special water quality parameters (i.e., pH, DIC and chloride concentrations etc.) and discusses the probable chemical processes that could explain the observed EIS results. The objective of the present study is to evaluate the stability of an arti®cially synthesized, relatively homogeneous, cuprous oxide (Cu2O) ®lm on a fresh copper coupon using synthetic and real ambient waters. Furthermore, it is our objective to compare the performance of the Cu2O ®lm to naturally formed corrosion scale as copper corrosion inhibitor and its capacity to protect the underlying copper metal. EIS was used to study the stability of the ®lms under di€erent water quality conditions. The synthetic waters used in the second phase of the study consisted of several combinations of typical distribution system chloride and DIC concentrations at pH values of 7.0 and 8.5. Finally, the performance and stability A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1805 of the arti®cially synthesized cuprous oxide ®lm was compared to that of naturally scaled copper coupons under typical distribution system water qualities. 2. Experimental 2.1. Electrochemical (DC) measurements Corrosion currents were used to quantify the instantaneous corrosion rates. To estimate the three key parameters: the corrosion currents, the corrosion potentials and the Tafel slopes, potentiodynamic polarization scans were carried out using a Gamry AC/DC corrosion testing machine with the CMS 100 software. A threeelectrode con®guration was adopted with a platinum auxiliary electrode and a Ag/ AgCl reference electrode. A summary of the parameter settings for the potentiodynamic polarization scans is given in Table 1. For every set of measurements to estimate the Tafel parameters by potentiodynamic polarization scans, the average values obtained from three concurrent scans (not di€ering from one another by more than 10%) are reported. As an additional quality control measure for selected samples, three polarization resistance scans were conducted as well to estimate Rp (polarization resistance). The Tafel slopes ba and bc † were calculated from the potentiodynamic scans using commonly used equations. Substituting the values of Rp (from the polarization resistance scans) and ba and bc into the expression Rp ˆ ba bc = 2:3  Icorr  ba ‡ bc ††, yielded an independent estimate of Icorr. If this estimate of Icorr was not within 20% of that obtained from the potentiodynamic scans, all the scans in that batch were repeated until all quality control requirements (i.e., within 20%) were satis®ed. The Rp values from the Tafel scans were compared to those from the EIS measurements (discussed later), as an additional quality control measure. 2.2. Synthesis of Cu2O coatings A ¯ame aerosol generation and deposition technique was used for the synthesis Table 1 Parameter settings for potentiodynamic scans Parameter Value Voltage scan range ÿ250 mV to +250 mV relative to open circuit voltage (OCV) 1 mV/s 2s 0.78 cm2 15 s O€ 600 s Scan rate Sampling time Sample area Auto-ranging time (delay) Conditioning Total scan time 1806 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 of the Cu2O coating. The apparatus consists of a honeycomb premixed ¯ame burner, the detailed design parameters of which is described elsewhere [24,25]. It has features of a uniform temperature pro®le over the burner cross-section and a steady one-dimensional ¯ow ®eld. Fig. 1 is a simpli®ed schematic of the premixed ¯ame aerosol reactor system. Particle-free dry oxygen (RH = 0%) was premixed with methane and subsequently ignited by the burner to produce the ¯ame. In some of the initial trials, the air was bubbled through heated copper acetate solution (temperature maintained at 508C) before being premixed with methane in order to provide adequate copper precursor (in the form of air-stripped copper acetate) for the formation of the cuprous oxide aerosol. However, this step was found to be unnecessary since the copper substrate (coupon) itself could be used as a precursor material. As a result, during the subsequent syntheses, this step was eliminated. The operating conditions that had to be optimized during the syntheses were: CH4 and O2 ¯ow rates, ¯ame temperature, actual ¯ame jet initial diameter, the distance of the coupon from the ¯ame and the number and the speed of passes of the copper substrate over the ¯ame. Identi®cation of the oxide coating was conducted by X-ray di€raction using a Siemens D500 Di€ractometer. A con®rmation of the identity and elemental composition of the oxide ®lm was obtained by conducting an SEM analysis of the ¯ame prepared Cu2O samples using a Hitachi S-4000 SEM with EDAX capability. The standardization of the cuprous oxide coating synthesis procedure resulted in excellent reproducibility of the ®lm morphology and thickness. 2.3. Electrochemical Impedance Spectroscopy (EIS) EIS is a technique to analyze the response of corroding electrodes to small amplitude alternating potential signals of widely varying frequency. The equivalent circuit for the oxide coated metal surface is shown in Fig. 2. Using the CMS 100 application software, reasonable initial values for the circuit components were plugged in and a number of iterations performed in order to get a close ®t Fig. 1. A simpli®ed schematic of the pre-mixed ¯ame aerosol reactor system. A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1807 between data points and the curves logjZj and y versus log frequency). If a good ®t was not obtained, then additional circuit components known as Warburg impedances would be added. These components simulate the di€usion-limited processes. From the Nyquist plot data one can conclude that most of the corrosion reactions were not under di€usion control. Fig. 3 shows a typical Nyquist plot. The stability of the coating is considered to be poor if Cdl increases with time along with a corresponding decrease in Rpo. EIS measurements were also carried out on fresh copper surfaces and naturally scaled copper coupons, after exposure to test solutions for 0 and 72 h. The naturally scaled coupons (after 10 weeks of exposure to synthetic and real waters) were tested using Cincinnati tap water and ozonated/bio-®ltered water from Lake Blu€ (IL) water treatment plant as electrolytes. The impedance of the equivalent circuit, Z o†, may be expressed in terms of real Z 0 o†† and imaginary Z 00 o†† components as follows: Z o† ˆ Z 0 o† ‡ Z 00 o† A summary of the parameter settings for the EIS scans is given in Table 2. The impedance behavior of the copper electrode was expressed as a Bode plot logjZj versus log o†). The EIS spectrum data was then ®tted and an estimate of the pertinent resistances and capacitances obtained. Fig. 2. An electrical circuit analog of the EIS system. 1808 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 Fig. 3. A typical Nyquist plot for fresh copper surface (t = 0 h, pH = 8.5, DIC = 4.9 mg/l, chloride = 25 mg/l), one time constant. 3. Results and discussion The purpose of this study was to investigate the e€ectiveness of an arti®cially synthesized Cu2O(s) ®lm in arresting the copper corrosion process under speci®c water quality conditions. The rationale behind this is to eventually isolate all the identi®ed oxide phases from the heterogeneous natural pipe scale and to quantify the e€ectiveness of each of these oxide phases as copper corrosion inhibitors. After the identi®cation of the best corrosion inhibiting copper oxide, this information could then be used to control the water quality, the temperature and Table 2 Parameter settings for EIS scans Parameter Value AC voltage amplitude 10 mV for a fresh electrode 20 mV for a coated electrode 0 mV vs. open circuit potential 5 5000±0.005 Hz 5.0 cm2 8.3 g/cm3 31.5 g/eq 100 s DC voltage Points per decade Frequency range Sample area Sample density Equivalent weight Initial delay A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1809 pipe hydrodynamics, in order to favor the deposition of a speci®c copper scale that is the best copper corrosion inhibitor under a given set of conditions. In the current study, the e€ects of chloride and DIC concentrations at pH values 7 and 8.5 were investigated. The results from synthetic Cu2O were compared to those from naturally scaled copper surfaces under controlled water quality conditions. The performance of the naturally scaled and arti®cially coated surfaces under the in¯uence of real waters (i.e., ozonated/bio-®ltered Lake Michigan water and Cincinnati tap water) was also investigated and compared to the results obtained from the synthetic waters. EIS measurements were carried out initially and after 72 h of exposure, to study the stability of the ®lms and evaluate the development of corrosion scale over the fresh copper surfaces. For the synthesis of the Cu2O coating, several trials had to be carried out to establish the optimum operating parameters of the ¯ame aerosol reactor. X-ray di€raction studies indicated that the ®nal coatings were not contaminated with unwanted oxide phases such as Cu(II) oxides, which could easily form under oxygen-rich ¯ame conditions. To con®rm the XRD ®ndings, SEM (magni®cation of 25,000 and electron beam accelerated at 20 kV) and EDAX analyses were conducted on the coated surfaces to ascertain the morphology and elemental composition of the coatings (Figs. 4±6)). The SEM analyses involved several sample preparation steps, including sputtering the sample with a 50 nm thick gold Fig. 4. An SEM photograph (25,000 magni®cation) of the arti®cially synthesized cuprous oxide coating. 1810 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 layer, the details of which are described elsewhere [26,27]. The copper to oxygen ratios for the samples (atomic percentages) were in the vicinity of 8:1, indicating that the only copper oxide present in the coating was Cu2O. EDAX analyses were conducted at four or ®ve di€erent locations of the sample, and the average of the atomic ratio values was reported. Occasionally, some of the samples were found to be contaminated with traces of zinc and silicon (<2%). However, this was not believed to be signi®cant enough to a€ect our results. To estimate the thickness of the coatings, EDAX analyses were conducted along the vertical cross-section of the samples. By observing a change in the elemental composition along the vertical cross-section and using the scale on the SEM micrograph, the thickness of the coatings was estimated. For each sample, the coating thickness was measured at three di€erent locations and the average value was reported. Under the operating conditions of the aerosol reactor, the Cu2O coating thickness in the copper samples was found to vary between 0.70 and 0.92 mm (i.e., the reproducibility of the coating thickness was quite good). Moreover, the thickness Fig. 5. An EDAX spectrum of the cuprous oxide coating. A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1811 Fig. 6. An SEM photograph (11,000 magni®cation) of the vertical surface of the cuprous oxide coated copper coupon. EDX analyses were conducted to determine the coating thickness. of the naturally formed scales (i.e., upon exposure of the copper coupons to simulated drinking water solutions: Scale A1 of 0.88 mm thickness and Scale B2 of 0.71 mm thickness) matched closely with those of the ¯ame synthesized Cu2O coatings. Synthetic solutions containing di€erent combinations of chloride and Ct concentrations (Table 3) were used for the EIS experiments. Typical Bode plots are shown in Fig. 7 (a single time constant) and Fig. 8 (two time constants). The following parameters were determined from the Bode plot using appropriate software for parameter ®tting (CMS 100) and techniques discussed elsewhere [28]: 1. 2. 3. 4. 5. 1 coating capacitance, CC electrical double layer capacitance, Cdl polarization resistance, Rp coating resistance, Rpo resistance of the solution, Rs Scale A: Scale developed after 10 weeks of exposure to a pH 7.0 solution with Ct = 250 mg/l (DIC = 49 mg/l) and Clÿ = 25 mg/l. 2 Scale B: Scale developed after 10 weeks of exposure to ozonated/bio-®ltered Lake Michigan water whose pH was adjusted to 7. 1812 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 Table 3 The composition of synthetic solutions used for EIS experiments ÿ Solution pH Ct = [COÿ2 3 ] + [HCO3 ] + [H2CO3] no. (mg/l) Clÿ Na+ Bu€er (anion only) (cation only) (mg/l) (mg/l) 1 2 3 4 5 6 7 8 125 25 125 25 125 25 125 25 7.0 7.0 7.0 7.0 8.5 8.5 8.5 8.5 a 250 250 25 25 250 250 25 25 175.3 110.5 90.4 25.6 175.3 110.5 90.4 25.6 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 (M) (M) (M) (M) (M) (M) (M) (M) MOPSa MOPS MOPS MOPS boric acid boric acid boric acid boric acid MOPS refers to 4-morpholinepropanesulfonic acid (pKa = 7.2). In the ®eld of corrosion and coating performance science and engineering, a coating is considered to be performing satisfactorily if it has high resistance and low capacitance, even after prolonged exposure to the corroding medium. If the coating breaks down under exposure, it may result in the entrainment of water within its structure. As a result, its capacitance goes up because water has a very Fig. 7. A Bode plot for fresh copper surface (t = 0 h, pH = 8.5, DIC = 4.9 mg/l, chloride = 25 mg/ l), one time constant. A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1813 Fig. 8. A Bode plot for cuprous oxide coated surface (t = 0 h, pH = 8.5, DIC = 4.9 mg/l, chloride = 125 mg/l), two time constants. high dielectric constant. In general, degradation of the coating may take place by any one or a combination of the following mechanisms: (a) delamination of the coating (i.e., the breakage of the adhesive forces (physical/chemical) between the coating and the interface, resulting in the coating layer being stripped o€ from the metal surface); (b) physical breakdown or dissolution of the coating at certain locations, resulting in seepage of the aqueous electrolyte within the coating. Eventually, parts of the underlying metal surface get exposed to the electrolyte directly. The electrolyte then tries to form a layer between the coating and the interface and breaks down the adhesive forces at the interface. Mechanism (a) usually takes place under conditions of very high hydrodynamic shear. Mechanism (b) appears more likely under laminar ¯ow or quiescent conditions (like the conditions in the present study). 3.1. E€ect of DIC and chloride concentrations at pH 7 Table 4 summarizes the parameters estimated from the EIS analyses for the 1814 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 Table 4 EIS parameters for fresh and Cu2O coated copper surfaces at pH 7, at t = 0 and 72 h of exposurea Cc (mF) Cdl (mF) Rp (O) Rpo (O) Cc (mF) Cdl (mF) Rp (O) Rpo (O) Ct = 250 mg/l, Clÿ = 25 mg/l Ct> = 25 mg/l, Clÿ = 25 mg/l Ct = 250 mg/l, Clÿ = 125 mg/l Ct = 25 mg/l, Clÿ = 125 mg/l (288)b 363 (213) 13,437 (11,004) (2580)b 37 (49) 33 (47) 53,430 (51,380) 22,167 (19,364) (237) 209 (195) 6350 (9050) (1955)b 173 (288) 166 (257) 3500 (4230) 2805 (1204) (346)b 525 (282) 18,838 (7700) (2190)b 53 (70) 49 (69) 45,500 (28,550) 15,270 (10,460) (355)b 363 (282) 18,360 (6611) (1762)b 421 (487) 407 (388) 1880 (10,300) 1010 (1241) a The numbers in parentheses refer to those at t = 72 h (i.e., at the end of the exposure period) while those outside are for t = 0 h (i.e., initially). Bold refer to the Cu2O coated surface. b The measurement is not applicable. fresh and Cu2O coated surfaces at pH 7. The EIS spectrum for the fresh copper surfaces shifts from a one time constant case (initially) to a two time constant case (after 72 h) in all the cases, due to the development of a corrosion scale. For the fresh copper surface at Ct = 250 mg/l, increasing the Clÿ dosage to 125 mg/l results in a 30% increase in corrosion rate after 72 h of exposure. This can be inferred from the respective Rp values (since Rp is inversely proportional to the corrosion rate). The higher chloride dosage also results in higher values of coating and double layer capacitances (i.e., Cc and Cdl) after 72 h of exposure. At a Clÿ dosage of 25 mg/l and Ct = 250 mg/l, the Cu2O coating is extremely stable even after 72 h of exposure and results in a decrease of the corrosion rates by nearly a factor of 5. Cc does not increase appreciably upon exposure. At a Clÿ dosage of 125 mg/l, there is a signi®cant reduction in Rp for the ¯ame synthesized Cu2O after 72 h of exposure. However, the increase in Cc is very low indicating that the increase in corrosion rates was likely due to local defects in the coating and not due to the seepage of the electrolyte through the microstructure of the coating. For the Ct = 250 mg/l and Clÿ = 125 mg/l case, the ¯ame synthesized Cu2O coating resulted in the reduction of corrosion rates by a factor of 3.6 relative to the fresh copper surface. However, at a Ct concentration of 25 mg/l, the Cu2O coating was found to be extremely unstable in the presence of Clÿ ions at concentrations as low as 25 mg/l. In fact, the corrosion rates were observed to be 50 to 60% higher than those obtained with the fresh copper surfaces. Increasing the Clÿ concentration to 125 mg/l resulted in an increase in the initial corrosion rate by a factor of nearly two for the ¯ame synthesized Cu2O. Finally, for the Cu2O coating, the increase in the coating capacitance after 72 h of exposure was quite signi®cant for the Ct = 25 mg/l cases. 3.2. E€ect of DIC and chloride concentrations at pH 8.5 Table 5 summarizes the parameters estimated from the EIS analyses for the A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1815 Table 5 EIS parameters for fresh and Cu2O coated copper surfaces at pH 8.5, at t = 0 and 72 h of exposurea Cc (mF) Cdl (mF) Rp (O) Rpo (O) Cc (mF) Cdl (mF) Rp (O) Rpo (O) Ct = 250 mg/l, Clÿ = 25 mg/l Ct = 25 mg/l, Clÿ = 25 mg/l Ct = 250 mg/l, Clÿ = 125 mg/l Ct = 25 mg/l, Clÿ = 125 mg/l (107)b 360 (79) 11,050 (41,600) (5692)b 247 (259) 273 (281) 6249 (4864) 2012 (1448) (76)b 263 (63) 6500 (55,379) (10036)b 195 (187) 182 (178) 6652 (6549) 2070 (1992) (115)b 502 (94) 16,200 (39,560) (5722)b 234 (363) 251 (341) 5380 (6049) 2585 (1838) (91)b 410 (72) 13,300 (40,270) (8094)b 289 (241) 252 (213) 7330 (12,427) 2112 (2802) a The numbers in parentheses refer to those at t = 72 h (i.e., at the end of the exposure period) while those outside are for t = 0 h (i.e., initially). Bold refer to the Cu2O coated surface. b The measurement is not applicable. fresh copper coupon and the ¯ame synthesized Cu2O at pH 8.5. For the fresh copper surface with a Clÿ concentration of 25 mg/l, the corrosion rate after 72 h of exposure decreased by 23% as the Ct decreased from 250 to 25 mg/l. The lower Ct concentration also results in a more compact natural scale as is indicated by a signi®cant increase in Rpo from 5692 to 10,036 O, and a reduction in Cc from 107 to 76 mF. Increasing the Clÿconcentration to 125 mg/l results in a minor increase in corrosion rates for the Ct = 250 mg/l case and a 27% increase for the Ct = 25 mg/l case, after 72 h of exposure. For the fresh copper surface at Ct = 250 mg/l, the resistance of the scale developed after 72 h was found to be independent of the Clÿ dosage. At pH 8.5, the ¯ame synthesized Cu2O coating was observed to be extremely unstable, irrespective of the Clÿ and DIC concentrations. In fact, the corrosion rates after 72 h were slightly higher or insigni®cantly lower than those measured initially, except for the case of Ct = 25 mg/l and Clÿ = 125 mg/l. The initial corrosion rates for the Cu2O case were comparable to those for the fresh copper surface for one case (Ct = 25 mg/l, Clÿ = 25 mg/l) and signi®cantly higher than the fresh copper surface for the other three cases. For the Cu2O coated surfaces, the change in corrosion rates after 72 h was not very signi®cant, except for the case of Ct = 25 mg/l and Clÿ = 125 mg/l, where the corrosion rate decreased by a factor of 1.7. However, at pH 8.5 for the fresh copper surface, the reduction in corrosion rates after 72 h was very signi®cant for all four cases. The corrosion rates decreased by a factor of nearly 8.5 for the Ct= 25 mg/l and Clÿ = 25 mg/l condition. A possible explanation for the instability of the Cu2O coating at pH 8.5 and its ability to aggravate the corrosion process is as follows. At pH 8.5, the DIC speciation shifts signi®cantly to CO2ÿ 3 , which di€uses through the defects in the coating and reacts with the free ionic Cu2+ species at the interface [18,29]:   DG0f ˆ ÿ675:4 kJ=mol ‰30Š Cu 2‡ ‡ CO32ÿ ‡ H2 O $ CuCO3 OH ÿ ‡ H‡ 1816 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 2ÿ Cu 2‡ ‡ CO32ÿ ‡ 2H2 O $ CuCO3 OH †2 ‡2H‡   DG0f ˆ ÿ861:9 kJ=mol ‰31Š These reactions result in the release of protons, which decreases the pH signi®cantly over small pockets at the coating/metal interface. This results in the solubilization of the coating as follows: Cu2O(s) + 2H+ t 2Cu2+ + H2O. As a result, not only does the coating gets disturbed at a number of locations, but several Cu2+ ions are generated which can subsequently react to form and further reduce the pH. Hence, the Cu(II) CuCO3OHÿ and CuCO3(OH)2ÿ 2 complex formation and the Cu2O(s) dissolution processes are synergistic in nature. For the ¯ame synthesized Cu2O coatings, an examination of the Cc values for the pH 8.5 cases reveals that there is very little change after 72 h, except for the case of Ct = 250 mg/l and Clÿ = 125 mg/l (Table 5). This indicates that the attack by CO23, and the resulting increase in acidity, which lead to the dissolution process must be quite localized. Even after 72 h of exposure, the electrolyte does not get entrained within the coating microstructure to result in a general failure. However, the general failure of the coating may occur after a longer exposure. The defects and pores in the microstructure of the Cu2O ®lm are extremely detrimental for its proper performance as a corrosion inhibitor at pH 8.5. At low DIC and high chloride concentrations, the hydrolysis of CuCl to form Cu2O(s) becomes dominant. As a result, it is hypothesized that the rate of deposition of Cu2O exceeds the rate of its dissolution by the processes described above. Experimental results show an increase in Rp from 7330 to 12,427 O for the case of Ct = 25 mg/l and Clÿ = 125 mg/l at pH 8.5 (i.e., corrosion rates decrease by a factor of 1.7 after 72 h of exposure). This clearly supports the hypothesis proposed earlier. 3.3. Comparison of the performance of naturally scaled coupons and synthetic Cu2O ®lms The stability of di€erent ®lms was evaluated upon the exposure to (a) ozonated/ bio-®ltered Lake Michigan water (pH 7.34) and (b) Cincinnati tap water (pH 7.29). The chloride and DIC concentrations of the two water samples are shown in Table 6. The conditions for the formation of the two natural scales were speci®ed earlier (refer to Scales A and B). A summary of the EIS parameters is given in Table 7 3.3.1. Solution 1: ozonated/bio-®ltered Lake Michigan water Scale A resulted in a very large reduction of the initial corrosion rate relative to the fresh copper surface, the ¯ame synthesized Cu2O coated surface and scale B. However, the performance of scale A deteriorated rapidly during the 72 h exposure to solution 1 (Rp decreased from 56,000 to 19,300 O and Cc increased from 37 to 113 mF). The changes in the values of Cc and Cdl for the Cu2O and scale B cases after 72 h were very minor compared to those for the case of scale A. This indicates that scale A deteriorates much more rapidly than the other two. A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1817 Table 6 DIC and chloride concentrations of water samples Water sample pH DIC (mg/l) Chloride (mg/l) Lake Michigan, ozonated/bio-®ltered water Cincinnati tap water 7.34 7.29 24.0 19.4 40.8 10.6 At the end of the 72 h exposure period, the corrosion rate for the Cu2O coating was comparable to that for scale A. 3.3.2. Solution 2: Cincinnati tap water Initially, scale A was found to be an excellent corrosion inhibitor. However, similar to the earlier case, scale A was extremely unstable during the exposure period and resulted in a sharp increase in the corrosion rate (Rp decreased from 57,174 to 25,800 O and Cc increased from 37 to 92 mF). However, scale B was found to be the most stable and resulted in the lowest corrosion rates after 72 h of exposure. The corrosion rate for scale B (after 72 h of exposure) was only 50% of the rates for scale A and the synthetic Cu2O coating. The performance of the Cu2O coating upon exposure to Cincinnati tap water was much better than that upon exposure to ozonated/bio-®ltered lake Michigan water. From the synthetic water experiments it was concluded that, at a DIC to chloride ratio of 1.96 at pH 7 (i.e., Ct = 250 mg/l, Clÿ = 25 mg/l), Cu2O coating Table 7 EIS parameters for fresh, Cu2O coated and naturally scaled Cu surfaces using (a) solution 1: ozonated/ bio-®ltered Lake Michigan water, (b) solution 2: Cincinnati tap water, at t = 0 and 72 h of exposurea Cc (mF) Cdl (mF) Rp (O) Rpo (O) Fresh copper Cu2O coated Scale A Scale B Sol. 1 Sol. 2 Sol. 1 Sol. 2 Sol. 1 Sol. 2 Sol. 1 Sol. 2 ±b (68) 60 (58) 5070 (7345) ±b (10,400) ±b (42) 79 (36) 8225 (6900) ±b (20,106) 93 (119) 125 (112) 16,480 (17,904) 4342 (3636) 118 (112) 95 (90) 24,600 (30,262) 4970 (5568) 37 (113) 40 (85) 56,000 (19,300) 19,894 (6370) 37 (92) 40 (85) 57,174 (25,800) 19,000 (7416) 46 (60) 58 (55) 11,500 (12,640) 10,200 (3800) 50 (40) 61 (35) 32,100 (60,000) 12,300 (17,706) a The numbers in parentheses refer to those at t = 72 h (i.e., at the end of the exposure period) while those outside are for t = 0 h (i.e., initially). b The measurement is not applicable. 1818 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 was very stable and did not result in an increase in the corrosion rates signi®cantly (Table 4). Using Cincinnati tap water with a comparable DIC to chloride ratio and a slightly higher pH (i.e., 1.94, pH 7.29), the Cu2O coating was once again found to be very stable during the 72 h exposure period (Rp increased by nearly 25% and there was very little change in Cc, Table 7). However, the Rp values for the Cincinnati tap water were signi®cantly lower. This indicates that besides the three key parameters: pH, DIC and chloride concentrations, other organic and inorganic species and water quality parameters can signi®cantly in¯uence the instantaneous corrosion rates. However, the deterioration of the Cu2O coating may be a strong function of the DIC to Clÿ ratio at a given pH, and is almost independent of the concentration of the other inorganic and organic species. To con®rm this hypothesis, additional control experiments were conducted to assess the variation of coating capacitance, Cc, as a function of the DIC to Clÿ ratio. The pH was maintained at approximately 7.0 and the chloride concentration was kept ®xed at 25 mg/l. The DIC adjustment was carried out by using freshly prepared 2 M stock solution of NaHCO3. To assess whether the other background water quality parameters a€ected the results, control experiments were conducted using Cincinnati tap water whose DIC to Clÿ ratio had been adjusted to certain speci®c values. Table 8 provides a summary of the typical values of pertinent background water quality parameters for Cincinnati tap water. All wet chemistry analyses for assessing water quality parameters were conducted using protocols described in Standard Methods for the Examination of Water and Waste water [32]. Fig. 9 shows that irrespective of the background water quality conditions, Cc was found to be a strong function of the DIC to Clÿ ratio at pH 7. At pH 7, Cc values were found to be the lowest in the vicinity of a DIC to Clÿ ratio of 2. The plots for the synthetic water and the Cincinnati tap water overlap greatly, Table 8 Pertinent water quality parameters for Cincinnati tap water at 258C (avg.2standard deviation) Parameter Value pH Alkalinity (mg CaCO3/l) DO (mg/l) Conductivity (mS) ORP (mV) DIC (mg/l) Chloride (mg/l) Sulfate (mg/l) TOC (mg/l) DOC (mg/l) BDOC (mg/l) UV254 (cmÿ1) 7.2920.03 46.525.8 7.020.1 21523 30225 19.421.2 10.620.6 21.421.7 1.120.02 0.920.07 0.220.1 0.2720.02 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1819 Fig. 9. Cu2O coating capacitance after 72 h as a function of DIC/chloride ratio at pH 7 (chloride = 25 mg/l). although at any given DIC to Clÿ ratio, the Rp values with the Cincinnati tap water were signi®cantly lower than those with the synthetic water (not shown in Fig. 9 (i.e., the corrosion rates were greatly dissimilar). Therefore, although earlier studies [33,34] have concluded that chloride (up to rather high concentrations) decreases the corrosion rate of copper, our results in this study indicate that the interplay between chloride and DIC is very interesting and complicated. Furthermore, both chloride and DIC should be considered in future studies, which attempt to predict and fully understand copper corrosion in distribution systems. 3.4. E€ect of bu€er concentration on corrosion behavior The choice of the bu€er solutions was primarily dictated by their inertness and has practically no in¯uence on the corrosion chemistry. Our objective was to limit 1820 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 Table 9 Comparison of EIS results at two di€erent bu€er concentrations, for cuprous oxide coated samplea EIS parameters Cc (mF) Cdl (mF) Rp (O) Rpo (O) pH 7.0, Ct = 250 mg/l and Clÿ = 25 mg/l pH 8.5, Ct = 250 mg/l and Clÿ= 25 mg/l 0.01 (M) MOPS 37 (49) 33 (47) 53,430 (51,380) 22,167 (19,364) 0.01 (M) boric acid 247 (259) 273 (281) 6249 (4864) 2012 (1448) 0.1 (M) MOPS 47 (59) 48 (61) 43,900 (42,101) 18,210 (15,480) 0.1 (M) boric acid 290 (317) 320 (345) 5600 (4344) 1807 (1300) a The numbers in parentheses refer to those at t = 72 h (i.e., at the end of the exposure period) while those outside are for t = 0 h (i.e., initially). the bu€er concentration to the bare minimum without, at the same time, compromising our ability to maintain a fairly constant pH over the exposure period. We agree that bu€er solutions increase the conductivity of the test solution and lead to potentially higher observed corrosion rates (due to the migration contribution). We were aware of this while performing the experiments. In order to get an idea about the e€ects of variable bu€er concentrations (and hence solution conductivity), a few control experiments have been performed (both at pH 7.0 and 8.5) at a bu€er concentration of 0.1 M (instead of 0.01 M, which was used in most of the experiments). Table 9 summarizes the EIS experimental data for the two di€erent bu€er concentrations. The 10-fold increase in the bu€er concentration led to a nearly 18% increase in the corrosion rate at pH 7 and a 10% increase at pH 8.5. However, one must remember that the main focus of our discussion in this paper is not to study the absolute values of corrosion rates, but to evaluate the relative performance of the synthesized and natural scales under di€erent pH conditions and at di€erent chloride and DIC concentrations. It is certain that the corrosion rate will depend on the solution conductivity. However, by keeping the bu€er concentration constant in all our experiments, we can almost completely eliminate the di€erences observed in corrosion rates due to conductivity/migration, when comparing the e€ects of chloride and DIC concentrations at the two pH values studied. This will provide a common baseline to compare the e€ects of chloride and DIC on scale stability at the two pH values. Hence, we feel that although conductivity/migration contribution is an important issue, it is even more important to try to establish a common baseline from which to compare the e€ects of chloride and DIC. 4. Conclusions The performance and the stability of the Cu2O scale are strongly dependent upon the pH, the DIC and the chloride concentrations. At pH 7, the coating was found to be very stable at a DIC to chloride ratio in the vicinity of 2. However, at A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 1821 pH 8.5, due to the shift in the DIC speciation toward the CO2ÿ and the 3 consequent development of small pockets of high H+ concentration, the performance of the Cu2O coating deteriorated very fast. Under the in¯uence of real waters, the performance of the Cu2O coating was found to be comparable to that of the two naturally developed scales only under certain conditions. Although the instantaneous corrosion rates are strongly dependent on several water quality parameters and inorganic and organic species, the deterioration of the Cu2O ®lm upon exposure at a given pH seems to depend most strongly on the DIC to chloride ratio. The present study by no means answers all the questions regarding ®lm stability in distribution systems, though it provides an excellent foundation for additional studies in this critical area of research. The performance of other Cu(I) and Cu(II) mineral phases that have been identi®ed in the water distribution corrosion scales needs to be quanti®ed under the in¯uence of a variety of real and synthetic waters. The e€ect of temperature, hydrodynamic shear and disinfectant residuals on the solubility, stability and morphology of these scales need to be ascertained in order to optimize the performance of the ®lms, when they are exposed to di€erent water quality conditions. Acknowledgements The authors wish to thank Prof. Pratim Biswas of University of Cincinnati for his generosity in using the aerosol reactor for cuprous oxide synthesis and Mr. Michael Schock of the USEPA Drinking Water Laboratory (Cincinnati, Ohio) for his useful comments and input to the scope of the research described herein. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] M. Schock, D. Lytle, J. Clement, EPA Report: EPA/600/R-95/085 (1995). Y. Feng, W.-K. Teo, K.-S. Siow, K.-L. Tan, A.-K. Hsieh, Corrosion Sci. 38 (1996) 369. Y. Feng, W.-K. Teo, K.-S. Siow, A.-K. Hsieh, Corrosion Sci. 38 (1996) 387. E. Mattson, Brass Corros. 15 (1980) 6. D.C. Atlas, J. Coombs, O.T. Zajicek, Water Res. 16 (1982) 693. H. Cruse, O. von Franque, Internal Corrosion of Water Distribution Systems, AWWA Research Foundation/DVGW Forschungsstelle, Denver, Colorado, 1985. S. Reiber, JAWWA 81 (1989) 114. M. Edwards, M. Schock, T. Meyer, JAWWA 88 (1996) 81. T.E. Meyer, The e€ect of inorganic anions on copper corrosion. A Ph.D. thesis at University of Colorado at Boulder, 1994. S.B. Adeloju, H.C. Hughes, Corrosion Sci. 26 (1986) 851. N. de Zoubov, C. Vanleugenhaghe, M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, 2nd ed., National Association of Corrosion Engineers, Houston, TX, 1974 (Section 14.1). C.F. Jr. Baes, R.E. Mesmer, The Hydrolysis of Cations, Wiley±Interscience, New York, 1976. F.J. Millero, M. Izaguirre, V.K. Sharma, Mar. Chem. 22 (1987) 79. J.N. Alhajji, M.R. Reda, British Corrosion J. 31 (1996) 125. 1822 A. Palit, S.O. Pehkonen / Corrosion Science 42 (2000) 1801±1822 [15] A. Palit, S.O. Pehkonen, Copper corrosion: the e€ect of speci®c water quality parameters, Student Paper Competition of the 91st Annual Meeting of the Air & Waste Management Association, 1998. [16] Y. Feng, K.S. Siow, W.K. Teo, K.L. Tan, A.K. Hsieh, Corrosion 53 (1997) 389. [17] F.A. Cotton, G. Wilkinson, Advanced Inorganic Chemistry, 5th ed., Wiley, New York, 1988. [18] M. Drogowska, L. Brossard, H. Menard, J. App. Electrochem. 24 (1994) 344. [19] C.H.J. Elzenga, H.J. Boorsma, Corrosion of copper pipes in various types of drinking water, International Committee on Corrosion and Protection of Underground Pipelines, 1981. [20] S. Yamauchi, Dissolution of copper from copper water tubes and new copper alloys showing anti-cuprosolvency, in: Proc. International Symposium on Corrosion of Copper and Copper Alloys in Buildings, Tokyo, Japan, 1982. [21] M.M., Benjamin Chemistry of Corrosion Inhibitors in Potable Water. AWWA Research Foundation, Denver, CO, 1990. [22] D.W. Shoesmith, et al., Electrochem. Acta 22 (1977) 403. [23] Y. Suzuki, H.J. Akahoshi, Electrochem. Soc. 57 (1989) 983. [24] H. Chang, P. Biswas, J. Colloid Interface Sci. 153 (1992) 157. [25] G. Yang, H. Zhuang, P. Biswas, Nanostructured Materials 7 (1996) 675. [26] C.O. Wells, Scanning Electron Microscopy, McGraw-Hill, New York, 1974. [27] L.B. Gabriel, SEM: A User's Manual for Material Science, American Society of Metals publication, Metal Park, OH, 1985. [28] S.W. Tait, An Introduction to Electrochemical Corrosion Testing for Practicing Engineers and Scientists, PairODocs Publications, Racine, Wisconsin, 1994. [29] M. Drogowska, L. Brossard, H. Menard, J. Electrochem. Soc. 139 (1992) 39. [30] L. Millette, D.S. Mavinic, Canadian Jour. Chem. 15 (1988) 79. [31] P.T. Gilbert, Water Treatment and Examination 15 (1966) 165. [32] A.D. Eaton, L.S. Clesceri, A.E. Greenberg, Standard Methods for the Examination of Water and Wastewater,, 20th ed., United Book Press, Baltimore, MD, 1996. [33] A. Elfstrom Broo, B. Berghult, T. Hedberg, Corrosion Science 39 (1997) 1119. [34] M. Edwards, T. Meyer, J. Rehring, JAWWA 86 (1994) 73.