organic matter and copper corrosion by-product release: a mechanistic study

18
Organic matter and copper corrosion by-product release: a mechanistic study Marc Edwards * , Nicolle Sprague Department of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University, Blacksburg VA 24061-2046, USA Received 19 July 1999; accepted 9 May 2000 Abstract The adverse eects of natural organic matter on copper corrosion by-product release to drinking water result from soluble complex formation, colloid mobilization/dispersion, and interference with aging processes that tend to decrease scale solubility. Organic matter can also reverse copper corrosion by-product release to drinking water by (1) fueling microbial removal of oxygen, causing re-deposition of Cu(I) to the pipe wall, and (2) sorption of soluble organic matter onto pipe surfaces during stagnation, which decreases the solutionÕs complexation capacity for copper. Ó 2000 Elsevier Science Ltd. All rights reserved. Keywords: Copper; Natural organic matter; Drinking water; Corrosion by-products; Equilibrium calculations; Thermodynamic diagrams 1. Introduction Although exhaustive practical experience has highlighted the importance of nat- ural organic matter (NOM) in corrosion by-product release from copper pipe in drinking water [1–7], there are few mechanistic studies in the literature which clearly identify the mechanism(s) by which NOM acts. Consequently, when considering the attractive yet costly prospect of controlling copper release by removing NOM or manipulating water chemistry, there is little basis for decision making. That is, there have been some reports that NOM has little or no eect, in other cases NOM caused substantial short-term reductions in copper release [3], and in still other cases the concentration of copper release directly increases with NOM concentration in the www.elsevier.com/locate/corsci Corrosion Science 43 (2001) 1–18 * Corresponding author. Tel.: +1-540-231-7236; fax: +1-540-231-7916. 0010-938X/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved. PII:S0010-938X(00)00071-8

Upload: marc-edwards

Post on 01-Nov-2016

217 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Organic matter and copper corrosion by-product release: a mechanistic study

Organic matter and copper corrosion by-productrelease: a mechanistic study

Marc Edwards *, Nicolle Sprague

Department of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University,

Blacksburg VA 24061-2046, USA

Received 19 July 1999; accepted 9 May 2000

Abstract

The adverse e�ects of natural organic matter on copper corrosion by-product release to

drinking water result from soluble complex formation, colloid mobilization/dispersion, and

interference with aging processes that tend to decrease scale solubility. Organic matter can also

reverse copper corrosion by-product release to drinking water by (1) fueling microbial removal

of oxygen, causing re-deposition of Cu(I) to the pipe wall, and (2) sorption of soluble organic

matter onto pipe surfaces during stagnation, which decreases the solutionÕs complexation

capacity for copper. Ó 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Copper; Natural organic matter; Drinking water; Corrosion by-products; Equilibrium

calculations; Thermodynamic diagrams

1. Introduction

Although exhaustive practical experience has highlighted the importance of nat-ural organic matter (NOM) in corrosion by-product release from copper pipe indrinking water [1±7], there are few mechanistic studies in the literature which clearlyidentify the mechanism(s) by which NOM acts. Consequently, when considering theattractive yet costly prospect of controlling copper release by removing NOM ormanipulating water chemistry, there is little basis for decision making. That is, therehave been some reports that NOM has little or no e�ect, in other cases NOM causedsubstantial short-term reductions in copper release [3], and in still other cases theconcentration of copper release directly increases with NOM concentration in the

www.elsevier.com/locate/corsci

Corrosion Science 43 (2001) 1±18

* Corresponding author. Tel.: +1-540-231-7236; fax: +1-540-231-7916.

0010-938X/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved.

PII: S00 1 0-9 3 8X (0 0 )0 00 7 1- 8

Page 2: Organic matter and copper corrosion by-product release: a mechanistic study

water [1,6,8]. Perplexing ``presence or absence'' e�ects have also been reported [4,5],in which even trace 0.1 mg/l levels of NOM produce marked (>0.8 Cu mg/l) increasesin copper release to water, while further increases in NOM concentration produceonly slight additional increases to copper concentration.

Recent work has provided some insights on mechanisms that can be readilyevaluated in laboratory experiments. The experiments conducted in this work weredesigned based on the following:

· notion that soluble copper concentrations within a relatively new pipe is con-trolled by Cu(OH)2 solid equilibrium, with associated formation of soluble car-bonate complexes [7,9];

· belief that overall solubility of the Cu(OH)2 is reduced as the pipe ages by poorlycharacterized solid aging e�ects, with concomitant reductions in copper release[7,10±12];

· discovery that predictions of copper corrosion by-product release in low alkalinitywater are signi®cantly improved by explicitly considering possible complexationof copper by NOM [6];

· ®nding that copper corrosion in high alkalinity water is dramatically increased atlow levels of NOM, but relatively insensitive to additional increases in NOM con-centration [4,5];

· hypothesis of Korshin et al. that NOM might cause mobilization of colloidal cop-per via particle stabilization and detachment [5];

· report that copper corrosion by-product release sometimes reverses during stag-nation under at least some circumstances [12].

With few exceptions [3,5], these ®ndings were obtained from ®eld observations ofcopper corrosion in real systems with all limitations inherent to such complex situ-ations [6,13,14]. Thus, the ®rst goal of this study was to attempt and reproduce thephenomena under laboratory controlled conditions. It was hoped that this wouldprovide a basis for experiments that would better identify mechanism(s) by whichNOM interacts with copper.

2. Experimental

2.1. Organic matter

Three types of organic matter were used including sodium alginate, soluble NOMand particulate NOM. The soluble NOM solution (Table 1) was concentrated fromLake Pleasant, WA by reverse osmosis, followed by ®ltration and ion exchange ofcalcium and magnesium hardness as described elsewhere [15]. Particulate NOM waspurchased as peat from a local garden supply store, rinsed three times with deionizedwater, and collected by ®ltration before use in experiments. The sodium alginate,which is representative of bacterial extracellular polymeric substances, was dosed tosolutions at a concentration of 2% (w/v) two days before experiments to simulateexperiments presented elsewhere [16].

2 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 3: Organic matter and copper corrosion by-product release: a mechanistic study

2.2. Total and soluble copper measurements

In preliminary experiments, copper was analyzed by an inductively coupledplasma emission spectrophotometer (ICP-ES) according to standard method 3120[17] or a colorimetric method described in standard method 3500 C [17]. Resultsfrom the ICP-ES and the colorimetric test were compared and were not signi®cantlydi�erent from one another, but the colorimetric test could be used in very viscousalginate solutions not suited to ICP-ES. Total copper was determined colorimetri-cally after adjusting the sample pH to 3.0, whereas soluble copper was operationallyde®ned using ®ltration through a 0.45 lm pore size syringe ®lter according tostandard methods [17]. It should be noted that in the presence of colloidal speciesthat can pass through the ®lter, the standard methods approach represents an upperbound to truly soluble copper.

2.3. Free Cu2�

Free cupric ion concentrations were measured during titration experiments usingan Orion 94±29 cupric electrode, an Orion 90±02 double junction reference electrodeand an appropriate ion analyzer. The instrument was calibrated using cupric ionstandards at the same ionic strength and stirring rate as samples, but at pH 6.0. Themeter was corrected for drift every hour and the entire apparatus was covered withblack plastic to prevent light interference.

At pH 6.0, copper is virtually 100% cupric ion, whereas at pH 7.0 cupric ion isonly about 50% of the total soluble copper even if carbonate, organic matter, andother complexing ligands are absent. To con®rm the general validity of the electrodeat pH 7.0, solutions were constituted containing malonic and dipicolinic acids whichhave functional groups similar to those in the alginate and NOM but with wellde®ned complexation constants. Comparison of measurements using the electrode to

Table 1

Key characteristics of concentrated soluble NOM stock solution

Parameter Value

DOC 1150 mg/l

Humic DOCa 219 mg/l

Sum hardness (Ca, Mg, Al) <4 mg/l

pH 6.2

Net anonic charge at

pH 6.2 9.7 meq/g DOC

pH 7.0 10.1 meq/g DOC

pH 11.0 11.2 meq/g DOC

KCu (at pH 7.0) 1:44� 10ÿ5

Lt 13 mM sites/g DOCb

Cu 2 mg/l

a Fraction precipitating at pH 1.0.b Valid at 7.0 over the concentration range of 0.015±1.5 mg/l free copper.

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 3

Page 4: Organic matter and copper corrosion by-product release: a mechanistic study

those predicted by equilibrium chemistry indicated that at very low levels of freecopper (<15 ppb), the measured [Cu2�] was about 25% higher than predicted;however, in samples containing above 15 ppb of free copper, the predicted andmeasured values were in good agreement (always less than 20% di�erence). Thiscon®rmed the general validity of the method and established a practical quantitationlimit of 15 ppb Cu2�.

2.4. Titrations and pipe rig tests

Two types of experiments were conducted including titrations and pipe rig tests.Titrations were designed to illustrate fundamental aspects of copper NOM chem-istry. All titrations were conducted in 1 l glass containers at 20� 0:1°C, pH 7.0, ineither 1 or 2 mM NaNO3 as speci®ed. The general approach was to dose di�erentlevels of cupric ion to stirred solutions containing organic matter. The pH was heldconstant at 7:0� 0:2 throughout the experiment by dropwise addition of NaOH.After a speci®ed time period, changes in particle and solution characteristics weremonitored through measurements of free Cu2�, soluble copper, total copper, organicmatter (UV254), and particle zeta potential.

Pipe rig tests used a modi®ed method of Johannson [18] to illustrate changes incopper corrosion by-product release in response to water quality changes and aging.The rigs were 12 in. long, 3=4 in. diameter type L tubes with a volume of approxi-mately 100 ml. Number two rubber stoppers were used to plug the ends of each tubeand the pipe after being carefully rinsed three times with reagent grade water. Allpipes were exposed to a soft, low alkalinity, NOM free water synthesized in thelaboratory (Table 2), or the same water with either 2 mg/l as the total organic carbon(TOC) soluble or 2% (w/v) sodium alginate. A parallel experiment was run at pH 9.5,making a total of 12 pipes after duplication.

After adjusting each water to either pH 7.0 or 9:5� 0:2, fresh solution was pouredinto the pipes each Monday, Wednesday and Friday, resulting in a regular 48, 48 and72 h stagnation sequence for the experiment which ran as long as one year. Solutions

Table 2

Low alkalinity soft water synthesizeda for use in pipe rig experiments

Ion Concentration (mg/l)

Magnesium 3

Chloride 9

Calcium 5

Nitrate 16

Sulfate 16

Sodium 17

Silica 20

Alkalinity (as CaCO3) 5

a Added as reagent grade MgCl2, Ca�NO3�2 � 4H2O, Na2SO4, Na2SiO3, and Na2CO3; pH was adjusted

to 7.0 or 9.5 with addition of reagent grade NaOH and HNO3.

4 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 5: Organic matter and copper corrosion by-product release: a mechanistic study

were collected weekly and monitored for soluble copper, total copper, dissolvedoxygen (DO) and pH after the 72 h stagnation. Data from duplicate pipes were notstatistically di�erent and were pooled for con®dence testing. The temperature of thepipes and solutions was 20� 0:1°C at all times.

2.5. Other measurements

Sample pH was measured using a Beckman 11 pH meter and DO was measuredusing a Hanna Instruments HI 9141 DO probe. TOC was quanti®ed in samples usinga Dorhmann DC-80 TOC Analyzer with a DC-80 sludge/sediment sampler ac-cording to standard method 5310 [17]. In titration experiments, the concentration ofNOM removed by sorption or precipitation reactions was tracked by measurementsof UV absorbance at 254 nm according to standard method 5910 B but at solutionpH 3.0. Experiments indicated that copper did not interfere with the UV measure-ment at this pH and that soluble UV254 was proportional to DOC. Zeta potentialwas measured using electrophoresis. Measurements were conducted in triplicate andthe general accuracy of the measurement was con®rmed using standard particlesolutions. As a general rule, if particle zeta potentials are above a certain criticalabsolute value (e.g., > �18 mV or < ÿ18 mV), the solution will contain highlydispersed colloids with a small particle size. In contrast, if particle zeta potentials areless than the critical absolute value, particle agglomeration can readily occur.

3. Results and discussion

Initial experiments illustrated fundamental interactions between copper and or-ganic matter including complexation, NOM removal from solution by sorption and/or precipitation reactions, and particle dispersion. Thereafter, the important impactof NOM on the solubility of copper solids with aging was revealed. The ®nal phaseof experiments clearly de®ned, for the ®rst time, reproducible circumstances that canactually reverse copper corrosion by-product release into water under some cir-cumstances.

3.1. Fundamental aspects of organic matter/copper interactions

Interactions between NOM and cupric ion may be viewed as variations of thebasic complexation reaction:

Cu2� � Lÿz ! CuL�2ÿz�; �1�where Cu2� is the cupric ion concentration, Lÿz, the organic ligand binding site withcharge ÿz, CuL�2ÿz�, the cupric organic matter complex.

For a given ligand, this association is expressed through the equilibrium constant

Ki � �CuLi��Cu2���Li�

: �2�

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 5

Page 6: Organic matter and copper corrosion by-product release: a mechanistic study

Di�erent classes of organic ligands are present in drinking water supplies. Ofthese, soluble anionic polymeric substances such as fulvic acids are dominant,although colloidal organic materials such as peat can also be signi®cant under somecircumstances [19,20]. Ligands from biological processes such as extracellularpolymeric substances (EPS) are generally not important in the bulk water phase,although they are present at extremely high concentrations on the pipe surface withinomnipresent bio®lms.

Even within a given type of organic matter, a wide range of copper binding sitesmay be present, each with a distinct complexation constant ``K'' and concentrationthat varies with pH and ionic strength. Considering this, several approaches can beused to model copper binding including distributed binding strength (in®nite vari-ation in ligand binding strength), multiple ligand binding strength (a ®nite number ofdiscrete sites with varying concentration) and single ligand binding [20]. Thereafter,cupric ion in natural waters can be predicted as the sum of the free copper (Cu2�),soluble inorganic complexes (Cu±Ing) such as Cu�OH�2ÿx

x or CuHCO�3 , soluble or-ganic complexes (Cu±DOM), particulate organic complexes (Cu±POM), and inor-ganic precipitates (cupric solids) i.e.,

�Cu�total � Cu2� � Cu±Ing� Cu±DOM� Cu±POM� Cupric solids: �3�

As corrosion and dissolution of scale proceed within an aerobic copper pipe, thedrinking water is equilibrated with the scale during stagnant ¯ow periods and viceversa, essentially titrating the solution with cupric ions. Consequently, a range ofphenomena relevant to corrosion by-product release can be examined by simplytitrating solutions containing various organic matter sources, while tracking changesin free copper and soluble copper. Of course, if total copper concentrations arebelow the solubility product of inorganic copper solids (i.e., about 6 mg/l at pH 7.0and 23°C for fresh Cu(OH)2) [10], formation of copper organic complexes can onlydecrease free copper concentrations compared to solutions in which organic matter isabsent (Fig. 1). Approximate linearity in plots of free copper versus total copper forsolutions containing 2 mg/l particulate organic matter and 2 mg/l soluble organicmatter demonstrate that, over the free copper concentration range of interest indrinking water (�0.015±1.5 mg/l free copper), multiple ligand models would not ®tthe experimental data signi®cantly better than models using a single binding site.Thus, a single ligand model with complexation constants valid over this range (Table1) was used throughout this investigation. This approximation is not likely to bevalid at lower levels of free copper [6,20].

The nature of the organic matter ligand determines whether complexation pro-duces particulate, soluble, or some combination of particulate and soluble copperreaction products. In the case of soluble NOM, the resulting reaction products arecompletely soluble, whereas they are virtually 100% particulate in the case of alginate(Fig. 1). Interestingly, although the alginate solution contained virtually 100% sol-uble organic matter, addition of cupric ion and subsequent binding to alginateproduced a visible insoluble precipitate that could be at least partly removed by®ltration. Thus, copper addition to solutions containing alginate produced a mixture

6 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 7: Organic matter and copper corrosion by-product release: a mechanistic study

of soluble and particulate copper. In the context of Eq. (1), formation of particulatecopper alginate solids may be appropriately viewed as a precipitation reaction, inwhich a copper±alginate solubility product is exceeded. Particulate NOM exhibitedbehavior between these extremes, in that about half of the copper was soluble andhalf the copper was particulate. The portion of the copper that was soluble in thepresence of particulate NOM is well predicted based on measured levels of freecopper and known inorganic complexes, so it is likely that the fraction of copperbound to organic matter was completely particulate in this case.

Another important reaction involves colloid mobilization or dispersion of pre-existing solids. That is, under some circumstances in natural waters, anionic organicmatter can sorb to pre-existing copper solid surfaces and disperse particles similar tothe action of anti-scaling agents or anti-coagulants [21]. The basic reaction beginswith sorption of organic matter to pre-existing copper surfaces by surface com-plexation reactions of the type

S±OH� Lÿz ! S±L�1ÿz� �OHÿ; �4�

where S±OH is the surface complexation site on a copper solid or particle. To il-lustrate this reaction, soluble NOM was titrated with solutions containing freshlyprecipitated Cu(OH)2 solids prepared according to methods published elsewhere(titration of 0.5 mM Cu(NO3)2 solutions to pH 7.0) [10]. After allowing 12 h of

Fig. 1. Free copper (above) and soluble copper (below) after titrating solutions containing di�erent types

of organic matter.

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 7

Page 8: Organic matter and copper corrosion by-product release: a mechanistic study

aging, the resulting particle zeta potential was determined. The amount of NOMsorbed onto the pre-existing particles was also determined by ®ltration through a0.02 lm pore size ®lter. At pH 7.0 and in the complete absence of organic matter,Cu(OH)2 solids have a highly positive surface charge in 1 mM NaNO3 solutions(Fig. 2). However, as the concentration of soluble NOM in the water at equilibriumincreases, the concentration of organic matter on the particle surface increases andparticle zeta potential (surface charge) decreases. In fact, the particle zeta potentialdecreases to below about ÿ20 to ÿ30 mV, values below which particle dispersionbegins to become visually signi®cant, when soluble NOM in the water is only about1±2 mg/l. Above soluble NOM levels of about 2 mg/l, only slight e�ects on particlezeta potential and sorption density are observed.

Dispersion of pre-existing colloids by sorbed NOM can be dramatic. To illustratethis, fresh Cu(OH)2 solids were prepared and after about 2 h of aging, soluble copperin the water was stable at 6.2 mg/l consistent with Cu(OH)2 equilibrium (Fig. 2) [10].Aliquots of this solution were then spiked with either 2 or 20 mg/l NOM. Threeminutes after spiking 20 mg/l NOM to the water, the zeta potential decreased tobelow ÿ35 mV and more than 97% of the copper passed through a 0.45 lm pore size®lter. Interestingly, however, only 12 mg/l of this copper passed through a 0.02 lmpore size ®lter, supporting the notion that the copper was mostly in the form of astable copper colloid with some soluble complex formation. No change in coppersolubility occurred with the 2 mg/l NOM spike, mostly likely because the particlezeta potential only decreased from �35 mV down to �10 mV.

Fig. 2. E�ect of NOM on particle zeta potential (above) ± added NOM can have either drastic or no e�ect

on soluble copper, depending on the concentration added due to a dispersion e�ect (below); conditions

include pH 7.0, I � 2 mM NaNO3, 20°C.

8 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 9: Organic matter and copper corrosion by-product release: a mechanistic study

Complexation, colloid stability, and NOM sorption to precipitated solids must allbe considered in order to interpret the results of titrations over a wide range ofcopper doses at pH 7.0 when soluble NOM is present. In the absence of NOM,soluble copper increases steadily until the solubility product is exceeded at about 6mg/l copper, after which time most additional copper added to the solution pre-cipitated (Fig. 3). In the presence of NOM, however, and at relatively low levels ofcopper addition well below the solubility product of Cu(OH)2 solids, formationof soluble NOM complexes increases solubility of copper. In fact, measurements ofNOM-bound copper (NOM bound copper � total copper ÿ free Cu2�ÿ calculatedcopper hydroxide complexes) were in very good agreement with that predicted by thesingle ligand binding model (Fig. 3). At very high levels of copper addition (above 10mg/l), as much as 80% of the soluble NOM is removed from solution by sorptionreactions (Fig. 3). In this range, soluble copper concentrations in solutions con-taining NOM can decrease markedly, since there is less NOM in solution to form

Fig. 3. Soluble copper concentrations are a complex function of total copper added and NOM concen-

tration (above); at low levels of total copper, increased solubility is consistent with complexation models

(middle); at high levels of copper, NOM is removed from solution by sorption onto precipitating cupric

solids (below).

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 9

Page 10: Organic matter and copper corrosion by-product release: a mechanistic study

soluble copper complexes. In the intermediate copper dose range, combined e�ects ofsoluble complexes and stabilized colloids tend to increase solubility compared tosolutions without NOM.

3.2. Role of NOM in Cu�OH�2 aging

Complexation and colloid mobilization are two important mechanisms leading toincreased copper corrosion by-product release in the presence of NOM [5,6]. Giventhe importance of scale identity in controlling overall by-product release [7,12,13],another key mechanism of interaction might involve altering the identity of the scalelayer itself. To investigate this possibility, the experiment depicted in Fig. 3 wasallowed to continue while stirring at 20°C for several weeks. Consistent with thegradual transition of fresh Cu(OH)2 to tenorite with aging in such circumstances,soluble and free copper decreased markedly in the absence of NOM [11]. At ®rstglance, the increased copper solubility in the presence of NOM might be suspected toresult from complexation; however, this is completely inconsistent with the higherfree copper concentrations observed when NOM is present (Fig. 4). This can only beexplained by the presence of a higher solubility solid phase compared to the systemwithout NOM.

For example, after 7 days aging in the absence of NOM, free copper was 0.55 mg/lwhile soluble copper was 1.17 mg/l (Fig. 4). Making the reasonable assumption thatthe di�erence of 0.62 mg/l is attributable to soluble hydroxide complexes such asCu�OH�02, the ®xed ratio of copper hydroxide complexes to free copper is calculated

Fig. 4. The presence of sorbed NOM interferes with aging of cupric solids, maintaining higher solubility

(above) and higher Cu2� concentrations (below).

10 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 11: Organic matter and copper corrosion by-product release: a mechanistic study

to be 1.12 to 1 in this system. In comparison, after 7 days in the solution that initiallycontained 8 mg/l NOM, the free copper concentration was 1.40 mg/l and the hy-droxide complexes are calculated to be 1.57 mg/l. The remaining soluble copper(3:94ÿ 1:40ÿ 1:57 � 0:97 mg/l) may be attributed to soluble organic complexes withthe small amount (�1.5 mg/l TOC) organic matter left in solution after sorption to thesolids that formed (Fig. 3). In summary, of the 336% increase in copper solubility dueto NOM in this case, only about 1/3 is directly attributed to organic matter com-plexes, the remainder is due to the higher free copper associated with a higher solu-bility solid phase. Even this underestimates the importance of the higher solubilitysolid, since according to Eqs. (1)±(3) the concentration of soluble copper organicmatter complexes will also increase as the result of higher free copper concentrations.Similar e�ects were observed when 2 mg/l NOM was initially present (Fig. 4).

Is there any evidence that NOM interferes with scale aging under conditionspractically relevant to corrosion? To answer this question, two experiments wereconducted. The ®rst experiment used the pipe rigs from which data at various ex-perimental times were compiled (Fig. 5). At pH 7.0 and in the absence of NOM,copper corrosion by-product release after 48 h of stagnation averaged 0.79 in the ®rstmonth, eventually decreasing to 0.63 mg/l after 8±10 months of experimental time. Incontrast, for the same condition but with 2 mg/l NOM present, copper release in-creased from 1.6 to 2.2 mg/l NOM over the same time period. This increase with

Fig. 5. If NOM is present during aging, by-product release increases with time, whereas it decreases with

time without NOM (above). When copper pipes are aged in the complete absence of NOM, increases in

copper at higher NOM are consistent with complexation, but at lower levels other e�ects are also operative

(below).

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 11

Page 12: Organic matter and copper corrosion by-product release: a mechanistic study

aging was signi®cant at 95% con®dence, as was the higher copper release in thepresence of NOM compared to the system without NOM at pH 7.0.

At pH 7.0, the vast majority of the copper corrosion by-products were soluble.Based on the soluble copper concentrations and the hydroxide formation constantsdescribed earlier, it is estimated that 0.37 and 0.29 mg/l Cu2� were present in thepipes without NOM at the new and old conditions, respectively. Using the singleligand model and the assumption of equilibration with 2 mg/l soluble NOM, thesefree copper concentrations would be predicted to form 0.75 and 0.65 mg/l solubleorganic matter complexes, respectively. In relatively new pipes, this predicted in-crease in copper is close to that predicted in the presence of 2 mg/l NOM, suggestingthat the scale in such pipes was equilibrated with the same concentration of freecopper with or without NOM present. However, in the older pipes, the increase insolubility was 0.87 mg/l higher than the prediction based on NOM complexationalone, and the assumption of identical free copper concentrations is not valid ± theimplication is that a scale formed in the presence of NOM that maintained a higherconcentration of free copper.

A second experiment provided another perspective on this issue. After the aboveexperiment was completed, three new pipes were started at the pH 7.0 condition andexposed for 2 months. Thereafter, solutions with 0, 2 and 8 mg/l NOM TOC wereadded to these new pipes and the old pipes in sequence. Before adding the sampleswith NOM, the pipe was ®lled three times over 15 min intervals in an attempt toequilibrate the NOM with scale on the pipe wall. In the pipes never exposed toNOM, copper release was slightly lower in the old pipes than in the newer pipes (Fig.5). When 2 and 8 mg/l NOM TOC were added to the pipes, only 0.4 and 3.5 mg/lNOM TOC remained in solution after stagnation, the remainder presumably sorbedto the wall of the copper pipe. Comparing new and old pipes with NOM, the ob-served slope of 0.37 mg/l ``extra'' copper per mg/l NOM TOC in solution is very closeto the 0.40 mg Cu/mg NOM predicted based on complexation models. However,comparing release from the pipes with ``no NOM'' to ``low NOM'' indicates a muchhigher slope, consistent with a higher solubility solid phase.

The key implications of these ®ndings are synthesized in Fig. 6. The concentrationof soluble copper equilibrated with Cu(OH)2 solids and, by implication, released tocopper pipe is [7]

Soluble Cu � Cu2� � Cu2�-hydroxide� Cu2�-carbonate� Cu2�-NOM: �5�The log K of Cu(OH)2 solids decreases from 9.36 down to 7.4 as solids age [10]. At agiven time in the pipes life, the actual log K will tend to be higher in systems held atlower pH and lower temperature as indicated by Hidmi et al. [10], or in the presenceof NOM as illustrated in this work (Fig. 4). This in turn increases the concentrationof free copper and cupric hydroxide complexes after equilibration in pipes. In thissimplistic model, the upper line in Fig. 6 denotes the minimum copper corrosion by-product release in the presence of Cu(OH)2 solids of the indicated solubility.

Depending on the NOM and bicarbonate present, this minimum will be increaseddue to formation of carbonate and NOM complexes (Fig. 6). For Cu(OH)2 with alog K of 8.56, which seems close to that estimated for relatively new pipes in water

12 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 13: Organic matter and copper corrosion by-product release: a mechanistic study

distribution systems [7], each 50 mg/l of CaCO3 alkalinity at pH 7.0 will add 0.75 mg/l soluble copper in carbonate complexes. Assuming the NOM described in this workis a suitable model for those present in other waters, each mg/l of NOM TOC willadd another 0.38 mg/l copper organic matter complexes. In contrast, in the presenceof low solubility tenorite solids, every 50 mg/l alkalinity adds only 0.08 mg/l copperand every mg/l NOM adds only 0.06 mg/l copper (Fig. 6). Thus, in well aged pipes,corrosion by-product release is likely to be very low and not strongly impacted byNOM or alkalinity.

As a ®nal point, the other key solid phase thought to be an important scale incopper pipes is malachite �Cu2�CO3��OH�2�. At pH 7.0 and above about 80 mg/lalkalinity, free copper in the presence of malachite is about equal or less than that fortenorite in Fig. 6. Thus, if malachite controls solubility, the maximum concentrationof soluble copper is expected to be very low (<0.2 mg/l) and is only a weak functionof NOM concentration and alkalinity. This is consistent with the data presented byKorshin et al. (1996) in the absence of NOM [5]. However, if traces of NOM were tocompletely prevent malachite formation, resulting in a cupric hydroxide scale in-stead, copper solubility could increase dramatically even at low levels of NOM. Thisidea has precedence in the ``trace organic'' e�ect noted by Campbell [22], in that verylow levels of organic matter completely change the type of calcite solids precipitatedon pipes. It is also consistent with the ``presence or absence'' adverse e�ect noted inlaboratory work with copper [5] and in utility monitoring data [4]. Additional workis necessary to determine whether this ``NOM inhibition of solids aging'' theory, theKorshin et al. (1996) ``colloid mobilization mechanism,'' or some combination of the

Fig. 6. Predictions for soluble copper in the presence of various cupric hydroxide phases (above) and in

the presence of key complexing agents in drinking water (below) ± model prediction: 20°C, I � 2 mM.

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 13

Page 14: Organic matter and copper corrosion by-product release: a mechanistic study

two theories is key to NOM impacts on copper corrosion by-product release in highalkalinity waters.

3.3. Mechanisms leading to reversal of corrosion by-product release

During the pipe rig tests, intensive sampling was conducted after holding water inthe pipe at various stagnation times to obtain by-product release pro®les. In general,the concentration of soluble copper released to the water increased and graduallyleveled o� at a maximum value, similar to the expectations for surface activated ordi�usion controlled dissolution processes. However, in two situations, by-productrelease would rise, reach a peak, and then decrease rapidly.

The ®rst of these situations occurred in pipe rig tests with soluble NOM. To il-lustrate the mechanisms involved, 1 l of synthesized drinking water solution wasrecirculated (at 1 gpm) in a closed pipe loop while holding pH constant at 7.2. 5 mg/lDOC was spiked to the solution at t � 0 min and DO was saturated throughout theexperiment. With time and as the pipe corroded, DOC in the water decreased (Fig.7), presumably due to sorption of NOM onto freshly formed Cu(OH)2 solids on thepipe wall. This resulted in a steadily decreasing solution complexation capacity, sinceorganic ligands were being removed from solution. The net result is that soluble

Fig. 7. As water is recirculated through the pipe, DOC is removed from solution (above); this decreases

complexation capacity (below), and ultimately decreases soluble copper release; conditions are constant

pH 7.2, alkalinity 5 mg/l and 1 year pipe age.

14 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 15: Organic matter and copper corrosion by-product release: a mechanistic study

copper in the loop rapidly rose to that predicted by the solubility model, and thendecreased rapidly as organic ligands were removed from solution (Fig. 7). This isanalogous to the experiment illustrated in Fig. 3, except that particulate copper isremoved onto the pipe wall instead of via ®ltration.

The second phenomenon occurred only in pipes that had biodegradable organicmatter present and in which the DO concentration rapidly decreased to 0 (Fig. 8). Insuch circumstances, very sharp peaks in copper by-product release occurred at thepoint where DO dropped to nearly 0 mg/l, after which time copper steadily de-creased. This phenomenon was most recently reported by Werner et al. (1994) [12],who hypothesized that when DO disappeared, previously released cupric ion couldbe reduced to form less soluble cuprite [Cu(OH)], which subsequently precipitatedonto the pipe wall.

We attempted to reproduce this phenomenon in well controlled beaker experi-ments. The general protocol involved addition of 1 mg/l as Cu2� to 1 mM NaNO3

solutions which also contained 30 g of copper metal (Cu0). The oxygen was thenremoved by purging the sample with a 95% N2=5% H2 mixture. Multiple experimentsusing this arrangement illustrated no detectable decrease in Cu2� concentration.After trying many di�erent approaches, it was discovered that when 10ÿ3 M NaClwas also present as a catalyst, soluble and free copper decreased rapidly when DOwas removed from solution. In sum, in the absence of DO, and in the presence ofcopper metal and a Clÿ catalyst, copper corrosion by-product release can be readilyreversed apparently in accordance with the hypothesis of Werner et al. (1994). Therole of organic matter is to promote microbial growth which, in turn, leads toreduced DO necessary to drive the reaction.

Fig. 8. During stagnation in pipes containing high concentrations of biodegradable organic matter (2%

alginate), copper levels peak after about 6 h and then plummet after DO is completely consumed (above);

the same trend was reproduced in batch tests, but only copper metal and chloride were present (below).

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 15

Page 16: Organic matter and copper corrosion by-product release: a mechanistic study

3.4. Overall summary of organic matter e�ects

Overall e�ects of organic matter on copper corrosion by-product release can beconceptualized quite simply (Fig. 9). In the presence of oxygen, copper metal cor-rodes forming a solid scale layer of Cu(OH)2 or malachite. This scale layer equili-brates with the water, controlling the free-copper concentration. Free copper canreact with organic matter in solution, forming either soluble complexes, particulatecomplexes, or precipitates depending on the nature of the organic ligand. All of thesereactions tend to increase copper corrosion by-product release. The interaction oforganic matter with scale, and interruption of solids aging, tends to maintain higherfree-copper concentrations in solution than in cases where NOM is completely ab-sent. This increases the concentration of both organic and inorganic copper com-plexes.

Under some circumstances, the presence of organic matter can lead to reversal ofcopper corrosion by-product release. If bacteria can utilize the organic matter andremove all of the DO (or DO is otherwise consumed), soluble copper concentrationswill decrease if chloride or other suitable catalysts are present. Similarly, sorption ofsoluble NOM onto freshly formed copper surfaces can decrease the solution com-plexation capacity, causing re-deposition of previously released cupric ion onto thepipe wall. Under at least some circumstances, NOM can sorb onto scale surfaces andcause detachment of particulate copper due to development of unfavorable surfacecharge or other dispersive e�ects, but it is unclear whether this e�ect is signi®cantunder circumstances typical to drinking water systems [5].

4. Conclusions

Thus based on the analysis done above, we arrive at the following set of con-clusions:

(1) NOM in drinking water can increase copper corrosion by-product release bycomplexation and/or colloid mobilization/dispersion.

Fig. 9. Mechanisms by which organic matter in¯uences copper corrosion by-product release.

16 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18

Page 17: Organic matter and copper corrosion by-product release: a mechanistic study

(2) Even at trace soluble NOM concentrations, incorporation of NOM into a pipescale layer will interfere with natural aging processes, maintaining higher free-copperconcentrations in drinking water. This e�ect will tend to increase the soluble con-centration of copper hydroxide, copper carbonate and copper±NOM complexes in awide range of waters.

(3) Colloid mobilization is certainly possible, but seems to be most important onlyat very high levels of NOM and in the presence of Cu(OH)2 solids.

(4) The presence of organic matter can reverse copper corrosion by-product re-lease during stagnation by

(1) serving as a food source for bacteria, leading to complete removal of DOand subsequent re-deposition of copper onto the pipe wall in the presence ofchloride or other catalysts;(2) gradual sorption of soluble natural organic matter onto scale on copperpipe surfaces, decreasing the soluble copper complexation capacity of waterand leading to reduced copper concentrations after long stagnation times.

Acknowledgements

This work was supported by the National Science Foundation (NSF) under grantBES-9729008. The opinions, ®ndings, conclusions or recommendations are those ofthe authors and do not necessarily re¯ect the views of NSF.

References

[1] H. Kristiansen, VATTEN 38 (1982) 181.

[2] M. Edwards, S. Reiber, J.F. Ferguson, J. AWWA 86 (7) (1994) 74.

[3] J. Rehring, M. Edwards, Corrosion 52 (4) (1996) 307.

[4] D.M. Dodrill, L. Hidmi, M. Edwards, Proceedings of the Toronto AWWA National Conference,

American Water Works Association, Denver, CO, 1996, p. 719.

[5] G.V. Korshin, S.A. Perry, J.F. Ferguson, J. AWWA 88 (7) (1996) 36.

[6] A.E. Broo, B. Berghult, T. Hedberg, Corros. Sci. 40 (1998) 1479.

[7] M. Edwards, T.M.R. Schock, T.E. Meyer, J. AWWA 88 (3) (1996) 81.

[8] H. Holmstrom, E. Swedling, VATTEN 53 (2) (1997) 191.

[9] M.R. Schock, D.A. Lytle, J.A. Clement, Research Report, USEPA, O�ce of Research and

Development, Risk Reduction Engineering Laboratory, Cincinnati, Ohio, EPA/600/R-95/085, June

1995.

[10] L. Hidmi, M. Edwards, Environ. Sci. Technol. 33 (1999) 2607±2610.

[11] J.W. Patterson, R.E. Boice, D. Marani, Environ. Sci. Technol. 25 (1991) 1780±1787.

[12] W. Werner, et al., Wasser. Abwassen 135 (2) (1994) 92.

[13] M. Schock, Environ. Monit. Ass. 15 (1994) 59.

[14] D. Dodrill, M. Edwards, J. AWWA 87 (7) (1995) 74.

[15] M. Edwards, M.M. Benjamin, J. AWWA 84 (6) (1992) 56.

[16] G.G. Geesey, T. Iwaoka, P.R. Gri�ths, J. Coll. Interface Sci. 120 (2) (1987) 370.

[17] APHA. Standard Methods for the Examination of Water and Wastewater, 20th ed., United Book

Press, Baltimore, Maryland, 1998.

M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18 17

Page 18: Organic matter and copper corrosion by-product release: a mechanistic study

[18] E. Lind Johansson, Diss., Department of Sanitary Engineering, Chalmer University of Technology,

Goteborg, Sweden, 1989.

[19] M. Edwards, J. AWWA 89 (5) (1997) 78.

[20] S.E. Canabiss, M.S. Shuman, Geochim. Cosmochim. Acta 52 (1998) 185.

[21] M. Edwards, M.M. Benjamin, J. AWWA 84 (8) (1992) 63.

[22] H.S. Campbell, M.E.D. Turner, J. Inst. Water. Engng. Sci. 4 (1983) 55.

18 M. Edwards, N. Sprague / Corrosion Science 43 (2001) 1±18