Efeitos do fluxo, a localização de latão, materiais do tubo e da temperatura sobre a corrosão dos dispositivos de canalização de latão.pdf

Corrosion Science 53 (2011) 1813–1824

Contents lists available at ScienceDirect

Corrosion Science

journal homepage: www.elsevier .com/ locate /corsc i

Effects of flow, brass location, tube materials and temperature on corrosionof brass plumbing devices

Emily Sarver ⇑, Marc Edwards 1

Department of Civil and Environmental Engineering, 418 Durham Hall, Virginia Tech, Blacksburg, VA 24061, USA

a r t i c l e i n f o

Article history:Received 20 October 2010Accepted 29 January 2011Available online 4 February 2011

Keywords:A. BrassB. Weight lossB. SEMC. De-alloyingC. Pitting corrosion

0010-938X/$ - see front matter � 2011 Elsevier Ltd. Adoi:10.1016/j.corsci.2011.01.060

⇑ Corresponding author. Permanent address: DepartEngineering, 108 Holden Hall, Virginia Tech, Blacksbur231 8139; fax: +1 540 231 4070.

E-mail addresses: [email protected] (E. Sarver), edwa1 Tel.: +1 540 231 7236; fax: +1 540 231 7916.

a b s t r a c t

Effects of plumbing-specific installation factors on brass corrosion were investigated in a series of pipe-loop experiments. Increased flow velocity increased corrosion rates, but did not affect corrosion type. Thepresence of copper tubing in the plumbing system increased selectivity of brass corrosion for zinc, unlessa galvanic connection was made between copper and brass, in which case corrosion became more uni-form and was accelerated. Plastic tubing allowed oxidant (i.e., free chlorine) to persist in water, increasingbrass exposure. Additionally, hot water significantly increased lead leaching from brass. These findingsmay inform future investigations into brass corrosion issues and plumbing designs.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Brass is a copper–zinc alloy, which contains small amounts oflead, iron, and other trace elements. In premise plumbing systems,brass devices (e.g., valves, couplings, faucets) are commonly useddue to their relative durability, machinability, intricate functional-ity and low cost. Although brass corrosion does not present seriousproblems in most systems, certain types of non-uniform corrosioncan significantly impact performance in some cases. Most notably,dezincification is a type of de-alloying corrosion wherein zinc isselectively dissolved, either from small (i.e., plugs) or large areas(i.e., layers) that can penetrate deep into the brass surface. Dezin-cification can leave brass porous and subject to leaks, and can alsolead to clogged water lines and burst failures when voluminouscorrosion deposits called ‘‘meringue’’ build-up within affected de-vices. On top of economic consequences, water damage can resultin harmful mould growth [1]. Brass pitting is another type of non-uniform corrosion, which may also result in leak failures [2]. Final-ly, waters that support dezincification or other brass corrosion mayalso cause lead leaching, which presents a potential health hazardif tap water is contaminated [3–7].

Red (i.e., low-zinc) and dezincification resistant (i.e., ‘‘DZR’’ or‘‘inhibited’’) brasses are made for use in plumbing [8], but thesealloys are not always utilised due to their relatively high costs,

ll rights reserved.

ment of Mining and Mineralsg, Va 24061, USA. Tel.: +1 540

[email protected] (M. Edwards).

limited availability, or unfamiliarity. Recently, use of inexpensive,yellow (i.e., high-zinc) brasses has resulted in widespread dezinci-fication failures in particular areas and product liability concerns(e.g., [9]). ‘‘Lead free’’ brasses are now required in many potablewater applications [10], but these alloys may contain up to 8% lead(by weight) and may not be adequately reducing lead leaching[11,12]. There is also some evidence that DZR brasses may be moreprone to lead leaching than un-inhibited brasses [7,13], and nei-ther DZR nor ‘‘lead-free’’ brass specifically offers protection againstpitting. Thus, while problems are relatively rare, they do occur.

In potable water systems, the type and extent of brass corrosionmay often be controlled by alloying constituents and metallurgy,and water chemistry (e.g., pH, alkalinity, chloride, oxidant) [2,13–17]. However, corrosion may also be influenced by physical orphysiochemical factors associated with specific installations,including flow conditions, location of brass within a system, pre-dominant tubing materials, galvanic connections, and watertemperature.

1.1. Flow and brass location

The effects of flow on brass corrosion are not straightforward.Although texts often state that dezincification is promoted byslowly moving or stagnant water at the affected surface (e.g.,[18,19]), practical observations reveal that the corrosion may beaccelerated by flowing conditions (e.g., [20,21]). As explained byKelly et al. [22], this discrepancy may be due to physical separationof anodic and cathodic sites on brass surfaces. If, for example,relatively small anode sites become occluded beneath meringuedeposits, these sites would be shielded from bulk water flow

Fig. 1. Hypothesized effects of differential flow and galvanic connection (to copper)on brass corrosion. Water flows through the copper tube, but is relatively stagnant

1814 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

conditions; whereas, larger cathode sites may be exposed to thebulk water conditions. In that case, increasing flow velocity wouldeffectively increase the supply of oxidant (e.g., free chlorine or dis-solved oxygen) to the cathodes2, aiding in development of a differ-ential concentration cell and driving dezincification at the anode.Separation of anodes and cathodes may also occur on a much largerscale in plumbing systems if the location of brass within a systemcauses it to be subjected to differential flow. Such a scenario can existwhen a brass ‘‘T-connection’’ branches one water line off of another,such that one section is subject to flow while another may not be, asis commonly the case in building plumbing systems. Furthermore,flowing conditions could promote selective lead leaching from brass,which may be mechanistically analogous to dezincification (i.e., dueto micro-galvanic currents between dissimilar alloy constituents).Additionally, increased flow could contribute to brass pitting – sim-ilar to observations of some copper pitting phenomena (e.g., [24]).

inside the brass. As zinc (or other alloy constituents) is oxidized, electrons areliberated at the brass surface and accepted at the copper surface where an oxidant(e.g., free chlorine as HOCl) is reduced. pH drops in the stagnant water due to Lewisacidity of metal cations (e.g., Zn2+), salt content increases as chloride (Cl�) and otheranions draw into satisfy electroneutrality, and corrosion deposits (e.g., meringue)may build-up on the brass surface. If flow velocity is increased, or oxidantconcentration is increased in the flowing water, brass corrosion is expected toaccelerate.

1.2. Plumbing tube material

Modern plumbing systems primarily utilise either copper orplastic (e.g., PEX) tubing [25], and the effects of the tube materialon bulk water chemistry in a system may impact corrosion of brassdevices. For example, while copper tubes can significantly con-sume oxidants in potable waters, plastics tubes cannot. Thus, longlengths of copper tubing might dramatically reduce free chlorineresidual3 of the water before it even reaches some brass components(e.g., a faucet), perhaps effectively reducing overall corrosion activityof those components. In contrast, plastic tubing can allow chlorine topersist in water for much longer periods of time, increasing the like-lihood of high dose and duration of exposure. It is also possible thatcopper released from copper tubing may limit copper leaching frombrass components through equilibrium and mass transport phenom-ena, thereby increasing selective zinc leaching from brass.

In the case of copper tubing, brass corrosion can also be affectedby galvanic connections between the two metals, which are com-monplace in plumbing systems. Given the typically small wetted-surface-area ratios between brass and copper in these systems,and that copper is usually noble to brass, galvanic currents maysignificantly accelerate brass corrosion. With respect to overallcorrosion rate, a worst-case scenario may be represented by a cop-per-brass connection oriented such that the brass is additionallysubject to differential flow conditions (Fig. 1).

1.3. Temperature

Temperature might also play a role in the type(s) and extent ofbrass corrosion that occurs in potable water systems. While it is ac-cepted that high temperatures can worsen dezincification-relatedfailures [28] via precipitation and accumulation of meringue[23,29], or acceleration of zinc leaching [2,13], effects of tempera-ture on other types of brass corrosion are not well understood un-der the chemical and physical conditions that may be encounteredin real plumbing systems. For instance, some field data suggeststhat lead levels can be higher in hot water tap samples than in cold

2 Dezincification of brass in potable water is often cathodically-limited [23].3 Free chlorine is common disinfectant for potable water supplies. Relatively high

concentrations may be dosed at a treatment facility for immediate disinfection, andresidual concentrations provide subsequent protection against microorganismsthroughout a water distribution system. In addition to disinfection and auto-decomposition, chlorine may be decayed (i.e., to Cl-) via redox reactions associatedwith corrosion of plumbing materials (e.g., brass, copper, iron) due to its highoxidizing power. The concentration of free chlorine (e.g., as mg/L) at any point in thesystem is referred to as the ‘‘free chlorine residual’’ at that location; in the U.S.,residuals may not exceed 4 mg/L in potable water systems [26], but even lowerconcentrations can contribute to significant metals corrosion over time (e.g., [27]).

when brass components are present (e.g., [30]), and it is highlydesirable to understand if lead-bearing brass is a probable culprit.

The objective of this work is to gain a broader understanding ofthe practical effects that the above factors of installation may haveon the type(s) and extent of brass corrosion that may occur in po-table water systems of specific buildings.

2. Materials and methods

Three experiments were conducted (Table 1) in pipe-loop appa-ratuses (Fig. 2). The first consisted of short-term ‘‘proof of concept’’tests (i.e., tests A–C) to study the hypothesized scenario in Fig. 1,while the second and third consisted of longer-term tests (i.e., testsD–K) designed to isolate the effects of individual factors on corro-sion of brass plumbing devices.

2.1. Pipe-loops

Pipe-loops were constructed using new C36000 (i.e., high-zinc,un-inhibited) brass fittings, effectively smooth-walled brass tubingin the loops, and C12200 copper tubing (where specified). Thebrass fittings were either couplings or caps that would practicallybe utilised with PEX, or a combination of PEX and copper, plumb-ing tube systems (Fig. 3). The actual composition of the brass alloy(by weight) was determined by a handheld X-ray fluorescence(XRF) instrument (Innov-X Systems, model Alpha 8000): 58.8%Cu, 37.2% Zn, 3.1% Pb, 0.3% Sn, 0.3% Fe and 0.1% Ni. Prior to assem-bling the pipe-loops, copper tubing was cut and de-burred, and allcomponents were rinsed with de-ionised water, allowed to dry andweighed.

All pipe-loops had continuously re-circulating flow of test water(chemistries and velocities in Table 1) from polypropylene plasticreservoirs (30L in exp. 1, 94 L in exp. 2 and 3), which were closedto limit atmospheric influences on water quality (e.g., changes inpH, temperature). All brass and/or copper sections were physicallyseparated using short lengths of clear vinyl tubing, but were elec-trically connected via external copper bridge wires (except in testsE and J as specified in Table 1). This allowed for collection ofelectrochemical data by temporarily replacing the wires with amulti-meter (as illustrated in Fig. 2). Magnet-drive centrifugalpumps (polypropylene construction) were used to re-circulate

Table 1Test conditions for three experiments.

Test Duration days Pipe-loop Flow velocity(m/s)

Water chemistry parameters Water change cycle

Maintained InitialTarget value (±standard deviation)

Experiment 1A 14 Loop 1 with variable free Cl2 2.1 Variable Cl2

a pH 7.6 (±0.3)11 mg/L alkb (±3)38 mg/L Na+ (±6)50 mg/L Cl� (±5)

–B 14 Loop 1 2.1 4 mg/L Cl2

a (±0.8) WeeklyC 21 Loop 1 with brass located in-flow (no T) 2.1 Weekly

Experiment 2D 490 Loop 2 1.3 pH 9.2 (±0.3)

4 mg/L Cl2a (±0.5)

34 mg/L alkb (±3)16 mg/L Na+ (±3)20 mg/L Cl� (±3)17 mg/L Ca2+ (±2)13 mg/L SO4

2� (±3)

Every other weekEvery other weekEvery other weekEvery other weekEvery other week

E 490 Loop 2 without galvanic connection to copper 1.3F 490 Loop 2 with PVC in-flow (instead of copper) 1.3

Experiment 3G 150 Loop 3 1.9 pH 8.3 (±0.2)

4 mg/L Cl2a (±0.5)

34 mg/L alkb (±3)29 mg/L Na+ (±5)40 mg/L Cl� (±5)17 mg/L Ca2+ (±2)13 mg/L SO4

2� (±3)

Every other weekEvery other weekEvery other weekEvery other weekEvery other week

H 150 Loop 3 at low velocity 0.6I 87 Loop 3 at 45 �C 1.9J 150 Loop 3 without electrical connection 1.9K 150 Loop 3 with copper in-flow (instead of brass) 1.3

a Free chlorine residual.b Alkalinity as CaCO3.

Fig. 2. Pipe-loops for three experiments. Flow direction is indicated by arrows.

Fig. 3. New C36000 brass coupling (left) and cap (right). Left photographs show thefittings as installed in the pipe-loops; right show cross-sections to display smooth,interior shape.

E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824 1815

water, except in test H (exp. 3), which had a small submersiblepump (polypropylene and stainless steel construction) that pro-vided lower flow velocity.

In experiment 1, the pipe-loops each had approximately 60 cm(total length) of Type M copper tubing (1.9 cm inner diameter) lo-cated in-flow, and one brass coupling (4.5 cm long, 1.75 cm inner

diameter). For tests A and B, the brass was located out-of-flowand plugged with a silicone stopper (i.e., as shown in Fig. 2, Loop1) to create differential flow conditions; for test C, however, thebrass was located in-flow, midway between two equal lengths ofcopper, such that water flowed straight through all sections of tub-ing. In experiment 2, each pipe-loop had one brass coupling(dimensions above) which was located out-of-flow and pluggedwith a silicone stopper (as shown in Fig. 2, Loop 2). Located in-flowwas approximately 91 cm (total length) of Type M copper (1.9 cminner diameter) in tests D and E; in test F, instead of copper, thesame length of equal-diameter of polyvinyl chloride (PVC) tubingwas used. In experiment 3, each pipe-loop had one brass cap(1.3 cm long, 1.3 cm inner diameter) located out-of flow (i.e., no sil-icone plug was needed to create the dead-end), and one brass cou-pling (dimensions above) located in-flow (as shown in Fig. 2, Loop

1816 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

3). The only exception was test K, in which the brass coupling wasreplaced with a Type M copper tube of equal length (1.9 cm innerdiameter). From this point forward, brass samples located in-flowin a pipe-loop are referred to as ‘‘in-flow,’’ and samples locatedout-of-flow are referred to as ‘‘T’’ due to their orientation at thedead-end of a T-connection.

2.2. Water chemistry

In order to maintain relatively constant water chemistries in thepipe-loops, test waters were changed regularly (Table 1). Just priorto each water change, the test waters were synthesised using de-ionised water and reagent grade sodium and/or calcium salts. Freechlorine was added as sodium hypochlorite from a concentratedbleach stock (i.e., 6% NaOCl solution). Where applicable, pH wasadjusted using NaOH and HNO3 stock solutions. pH was measuredusing a double-junction Ag–AgCl electrode, and free chlorine resid-ual was measured with a Hach Chlorine Pocket Colorimeter II usinga DPD (diethyl phenylene diamine) colorimetric test per StandardMethod 4500-Cl G [31]. Except for in test I (exp. 3), water temper-atures were governed by ambient room temperature and the heatgenerated by the pumps; temperatures ranged from 24–33 �C,which is within the ‘‘cold water’’ range for premise plumbingapplications. In test I, temperature was maintained by an immer-sion heater at 45 �C, which is considered ‘‘hot water’’.

In experiment 1, the synthesised potable water chemistry had arelatively high chloride to alkalinity ratio with a high dezincifica-tion propensity [13]. The water made using only sodium salts(i.e., it had no hardness), and had a relatively stable pH of 7.6,which was not purposefully maintained. Chlorine was not initiallydosed to test A, but on day 3, free chlorine was dosed to the waterto achieve a 4 mg/L residual concentration (as Cl2). This was man-ually maintained until day 6, at which time the chlorine was al-lowed to decay until the end of the test. In tests B and C,chlorine was manually maintained over the entire test duration.

The synthesised potable water chemistry in experiment 2 hadhigh pH, high free chlorine residual and low alkalinity – a combi-nation known to cause pitting of copper [24] and thus also ex-pected to attack brass. A similar water chemistry was used inexperiment 3, except that it had lower pH (i.e., 8.3 instead of 9.2)and higher chloride (i.e., 40 instead of 20 mg/L). These conditionswere expected to produce severe meringue dezincification basedon prior studies (e.g., [13,16]). In experiments 2 and 3, pH and chlo-rine targets were maintained via automatic feed systems in con-junction with manual adjustment (at least four times per week).

2.3. Data and analysis

Corrosion currents (I) between metal sections were monitoredfor all tests having metals in electrical contact (i.e., all tests exceptE, F, and J). Measurements were taken at least once per waterchange cycle using a Fluke 189 True RMS multi-meter (internalresistance <100 X). In the convention utilised for this work, posi-tive I values indicate anodic behaviour of a given metal sectionvs. all other metal in a system. Additionally, re-circulating bulk(i.e., from the pipe-loop reservoirs) and stagnant (i.e., from thedead-end T sections shown in Fig. 2) water samples were collectedon a regular basis (i.e., weekly in exp. 1, every 6 weeks in exp. 2,and every 4 weeks in exp. 3), and metals concentrations (i.e., Zn,Cu, and Pb) were determined via inductively couple plasma massspectrometry (ICP-MS).

At the end of each test, brass samples were subject to forensicinspection. They were removed from pipe-loops and allowed todry prior to removal of corrosion scales and cross-sectioning for vi-sual and surface analyses. Copper enrichment (i.e., the relative in-crease in copper weight fraction of the brass surface after testing),

and zinc and lead depletion (i.e., the relative decrease in the zinc orlead weight fraction of the brass surface after testing) were deter-mined by comparing elemental analyses (obtained via XRF) ofun-exposed brass surface areas with (cleaned) areas which wereobviously corroded during testing. The XRF instrument analysedan area of approximately 8 mm2, to a penetration depth of 1 lmor more. Enrichment and depletion were computed using Eq. (1)and (2):

EM ¼ 100� MC �MU

MU

� �ð1Þ

DM ¼ 100� MU �MC

MU

� �ð2Þ

where EM and DM are enrichment or depletion, respectively, of a gi-ven element (e.g., Zn) during testing (expressed as%); MC is theweight percentage of that element on the corroded brass surface;and MU is the weight percentage of that element on the brass sur-face not exposed to test water. Selected samples from experiment3 were also imaged and analysed using an FEI Quanta 600 FEG envi-ronmental scanning electron microscope (ESEM), equipped with aBruker Quantax 400 energy dispersive X-ray spectroscope (EDX)(penetration depth of 1 lm). This allowed for precise geometricmeasurements and elemental analyses of specific regions of interest(e.g., pits, areas of plug dezincification). Eq. (1) and (2) also wereused to calculate metal enrichment and depletion values for theseregions from the ESEM–EDX, which had surface areas of about0.005–0.05 mm2. Additionally, corrosion scales from experiments2 and 3 were dissolved and analysed for metals contents via ICP-MS.

3. Results

Dezincification was observed in all tests, though the severity ofattack and particular manifestations (i.e., layer, plug and/or merin-gue) varied (Table 2). Brass pitting was also observed in some tests(i.e., G, H, and K in exp. 3), and selective lead leaching was evidentfrom water, scale and brass surface analyses in another (i.e., I, exp.3). The sections below provide detailed results from each of thethree experiments.

3.1. Experiment 1: proof of concept

Results from experiment 1 confirmed that the general hypothe-sis presented in Fig. 1 (i.e., acceleration of brass corrosion via acombination of differential flow and galvanic connection to cop-per) is valid. In test A, the galvanic corrosion current density (brassT vs. copper in-flow) was closely related to the free chlorine con-centration in the bulk water (Fig. 4), demonstrating that corrosionon the anodic brass surface was affected by the supply of oxidantto the cathodic copper. When chlorine dosing began on day 3,the current jumped and then rose over about a two-day period,reaching a peak of nearly 14 lA/cm2; when chlorine dosing wasstopped on day 7, the current fell slowly with chlorine residualand stabilised around 0.8 lA/cm2 on day 12. No meringue formedin this test (or other tests in exp. 1), but based on metal leaching(not shown), visual observations and XRF data (Table 2), the mea-sured galvanic corrosion currents can be attributed to dezincifica-tion of the brass.

In test B, where chlorine concentration was maintainedthroughout, comparison of the stagnant water localised insidethe brass T to the bulk water proved that a large-scale differentialconcentration cell did indeed develop, as hypothesised. Watersamples were taken on days 7 and 14, and on both occasions thewater in the T had lower pH (i.e., by 0.4 pH units) and lower freechlorine concentration (i.e., by 1.1–1.3 mg/L) than the bulk water.

Table 2Summary of selected results for all tests.

Test Current densitya

(T vs. In)Weight loss(%/year)

Observedb corrosiontype(s)

Surface Zndepletionc (%)

Surface Cuenrichmentc (%)

Surface Pbenrichmentc (%)

Scale Zn contentd

(Weight%)Scale Cu contentd

(Weight%)Scale Pb contentd

(Weight%)

lA/cm2 95% CI T In. T In. T In. T In. T In. T In T In T In.

Experiment 1A NA LD (shallow) NA 18.1 NA 10.6 NA 41.5 NA ND ND ND ND ND NDB 5.9 1.0 ND LD (shallow) NA ND ND ND ND ND ND ND ND ND ND ND NDC 3.2 0.5 NA LD (shallow) ND ND ND ND ND ND ND ND ND ND ND NDExperiment 2D 0.9 0.1 1.2 NA MD,

LD (shallow)NA 35.9 NA 22.9 NA 26.6 NA 47.6 NA 4.7 NA 2.9 NA

E NA 1.0 NA MD,LD (shallow)

NA 69.5 NA 42.5 NA 88.3 NA 57.1 NA 0.6 NA 2.3 NA

F NA 0.9 NA MD,LD (shallow)

NA 53.9 NA 35.3 NA 23.6 NA 56.8 NA 0.1 NA 2.6 NA

Experiment 3G 0.8 0.5 1.3 1.9 BP, PD,

LD (shallow)LD (shallow) 30.6 51.8 20.1 30.9 �15.8 19.0 48.8 46.9 2.0 4.1 1.0 2.6

H 0.2 0.2 0.8 1.8 LD (shallow) BP,LD (shallow)

29.8 47.5 18.4 31.0 10.4 �8.6 49.3 52.1 5.6 3.1 1.4 2.2

I 0.5 1.3 1.5 2.1 MD, PD,LD (deep)

MD,LD (shallow)

19.0 28.9 13.9 21.4 �43.5 �32.7 64.6 47.9 5.4 1.0 4.4 3.1

J NA 1.3 2.0 PD,LD (deep)

LD(shallow)

35.1 37.0 21.0 23.3 14.8 22.1 58.9 40.0 5.1 7.8 1.3 2.7

K 2.7 2.0 1.7 NA BP,LD (deep)

NA 16.3 NA 8.4 NA 10.4 NA 44.8 NA 12.2 NA 1.4 NA

T Brass located out-of-flow, In. Brass/copper located in-flow, NA Not applicable, ND Not determined, LD Layer dezincification, MD Meringue dezincification, BP Brass pitting, PD Plug dezincification.a Average over duration (exp. 1, 2) or week 8 through end of test (exp. 3).b Based on visual inspection of cross-sectioned samples.c Computed from XRF data; negative enrichment values indicate depletion of element from brass surface.d Computed from ICP-MS data as the weight% of element contained in solid scale removed from brass sample following testing.

E.Sarver,M.Edw

ards/CorrosionScience

53(2011)

1813–1824

1817

0

5

10

15

Day

Anod

ic C

urre

nt D

ensi

ty (u

A/cm

2 )

0

2

4

6

Bulk Chlorine C

oncentration (mg/L)

CurrentChlorine

A B C

0 4 8 12 0 4 8 12 0 4 8 12 20 16

Fig. 4. Brass current density (vs. in-flow copper) and bulk chlorine residual over time for tests A, B and C.

Fig. 5. Brass T’s from tests in experiment 2. Meringue scale build-up is shown onthe left; non-uniform layer dezincification (revealed after scale removal and cross-sectioning) is on the right.

Table 3T water chemistry and free chlorine consumption rates for tests in experiment 2.

Test Average T water chemistry Average Cl2

consumptiona

pH Cl2 Cl� Zn Cu Pb

(mg/L) (mg/h)

D 8.6 1.4 54 280 417 1.3 7.8E 8.7 2.0 44 210 580 0.5 8.7F 8.9 2.5 9 208 2 1.2 0.5

a Rate of chlorine addition required to maintain target concentration in bulkwater.

1818 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

Comparison of current densities from tests B and C (Fig. 4) addi-tionally indicated that dezincification was more severe when thebrass was located out-of-flow (as depicted in Fig. 1, Loop 1) thanwhen it was located in-flow. The average current density wasabout 5.9 lA/cm2 in test B, but was only about 3.2 lA/cm2 in test C.

Brass from tests A, B, and C all looked very similar upon removalof corrosion scales. The entire exposed surface was reddened indi-cating that layer dezincification occurred; however the depth of

dezincification into the brass wall was too shallow to measure, atleast on the two planes (i.e., wall thicknesses) revealed by cross-sectioning the brass samples lengthwise. Significant lead leachingwas not observed in any of these short-term tests.

3.2. Experiment 2: plumbing tube materials

In experiment 2, the influences of plumbing tube materials onbrass corrosion were tested. Meringue build-up was quite severein all tests D–F (Fig. 5), but was visually worst in the test (F) in whichPVC tubes were installed in-flow instead of copper. While this testcondition produced the least brass weight loss (Table 2), it main-tained the highest pH levels in the stagnant T water (Table 3); sothe increased build-up of meringue was likely due to reduced solu-bility despite the lower overall corrosion rate. In contrast, test D,where the brass T was galvanically connected to copper tubing in-flow, had the least meringue build-up, the greatest brass weight loss,the greatest pH drop in the stagnant water, and the greatest averagezinc leaching (Fig. 6). In test E, where copper tubes were located in-flow but were not galvanically connected to the brass T, most resultsfell somewhere in between tests D and F.

Beneath the meringue and other corrosion scales, the brass sur-faces in all three tests exhibited shallow layer dezincification,which covered large non-uniform areas (Fig. 5). But, like in exper-iment 1, the depth of dezincification was too small to be quantified.Somewhat unexpectedly, brass pitting was not observed in thisexperiment even though the copper tubes in-flow (tests D and E)were severely pitted and began exhibiting pinhole leaks about200 days into the testing. Brass pitting was, however, observed inexperiment 3 (discussed below) where waters had lower pH andthere was less meringue formation.

As anticipated, chlorine consumption was impacted by the pre-dominant tube material in the pipe-loops. The chlorine consump-tion rate with PVC tubes in-flow (test F) was at least 15 timeslower than the rate with copper tubes in-flow (test D and E), andthe chlorine residual in the stagnant T water (i.e., to which thebrass was constantly exposed) also remained higher (Table 3).Since chlorine was expected to be aggressive with respect to corro-sion, it was somewhat surprising that the brass in test F exhibited alower weight loss than its counterpart in E (i.e., 0.9% vs. 1.0%,respectively); but the extreme difference in chloride concentra-tions in the T waters in these tests may have played a role as chlo-ride is also known to be very aggressive towards brassdezincification [16]. Chloride built-up very quickly in test E dueto the high chlorine consumption rate – which necessitated a highchlorine feed rate to the pipe-loop in order to maintain the

0

2

4

6

8

10

D E F G H I J K

Aver

age

Zinc

or C

oppe

r Ac

cum

ulat

ed (m

g/w

eek)

0

50

100

150

200

250

Average Lead Accum

ulated (ug/week)

Zn Cu Pb

Experiment 2 Experiment 3

Fig. 6. Average zinc, copper, and lead accumulated (either due to leaching or settling from bulk water) in stagnant T waters in experiments 2 and 3. The bulk water pH was 9.2in experiment 2, and 8.3 in experiment 3. Data are normalized based on T water volume and sampling frequency; error bars show 95% confidence intervals.

-100

-75

-50

-25

0

25

50

75

100

0 25 50 75 100Zn Depletion on

Exposed Surface (%)

Pb D

eple

tion

on

Expo

sed

Surfa

ce (%

)

T brassin-flow brass

Fig. 7. Zinc vs. lead depletion on exposed brass surfaces from experiments 2 (un-filled points) and 3 (filled points). Negative lead depletion values indicate leadenrichment.

E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824 1819

targeted 4 mg/L concentration – and could have increased theoverall corrosion rate of the brass in this test as compared to thatin test F, where chloride build-up was minimal.

In addition to accelerating the overall brass corrosion rate, as byevidenced by weight loss, the galvanic connection between coppertubes installed in-flow and the brass in test D also appears to haveinfluenced the selectivity of brass corrosion. The significantly higherpercentage of copper measured in the meringue from test D thanthat from E (i.e., about 4.7% vs. 0.7% by weight) suggests that the gal-vanically-influenced brass may have leached much more copperthan its uninfluenced counterpart, effectively reducing the selectiv-ity of brass corrosion for zinc (Table 2). Zinc depletion and copperenrichment data (as indicated by XRF) supported this finding, andalso showed that, conversely, the brass corrosion in test E was highlyselective for zinc – indeed more so than in any other test presentedhere. This might be explained, at least partially, by the extremelyhigh copper concentrations found in the T water from test E(Fig. 6), which were undoubtedly due to corrosion of the coppertubes installed in-flow in this test. While copper was also quite highin the T water in test D, again primarily due to corrosion of the in-flow copper tubing, it did not appear to inhibit copper leaching fromthe brass; perhaps this is because the galvanic voltage in this test wassufficiently anodic to oxidize both zinc and copper from the brass. Intest F, which was not influenced by a galvanic connection or highbackground copper concentrations, corrosion was moderately selec-tive for zinc (with respect to tests D and E), and nearly no copper wasfound in the T or bulk water, or in the meringue.

From the perspective of lead leaching, all tests in experiment 2behaved fairly similarly (e.g., Fig. 6). In addition to concentrationsof lead measured in the T and bulk waters (i.e., about 0.5–2 mg/Land 0.1–1 lg/L on average, respectively), lead was also found inthe meringue deposits in each condition. Lead ranged from about2.3–2.9% by weight, which is relatively similar to the weight frac-tion of lead in the brass alloy itself. Interestingly, it was also ob-served that zinc and lead depletion were inversely related, suchthat as brass corrosion became more selective for zinc, it becameless selective for lead, and vice versa (Fig. 7).

4 Silica has been previously observed to limit dezincification corrosion over shortdurations (e.g., [14]).

3.3. Experiment 3: other factors

In experiment 3, effects of flow, temperature and galvanic con-nection to copper were studied in head-to-head tests (G–K); test G

served as a control (i.e., high velocity; electrical connection be-tween brass T and brass in-flow to create differential flow condi-tions; cold water; no galvanic connection to copper). Beforedescribing the results, several notes should be made regarding dis-ruption to the experiment. First, the de-ionized water system uti-lised for these tests temporarily leached silica (i.e., 3–15 mg/L asSi) which briefly affected all tests in experiment 3 between the firstand second water changes (i.e., days 14–28)4. Second, a decisionwas made to interrupt chlorine dosing between the third and fourthwater changes (i.e., days 28–42), during which time it was confirmed

0

2

4

6

0 25 50 75 100 125 150Day

Anod

ic C

urre

nt D

ensi

ty (u

A/cm

2 )

0

2

4

6

Bulk Chlorine C

oncentration (mg/L)

CurrentChlorine

no Cl2

dosing

Si in DI system

Fig. 8. T brass current density (vs. in-flow copper) and bulk chlorine residual over time for test K. Periods of silica contamination and chlorine dosing interruption areindicated.

0

1

2

3

G H I J K

Wei

ght L

oss

(%/y

ear)

0

2

4

6

Average A

nodic Current D

ensity (uA/cm

2)

T brass weight loss

In-flow brass weight loss

Current density

Fig. 9. T and in-flow brass weight loss rates, and average T brass current densities (vs. in-flow brass or copper) for tests in experiment 3. Current densities represent data fromday 42 (i.e., after chlorine dosing resumed) through the end of tests; error bars show 95% confidence intervals.

1820 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

that electrochemical activity was significantly depressed in all tests– consistent with results from experiment 1. Fig. 8 shows the effectsof the above events on the brass current density in test K (galvanicconnection between brass T and copper in-flow). Additionally, testI (hot water) was only run for 87 days (vs. 150 days for other exp.3 tests); and due to the high rate of chlorine auto-decompositionin this test, the average chlorine residual in the bulk water was onlyabout 2.6 mg/L (vs. the maintained 4 mg/L target in all otherconditions).

Tests in experiment 3 produced various manifestations of brasscorrosion (Table 2). Based on weight loss measurements, corrosionwas most evident in the tests with hot water (I) or a galvanic con-nection between the brass T and copper in-flow (K), and was leastaggressive in the low velocity test (H) (Fig. 9). The control (G) andthe test where the T and in-flow brass samples were not electri-cally connected (J) produced similar weight losses, which weremoderate with respect to the other tests.

Overall, the tests with high weight losses also tended to havehigh current densities (measured between brass T vs. brass or

copper in-flow). However, the brass T’s lost less weight in every testcondition than the in-flow brass – despite the facts that the T’swere generally anodic to the in-flow samples, and that the stag-nant T waters maintained consistently lower pH than the bulkwaters (data not shown). The severity of dezincification (basedon zinc depletion and copper enrichment as indicated by XRF data)was also determined to be lower on the T than on the in-flow brassin all tests. Thus, even though corrosion on the T’s was acceleratedby their connection to in-flow metals, the overall corrosion rates ofthe T’s were still lower than those of the in-flow metals. It is con-sidered likely that this was because the in-flow metals were ex-posed to relatively high, constant chlorine concentrations,whereas chlorine levels in the brass T’s were consistently muchlower (i.e., about 2 mg/L on average) due to the differential concen-tration cells developed.

All tests produced a relatively smooth, dark grey scale, whichwas covered by at least some meringue (Fig. 10). Test I producedmeringue that covered a large portion of the T and in-flow brasssurfaces, despite its shorter run time and lower chlorine residual,

Fig. 10. T (top) and in-flow (bottom) brass from tests in experiment 3, prior to scale removal and cross-sectioning.

1100 µm

600 µm

Fig. 11. Photograph (lower left) and ESEM image (right) of pitting on in-flow brass from test H. Direction of sample cross-section is shown (upper left) and red box indicateslocation of imaging; white arrows indicate pit and dimensions are specified. This particular pit penetrated about 220 lm (i.e., roughly 15%) of the brass tube wall.

E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824 1821

which is consistent with observations by other authors that hotwater can increase meringue build-up [29]. In other tests, merin-gue deposits were round and isolated, and tended to be largerand fewer in number on the T’s than on the in-flow surfaces. Afterremoving the corrosion scales and cross-sectioning the brass, vary-ing degrees of layer dezincification were revealed on all samples.Some also exhibited plug dezincification and/or pitting. Accordingto XRF analyses, corrosion was most selective for zinc in tests G, Hand J, whereas corrosion was more uniform (i.e., lower zinc deple-tion and copper enrichment) in test K (Table 2), which is consistentwith the findings in experiment 2 (test D). Additionally, significantlead leaching was observed in test I, which is discussed furtherbelow.

All in-flow brass samples exhibited very shallow layer dezinci-fication, characterized by copper enriched surfaces (Table 2),although the depth of dezincification could not be quantified.While no plug dezincification was observed on any of these sam-ples, pitting was found on the in-flow brass from the low velocitytest (H) (e.g., Fig. 11). ESEM–EDX analyses indicated severe dezin-cification at the pit site shown in Fig. 11; zinc depletion associatedwith the pit was about 20% greater than that associated with theshallow layer dezincification observed on adjacent, un-pitted sur-faces (Table 4). However, it is not possible to know whether thepitting observed here, or that observed in other tests (i.e., G andK), occurred via dissolution of all alloy constituents at the pit sites

or via a sequence of primarily zinc dissolution followed by fractureand detachment of the residual, porous material.

The brass T samples from tests I, J, and K exhibited regions ofdeep layer dezincification (of varying size) that penetrated far en-ough into the surface to view at low magnification (e.g., Fig. 12).The deepest observed layers (i.e., on the surfaces revealed bycross-sectioning) were on the brass T from the hot water test (I),with a maximum depth of approximately 260 lm. However, themost uniform layer appeared to occur on the brass T in the galvanictest (K); upon cross-sectioning, nearly the entire surface that hadbeen exposed to the test water was dezincified to a depth ofroughly 100–200 lm. Some pitting was also found on this sample,as well as the T from the control test (G). Significant plug dezinci-fication was additionally observed on the T’s from tests G, I, and J(e.g., Fig. 13). ESEM–EDX analyses indicated that corrosion wasmore selective for zinc at sites of plug dezincification than at adja-cent sites of layer dezincification (Table 4), as was generally thecase for pitting sites.

Metal concentrations in the bulk (data not shown) and T (Fig. 6)waters were quite variable over the test duration. On average, highlevels of zinc were found in the T waters for tests H, I, and K. Con-sidering the high brass weight loss rates in tests I and K, this resultis not so surprising. For test H, this was unexpected, but might beexplained by increased settling of metals into the T water due tothe low flow velocity in the test. The high concentrations of copper

Table 4Maximum depth of corrosion and metal enrichment/depletion (vs. un-exposed surfaces) on brass surfaces at specific corrosion sites as computed from ESEM–EDX data.

Test Sample Corrosion type Maximum deptha (lm) Zinc depletionb Copper enrichmentb Lead enrichmentb

% 95% CI % 95% CI % 95% CI

G T BP 150 NDT PD 240

H In. BP 220 79.3 0.7 28.7 2.6 318.1 7.4In. LD (shallow) ND 59.6 14.6 18.3 7.0 98.1 33.3

I T LD (deep) 260 NDT PD 170 67.4 14.5 29.2 6.5 142.1 56.2

J T LD (deep) 160 NDT PD 340 85.0 3.2 32.2 6.1 220.1 119.1

K T BP 100 NDT LD (deep) 160 45.1 21.1 18.0 11.4 94.5 0.0

ND Not determined, T Brass located out-of-flow, In. Brass located in-flow, BP Brass pitting, LD Layer dezincification, PD Plug dezincification.a Observed on surfaces exposed upon cross-sectioning; site of maximum depth not necessarily site of ESEM–EDX analysis.b Computed from ESEM–EDX data; negative enrichment values indicate depletion of element from brass.

90 µm

Fig. 12. Photograph (lower left) and ESEM image (right) of deep layer dezincification on T brass from test K. Direction of sample cross-section is shown (upper left) and redbox indicates location of imaging; white arrows indicate dezincification layer and depth at imaged point is specified.(For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)

190 µm

280 µm

560 µm

Fig. 13. Photograph (lower left) and ESEM image (right) of plug dezincification on T brass from test J. Direction of sample cross-section is shown (upper left) and red boxindicates location of imaging; white arrows indicate plug and dimensions are specified. Only a portion of the plug in the photograph is shown in the SEM image.(Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

1822 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

accumulated in the T water in test K were largely due to corrosionof the copper tubing located in-flow in this test, similar to tests Dand E in exp. 2. A very high weight fraction of copper (i.e., morethan 12%) was also found in the corrosion scale on the T brass, rem-iniscent of the result from test D too.

Extremely high concentrations of lead were observed in the Twater from the hot water test (I), which was somewhat surprising,

but nevertheless consistent with other data. Lead was significantlydepleted from the exposed brass surfaces from this test; as indi-cated by XRF data, the lead weight fraction on the T and in-flowsurfaces decreased by approximately 44% and 33%, respectively.These values are even higher than corresponding zinc depletionvalues (19% and 29%, respectively) for this test. Brass samples intests G and H were also depleted in lead following testing, but to

E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824 1823

a much lesser extent. Additionally, there was significant lead up-take to the corrosion scales. The T and in-flow scales contained4.4% and 3.1% lead by weight, respectively, which were the highestlead contents measured for any test (Table 2). Moreover, these re-sults do not appear to be merely an artefact of increased metal solu-bility with temperature. As evidence, the lead-to-zinc weight ratiosof the T water and solid corrosion scales from the hot water test (I)were about 3–4 times higher than those from the control (G).

While ESEM–EDX analyses showed that brass corrosion washighly selective for zinc at specific plug or layer dezincificationsites in test I (as well as in other tests), significant lead leachingwas not indicated at these sites (Table 4). Thus, while the most se-vere zinc leaching seems to have been localised, lead leaching ap-pears to have been more generalised. Additionally, like inexperiment 2, zinc and lead depletion on the exposed brass sur-faces in experiment 3 were generally inversely related (Fig. 7). Thisis consistent with Zhang’s [13] recent observations from both hisown original tests and his review of metal leaching data from pre-vious works (by other authors, e.g., [4]) on brass corrosion. The in-verse relationship suggests that the tendency for lead release frombrass may not be mechanistically coupled with severe dezincifica-tion – although both can certainly occur on the same brasscomponent.

4. Discussion

4.1. Effects of flow velocity and differential flow conditions

The enhanced delivery of oxidant (i.e., free chlorine) to a cath-ode surface under flow did influence corrosion on a remote (i.e., re-moved from flow) brass anode. Kelly et al. [22] first demonstratedthis effect (in an analogous but smaller-scale version of the controltest apparatus in exp. 3) by varying flow velocity of oxygenatedwater past one brass sample and noting the effects on the corro-sion current of a second brass sample, which was in electrical con-tact with the first but removed from flow. In that work, increasingflow velocity (i.e., increasing oxidant delivery rate) tended to in-crease corrosion current. In the present work, oxidant delivery ratewas varied by varying bulk oxidant concentration (i.e., test A, exp.1), as well as flow velocity (i.e., tests G and H, exp. 3); and, in bothcases, it was confirmed that increasing oxidant delivery rate to thecathodic metal surface can result in increased corrosion of anodicbrass located out-of-flow. In experiment 3, not only did averagecorrosion current on the remote T brass increase with flow veloc-ity, but so did the rate of weight loss. The T weight loss in test G(high velocity) was nearly 40% higher than in H (low velocity).

Influences of flow velocity on mode of corrosion were subtle.Apart from the fact that plug dezincification was observed in testG but not in H, brass samples from the two tests appeared verysimilar. Additionally, computed zinc depletion and copper enrich-ment values were nearly identical between tests G and H. This sug-gests that only the corrosion rate changed with velocity, not thecorrosion mechanism(s). By comparing results (e.g., metal deple-tion/enrichment) of the T and in-flow brass samples from test J(which were not electrically connected), it appears that the corro-sion mechanisms were even similar between stagnant and flowingconditions. Again, the predominant influence was on corrosionrate, which was higher for the in-flow brass than for the T.

With respect to differential flow, results from tests B (differen-tial flow) and C (uniform flow) in experiment 1 indicated that,when connected to a noble cathode (e.g., copper) being suppliedwith oxidant, brass corrosion may be accelerated if the brass is lo-cated out-of-flow (vs. in-flow). When brass is not connected to amore noble metal, the effects of differential flow conditions maynot be as significant. Results from tests G (differential flow) and J

(T and in-flow brass not connected) in experiment 3 suggested thatelectrical connection between in-flow (cathode) and T (anode)brass did not influence the overall corrosion rate of either sample(i.e., weight losses were similar in both tests). However, the modeof corrosion did appear to be affected. Specifically, the selectivity ofcorrosion for zinc was greater in test G than in J. For example, zincdepletion on the in-flow brass surface was about 15% greater intest G than on its counterpart in J, and the weight ratio of zinc tocopper in the corrosion scales were also much higher in test G thanin J.

4.2. Copper vs. plastic plumbing systems

In comparing plastic to copper plumbing tube materials, exper-imental observations generally confirmed initial hypotheses. Plas-tic (i.e., relatively inert) tubing tended to keep free chlorine levelshigh, and therefore allowed brass to be exposed to higher free chlo-rine than was the case in systems with copper tubes. Although thisdid not result in higher corrosion rates here, in more realistic sys-tems where chlorine residuals in bulk water are not maintainedconstantly and chloride does not build-up, relatively high corro-sion rates and meringue build-up (as was demonstrated in test F)would be expected. This might partly explain why some PEXplumbing systems have been associated with high rates of brassfailures (e.g., [9,32]), which are prompting responses from the per-spective of both alloy certification (i.e., [33] and regulation (e.g.,[34]). While copper tubing, on the other hand, quickly consumedfree chlorine, it also released copper ions to the water – signifi-cantly limiting copper dissolution from brass surfaces which werenot galvanically connected to copper tubing. Although backgroundlevels of dissolved copper could be somewhat lower in real sys-tems than in the pipe-loop tests presented here (i.e., dependingon water chemistry and exposed copper surface areas), the pres-ence of copper tubing in a plumbing system may promote moreselective zinc leaching from brass.

Galvanic connections between copper tubing and brass werefound to be highly influential on both the rate and type of corrosionthat occurred on brass in this work. Weight loss rates of the brassT’s (connected to copper tubing in-flow) in tests D and K were 20%and 31% higher, respectively, than their counterparts in tests E andG (no galvanic connection). Moreover, the galvanic connection intests D and K resulted in brass corrosion being less selective forzinc. This was somewhat unexpected, as it was originally thoughtthat the galvanic current may worsen dezincification; but it islogical given the overall highly anodic behaviour of brass whenconnected to copper. The unique, very uniform layer of dezincifica-tion in test K – which was relatively less selective for zinc thanother layer or plug dezincification or pitting observed in other tests– was likely the result of the consistently high corrosion currentsmeasured on brass in this test (e.g., Fig. 8). In contrast, the brasspitting that occurred in this test was likely induced by water chem-istry, given that pitting was also found in tests G and H, which didnot have galvanic connections to copper. Pit depths were relativelyshallow in test K as compared to G and H (Table 4), which isconsistent with the observations that brass corrosion was moreuniform when the brass was connected to copper.

4.3. Hot vs. cold water systems

The most significant effects of hot water on brass corrosionwere observed in the meringue formation and lead release in testI (exp. 3). The increased meringue build-up seen in this test is con-sistent with prior observations that meringue dezincification prob-lems are often worse in hot water plumbing lines than in cold (e.g.,[23,29]). Additionally, weight losses of the brass samples in test I

1824 E. Sarver, M. Edwards / Corrosion Science 53 (2011) 1813–1824

were higher than their counterparts in test G (cold water), which isalso consistent with findings of other researchers (e.g., [2]).

Lead release was considerably higher at higher temperature,and this could not simply be attributed to increased total metalleaching or increased solubility. Instead, lead appeared to be moreselectively leached by hot water. This result was not anticipated,especially considering that the average chlorine residual in test Iwas relatively low as compared to that of other tests. The literaturesurrounding effects of water temperature on brass lead leaching isscarce, and while some field reports indicate that hot water mayincrease lead concentrations (e.g., [30]), other laboratory testsshow that increased temperature may have no effect or may evendecrease lead release [35]. Given the limitations of the currentwork (e.g., testing of a single brass alloy, lack of replication) andheterogeneities (e.g., pooling) associated with lead distribution inbrass devices, further research on this topic is certainly warranted.

5. Conclusions

Some factors associated with specific plumbing installationscan affect the type and extent of corrosion that occurs on brass de-vices. For high-zinc, un-inhibited brass alloys exposed to aggres-sive water chemistries, the current findings may informinvestigations into existing corrosion issues, as well as design of fu-ture plumbing systems:

(1) Increased flow velocity of chlorinated water tends to accel-erate brass dezincification and/or uniform corrosion ratesvia increased delivery of oxidant to cathodic surfaces; how-ever, at the velocities tested, the type of brass corrosion isnot significantly affected.

(2) Differential flow conditions could accelerate corrosion ofbrass surfaces that are removed from flow but are in electri-cal contact with brass or a more noble metal (e.g., copper)located in-flow.

(3) Galvanic connections between copper and brass tend toincrease overall brass corrosion rates, while making corrosionless selective for zinc. In these tests, the galvanic connectionwas found to increase brass corrosion rate, reduce meringuebuild-up, and render brass corrosion less selective for zinc.

(4) Plastic plumbing tubes do not significantly consume chlo-rine, and therefore allow brass devices to be exposed tohigher chlorine residuals than do copper tubes.

(5) Aside from potential galvanic effects, copper plumbing tubesmay also inhibit copper leaching from brass devices, therebymaking brass corrosion more selective for zinc.

(6) Increased water temperature might significantly increaselead leaching from brass, while reducing selectivity of corro-sion for zinc and promoting meringue build-up.

(7) In addition to other brass corrosion types, pitting may occurin waters with high chlorine residuals.

Acknowledgments

The authors thank the Water Research Foundation (WaterRF)for funding experimental work under project #4289, as well asThe National Science Foundation (NSF), which supported advancedanalyses and writing of the manuscript under grant CBET-0933246.The views, conclusions and/or recommendations expressed hereinare those of the authors, and do not necessarily represent those ofWaterRF or NSF. Additionally, special thanks are extended to PaoloScardina, Jaquelyn Dalrymple and Caitlin Grotke for their assis-tance with experimental work and analyses.

References

[1] E. Sarver, Y. Zhang, M. Edwards, Review of brass dezincification corrosion inpotable water systems, Corros. Rev. 28 (2010) 155–196.

[2] D. Nicholas, Dezincification of Brass in potable waters, Urban Water ResearchAssociation of Australia, Research Report No. 84, (1994).

[3] M. Schock, C. Neff, Trace metal contamination from Brass fittings, J. Am. WaterWorks Assoc. 80 (1988) 47–56.

[4] D. Lytle, M. Schock, Stagnation Time, Composition, pH and OrthophosphateEffects on Metal Leaching from Brass, US EPA, EPA/600/R-96/103, (1996).

[5] D. Kimbrough, Brass corrosion and the LCR monitoring program, J. Am. WaterWorks Assoc. 93 (2001) 81–91.

[6] S. Triantafyllidou, M. Edwards, Critical evaluation of the NSF 61 Section 9 testwater for lead, J. Am. Water Works Assoc. 99 (2007) 133–143.

[7] R. Sundberg, S. Hertzman, M. Linder, Intergranular Corrosion (IGA) of Brass, in:Copper 06, Compiègne, France, 2006. p. 153.

[8] A. Cohen, Corrosion by potable waters in building systems, Mater. Perform. 32(1993) 56–61.

[9] M. Massey, Covington residents sue over corrosive plumbing system, in:Louisiana Record (Online), 2010, available: <http://www.louisianarecord.com/news/229642-covington-residents-sue-over-corrosive-plumbing-system> (citedSeptember 20, 2010).

[10] Safe drinking water act amendments of 1986, Public Law No. 99-339, Dec.(1986).

[11] R. Maas, S. Patch, A. Parker, An assessment of lead exposure potential fromresidential cutoff valves, J. Environ. Health 65 (2002) 9–14.

[12] R. Maas, S. Patch, D. Morgan, T. Pandolfoa, Reducing lead exposure fromdrinking water: recent history and current status, Public Health Reports 120(2005) 316–321.

[13] Y. Zhang, ‘‘Dezincification and brass lead leaching in premise plumbingsystems: effects of alloy, physical conditions and water chemistry,’’ Master’sThesis. Virginia Polytechnic Institute and State University (2009).

[14] V. Lucey, The Mechanism of Dezincification and the Effect of Arsenic, Br.Corros. J. 1 (1965) 9–14. 53–55.

[15] D. Davies, A Note on the Dezincification of brass and the inhibiting effect ofelemental additions, Copper Development Association, Technical Report,(2003).

[16] M. Turner, The influence of water composition on the dezincification of duplexBrass fittings, in: Proceedings of the Society for Water Treatment andExamination, (1961) 162–174.

[17] K. Skidmore, H. Schwartzbart, Corrosion and Dezincification of Brasses inWater, J. Test. Eval. 4 (1976) 426–433.

[18] F. LaQue, H. Copson, Corrosion Resistance of Metals and Alloys, second ed.,Reinhold Publishing Co., New York, 1963.

[19] H. Kaiser, G. Eckstein, Corrosion of alloys, in: A. Bard, M. Stratmann, G. Frankel(Eds.), Encyclopedia of Electrochemistry, Wiley-VCH, Weinheim, 2003.

[20] G. Bengough, R. May, Seventh report to the corrosion research committee, J. ofthe Institute of Metals, 32 (924) 96–296.

[21] L. Tabor, Effects of softened water on equipment, J. Am. Water Works Assoc. 48(1956) 239–246.

[22] G. Kelly, G. Lesbanft, J. Venning, The effects of flow on dezincification, Proc. OfAustralasian Corros. Assoc. Conf. (1980). H-3- 1-14.

[23] V. Lucey, Relationship between water composition and dezincification ofduplex brass, in: Proceedings of BNF’s Seminar on Dezincification in BrassFittings for Water Services, Copenhagen, (1973).

[24] E. Sarver, K. Dodson, P. Scardina, R. Lattyak-Slabaugh, M. Edwards, copperpitting in chlorinated, high pH potable waters, J. Am. Water Works Assoc.,(2011, in press)..

[25] National Association of Homebuilders, Annual Builder Practices Survey, UpperMarlboro, MD, 2005. 2005.

[26] National Primary Drinking Water Regulations, 40 CFR, Part 141.54. 63 FR69465, Dec. (1998).

[27] P. Hong, Y. Macauley, Corrosion and leaching of copper tubing exposed tochlorinated drinking water, Water Air Soil Poll. 108 (1998) 457–471.

[28] R. Oliphant, M. Schock. Copper Alloys and Solder, in: Internal Corrosion ofWater Distribution Systems, American Water Works Association, Denver, CO.,1996, 269-312.

[29] M. Simmonds, W. Huxley, Dezincification of water supply fittings, AustralasianCorros. Eng. 11 (1967) 9–16.

[30] P. McCafferty, R. Schulz, R. D’Ercole, Lead contamination in Perth drinkingwater, Chemistry in Australia (1995) 21–22.

[31] AWWA, APHA, and WEF, Standard Methods, 20th ed., American Public HealthAssociation, (1998).

[32] B. Eckhouse, M. Trask, As Justice’s Wheels Grind, Frustration Over Failure-prone Plumbing Grows (Online), The Las Vegas Sun, 2010, available: hhttp://www.lasvegassun.com/news/2008/mar/23/31000-homes-county-may-need-7000-worth-new-plumbin/i (cited March 23, 2008).

[33] NSF/ANSI Standard 14, Plastic Piping System Components and RelatedMaterials, National Sanitation Foundation (NSF), Ann Arbor, MI, 2009.

[34] California State Legislature, California Plumbing Code, Title 24, Part 5.Sacramento, California, (2007).

[35] J. Paige, B. Covino Jr., Leachability of Lead from Selected Copper-Base AlloysUsed by the Plumbing Industry, in: NACE Corrosion/92, Nashville, TN, 1992.