EPA/600/A-96/016
EVALUATION OF CORROSION ISSUES FOR NANOFILTRATION
TREATED WATER
Darren A. Lytle, Michael R. Schock, Thomas F. Speth
U.S. Environmental Protection Agency
NRMRL, WSWRD, TTEB
Cincinnati, Ohio 45268
and
Jeffrey Swertfeger
Richard Miller Treatment Plant
City of Cincinnati Dept. Of Water Works
Cincinnati, Ohio 45228
Introduction
Membrane technologies (ie. reverse osmosis, nanofiltration, ultrafiltration,
and microfiltration) have been gaining attention as means of meeting existing and
upcoming drinking water regulations. For example, microfiltration has been
demonstrated to be effective for meeting the microbial and turbidity standards of
the Surface Water Treatment Rule (SWTR).1,2 The U.S Environmental Protection
Agency (USEPA) has listed reverse osmosis (RO) as a best available technology
(BAT) for meeting a number of inorganic drinking water standards including
cadmium, nitrite, and nitrate3, antimony and beryllium4, bariurrf, and proposed
standards such as sulfate6, and proposed changes to the radium-226 and -228
standard.7 Future new disinfection by-product (DBP) regulations and growing
microbiological concerns such as Cryptosporidium have further contributed to an
increased enthusiasm towards membrane technology and has brought about an
increase in research efforts.
Ultrafiltration and microfiltration processes are proven solid-liquid
separation technologies suitable for removing particulates from water. However,
they are limited in their ability to remove dissolved organic and inorganic
parameters. Reverse osmosis is very effective for rejecting high percentages of
dissolved and suspended constituents, but may require a great deal of pretreatment
and be associated with high operating and energy costs due to pressure
requirements. Nanofiltration (NF) shares many of the same treatment advantages
as RO, but operates at far lower pressures resulting in a lower cost. For this
reason, nanofiltration technology has attracted attention from some water utilities
as a potential means of meeting the increasing complexity and number of drinking
water standards. Of particular interest is the exploration of nanofiltration as a

-------
treatment strategy for the removal of dissolved DBP precursor material and
microbial contaminants.
Because nanofiltration effectively rejects a majority of water constituents,
the product water is likely to be corrosive towards distribution system materials.
Product water often tends to have a low pH and dissolved inorganic carbon (DIC)
concentration, and is stripped of ions such as silica, phosphate, calcium, and
aluminum which are frequently incorporated in corrosion-inhibiting surface films
that form in distribution systems. The corrosiveness of water treated by
nanofiltration depends direcdy on its application. For example, the use of NF for
the removal of dissolved DBP precursor material may dictate that only a portion
of the source water be treated to meet a desirable DBP level. Therefore, a
proportion of the conventionally-treated water can bypass the membrane and be
blended back with membrane-treated water to achieve a reconstituted stable final
water. However, if the goal of treatment is to achieve 4 to 5 log removal of
microorganisms such as Cryptosporidium, complete water passage through the
membrane will be mandatory.
The goals of this paper are to examine the secondary inorganic water quality
impacts of nanofiltration water treatment relating to corrosion control issues, and
to identify many often-overlooked operational constraints that will cause
substantially increased costs in practical applications. The paper uses data
gathered from two pilot-scale nanofiltration membrane (1.75"xl2" and 4"x40")
systems to address the issues. Computer modeling is used to calculate theoretical
final water qualities under various blending scenarios and to predict corresponding
copper and lead solubilities. Pipe-loop experimentation is used to verify the
anticipated impact of nanofiltered water on copper and lead solubility. A general
discussion of the implications and roles the water quality parameters impacted by
nanofiltration have on the distribution system is made. A brief discussion of
potential corrosion control strategies are made as well as pros and cons and costs
issues. Post-treatment water quality characteristics of nanofiltered water are
compared with full-scale GAC treated water and differences between corrosion
control requirements are described.
Experimental
Nanofiltration System
Two pilot-scale nanofiltration systems were evaluated in the study. The
first NF system consisted of three 4"x40" (10.2 cm xl02 cm) Fluid Systems TFCS
membranes*. The second system consisted of three 1.75"xl2" (2.5 cm x30.5 cm)
f Fluid Systems, San Diego, CA.

-------
Fluid Systems TFCS membranes (rolled by Purification Products Company1) in
series. Because the 1,75"xl2" elements were rolled by a different company than
the 4"x40" membranes, the spacer material was slightly different. The membrane
materials were manufactured on different dates..
In these series systems, the reject water from first membrane was fed to the
second membrane, and the reject of the second was fed to the third. The reject of
the third membrane was sent to waste. Permeate water from the three 4"x40"
membranes were combined in a covered 450 gallon (1700 L) stainless steel storage
tank. The combined permeate water was then fed to the "treated" pipe loop board
for the corrosion control portion of the study. Combined permeate water from the
three 1.75"xl2" membranes was not used to evaluate corrosion control and was
sent to waste. Instead, the performance of the 1.75"xl2" membrane system was
compared to the 4"x40" to evaluate scale-up issues.8 The 4"x40" and 1.75"xl2"
membranes produced an average of 2.0 gpm (7.56 L/min) and 0.13 gpm (0.48
L/min) of water, respectively. A general schematic of the system configuration
is shown in Figure 1. Sample ports, pressure gauges, and rotometers were located
between the membranes to measure the feed and concentrate flows. Sample ports
and rotometers were also located on each permeate line. Stainless steel piping was
used to make water line connections throughout the system.
Pipe Loop System
Two pipe loop boards, a "control" and a "treated" board, were used to
evaluate the impact of nanofiltered water on copper and lead; tin solder corrosion.
Each board consisted of duplicate Type-L soft copper tubing loops and duplicate
Type-L hard copper pipe loops with soldered joints. Each loop was constructed
of new 50 ft. (15.2 m) long sections of 0.5-in (1,27 cm) inside diameter pipe or
tubing. The soldered copper loops were constructed by a skilled technician using
18-90°elbows and 36 soldered joints with 50:50 Sn:Pb solder. The copper tubing
was cleaned prior to use following a previously described procedure which
employed dilute acid and degreasing cleaning steps.9 The cleaning procedure was
not used to clean the soldered copper loops for fear of excessive stripping of the
lead, which would make that comparison less relevant to actual domestic
conditions.
Piping and fittings such as flowmeters and valves leading from the water
sources to the pipe loops were constructed of non-metallic materials such as poly-
vinyl chloride (PVC). Each pipe loop was fed water off a 1.5 in (3.8 cm) PVC
manifold line. Each loop consisted of a check valve, needle valve (to control
flowrate), rotometer (to measure flowrate), metal pipe section, sample port, and
a ball valve in that order. Plumbing fittings and metal pipe were connected by 0.5
Purification Products Company, San Marcos, CA.

-------
in (1.27 cm) schedule 80 PVC pipe. Flowmeters were routinely monitored during
flow cycles and adjusted when necessary to maintain a rate of 1 gpm (3.78
L/min). Pressure gauges and pressure relief valves were place at the beginning of
each manifold to monitor and maintain a pressure of approximately 30 psi (207
kPa) in each board during stagnation periods. A flow totalizer was also placed at
the beginning of each manifold to monitor total flow to each board.
Pipe loop wastes during flow periods went to a common drain on each
board. The pipe loop boards had electronic-timer-controlled solenoid valves
located at the end of each drain line. Because the treated water board was not
sufficiently pressurized under the head of the water in the storage tank alone, a
pump was installed between the base of the storage tank and inlet to the manifold.
A pressure sensor capable of sending an electronic impulse upon a pressure drop
below a preselected pressure was placed strategically in the manifold. When the
pressure in the system dropped below the set point (30 psi), such as during solenoid
opening or during sampling, a impulse was sent to the pump which initiated
activation and flow until pressure returned to the set point. No pump was required
in the control line because the pressure in the source water line was normally well
above 30 psi and required a pressure relief valve. The solenoids were opened 7
times a day (9:30 AM, 12:00 PM, 2:00 PM, 4:00 PM, 6:00 PM, 8:00 PM, and
10:00 PM) for 15 to 20 minutes each time. This permitted a number of flowing
cycles and standing periods to simulate household water usage. Over weekends the
nanofiltration system was shutdown due to safety issues with regard to membrane
operation. As a result, only one flow period was permitted on Saturdays and
Sundays (9:30 AM).
Test Water
The feed water used in the study was Ohio River water treated at City of
Cincinnati's Richard Miller Drinking Water Treatment Plant. Raw Ohio River
water was coagulated with alum and polymers and passed through plate settlers.
Following the plate settlers, water flowed into two sedimentation basins with a
combined maximum detention time of three days. After the storage basins, ferric
sulfate and a small amount of lime were added prior to sand filtration. The
average measurable source water quality characteristics during the time-frame of
the study are listed in Table 1 as "feed water before acid addition" and the
corresponding sampling point is labeled in Figure 1 as sampling position "FBA".
This water was considered the "control" water because it fed the control corrosion
test board.
In order to prevent CaC03 scaling of the NF membranes, 0.75 N I^SQ
was added to the control water prior to the membranes. The resulting water quality
is shown in Table 1 as "feed water H2S04 addition" and represented in Figure 1
as sampling position "FAA". Because this pretreatment was required for operation
of the membrane, it was considered as a contributor to the final water quality

-------
produced by the NF system.
Stored permeate water was fed the "treated" corrosion control test loops.
The permeate water quality characteristics are shown in Table 1 as "4"x40"
combined permeate" and is represented in Figure 1 as "FS-P123". No post-
treatment was performed on the permeate water. Water characteristics of the
combined permeate of the 1.75"xl2" membrane and the reject waters of both
membranes are shown in Table 1.
Sampling
Samples from the nanofiltration systems were taken once a week (typically
on Wednesdays). Samples for metals (Al, Ba, Ca, Cu, Fe, Mn, Mg, Pb, Si, S,
Zn), wet chemistry (CI, Fl, P04, NH3, N03 and alkalinity), and TIC analysis were
taken from the feed line before acid addition, feed after acid addition, and the
combined permeate and the final concentrate lines from both units (locations shown
in Figure 1). Pipe loop samples (500 mL) were taken once a week (Wednesday)
following an 11 hour stand time and measured for metals only.
Analytical
Water samples taken for metal analysis were preserved with ultrapure nitric
acid (0.15% v/v).10 Pipe loop samples and background samples for metals analysis
were analyzed for lead, copper, zinc, iron, calcium, potassium, and magnesium
according to recommended techniques." Background wet chemistry samples were
not preserved and analyzed for alkalinity, phosphate, silicate, ammonia, sulfate,
nitrate, and chloride.1215
Results
Water Quality and Fouling
The inorganic water quality parameter removal performance was measured
by the percent feed rejection of each parameter of interest. The discussion in this
paper is primarily limited to inorganics with emphasis on those that impact
corrosion control. A detailed discussion of organic and particulate parameters
removal characteristics is made elsewhere.8 The rejection of an analyte was
defined as:
£0.100%	(1)
F
where CF (mg/L) was the feed concentration after acid addition and CP (mg/L) was
the combined (3 membranes) permeate concentration. Additionally, the recovery
of an analyte was also determined by:

-------
Cp*0p+CR*0R
p p—* 100%	(2)
cf*of	w
where, CR is the reject concentration (mg/L), Q: is the feed volumetric flowrate
(L/min), Qp is the permeate volumetric flowrate (L/min), and Q is the reject
volumetric flowrate (L/min). The analyte recovery is simply a mass balance and
is useful in determining the tendency of an inorganic parameter to precipitate onto
the membrane which may (or may not) result in fouling (ie. a flux loss).
The flow recovery (ratio of permeate flow to feed flow) that a system
operates at can dictate the final water quality. Figure 2 shows the feed rejection
plotted versus bulk rejection and recovery for a membrane system. The bulk
rejection is determined by averaging the feed and concentrate concentrations. This
better defines what concentration the membrane experiences. The figure shows
how increasing the recovery of a system will result in a final permeate water that
is higher in contaminant concentrations. For a single element operating at low
percent recovery (10 to 15%), the feed rejection and the bulk rejection are very
similar. An example is shown where "Low" is printed. At this point, the recovery
is near 15%, and both rejections are near 80%. If many membranes in series are
used, or if the membranes are staged, the recovery will increase. Full-scale plants
often operate at high recoveries near 85%. If a recovery near 85% is used, the
system will have a total feed rejection near 50%. This is shown where "High" is
printed on the figure. Therefore, the quality of the final water leaving the plant
can change depending on the recovery of the system.
The above example assumes that the bulk percent rejection will remain the
same regardless of the concentration that is in contact with the membrane. This
is commonly observed when the transport of constituents across the membrane
varies directly with the concentration above the membrane. This is observed in
diffusion and convection mechanisms associated with nanofiltration elements.
Sieving mechanisms do not show this property. As a result of these rejection
properties, a permeate water from a membrane that is in a latter stage can actually
be higher in concentration than the original membrane feed water. This is because
the feed water of the final stages is very high in concentration because of the loss
of permeate water in earlier stages. Hence, the total flux of the constituent through
the final membrane will be increased due to the increased concentration gradient
across the membrane.
Table 2 shows that rejection of all measurable water constituents were
generally high. Multivalent ions tended to see higher rejection percentages as
anticipated. Monovalent ions such as chloride and sodium tended to be rejected at
less than 80% while multivalent ions such as aluminum and sulphate were rejected

-------
at rates greater than 90%. Total inorganic carbon (TIC) which is composed of
dissolved carbon dioxide, and the bicarbonate and carbonate ions and carbonate-
containing particulate material, was rejected at a rate of approximately 85%.
Carbon dioxide readily passes the membrane and as a result lowers the pH of the
permeate water. TIC is an important corrosion control parameter and is related to
alkalinity and buffering ability of a water. TOC and UV@254 nm were rejected
at 92% and 98%, respectively.
Recording an accurate measurement of the permeate water's pH was
difficult due its low ionic strength. In addition, the water appeared to be over
saturated with carbon dioxide, and as a result the sample was subject to
degasification upon exposure to the atmosphere. Despite the measurement
complications, the pH drastically decreased following nanofiltration, dropping from
7.8 to approximately 6.1.
The l"xl2" membrane provided consistently higher rejection rates than the
4"x40" membrane. This observation demonstrates the inherent variability in
membrane performance and has been observed elsewhere.8
Recoveries were good for most analytes, averaging within 10% of 100%.
The major exception was iron which only averaged 41% recovery. The low
recovery may have been slightly overstated due to analytical uncertainty of the low
iron measurements in the permeate water near the instrumental detection limit.
However, the error would be relatively small and the outcome still suggests that
iron was a likely foulant and was either adsorbing to the membrane surface or
precipitating onto the membrane.
Analysis of used membrane cleaning solution supported the theory that iron
was a significant membrane foulant. Table 3 shows the inorganic analyses of the
cleaning solution taken on 9/5/95. The cleaning solution was high in aluminum,
calcium, iron, magnesium, sodium, and silicon dioxide. The mass balance
completed over the previous week of operation is also shown. As with daily mass
balances shown in Table 2, the mass balances over this week close (add to 100%)
reasonably well for most analytes except iron and aluminum. To close the mass
balances, the percentage of the feed for the cleaning solution was calculated (Table
3). The cleaning solution data supports the previous assertion that iron and
aluminum were depositing on the membrane. The data also indicate that cleaning
the membrane releases more iron than had deposited on the membrane during the
previous week. This is true to a lesser extent for aluminum. Therefore, this
cleaning was removing iron and aluminum that had deposited early in the
membrane run before routine consistent cleaning.
Analysis of pressure drop and flux decline data shows that the first
membrane element had the greatest amount of fouling.8 This suggests that the

-------
foulant was particulate in nature. Dissolved foulants would cause the greatest flux
decline in the third element in series due to the higher concentrations present in the
element. Therefore, this information combined with the mass balance data,
suggests that the foulant was iron containing-particulate material.
In this study, membrane fouling was a considerable operational and cost
issue. The membranes had to be cleaned weekly due a significant drop in flux
from the apparent build-up of iron, aluminum, and microbiological growths. Iron
fouling can jeopardize nanofiltration efficiency unless pretreatement for iron
removal is employed. Removal of iron using oxidants such as chlorine and
potassium permanganate is not acceptable since the membranes are not tolerant of
oxidants. Removal of iron by aeration and filtration is an alternative, however, air
binding of membranes can also be problematic.
Impact on Corrosion
Surfwial Coatings
Surficial coatings refer to films which cover a pipe and prevent the contact of
water with the underlying pipe material. There are natural diffusion barrier films
which include materials such as aluminum hydroxides and silicates, manganese
dioxide, and organic-containing films; calcium carbonate deposits, and silicate-based
coatings, Historically, calcium carbonate deposition has been a commonly used
corrosion control approach. The treatment involves chemically adjusting the pH,
DIC, and/or calcium concentration of a water to favor the deposition of calcium
carbonate. Recently the reliability of this practice has been questioned due to lack
of evidence showing a uniform protective coating throughout an entire distribution
system. However, calcium-containing solids may be significant in protective films,
even if not exactly calcium carbonate.
Aluminum-containing films (often associated with silica) built up in
domestic plumbing over many years of normal plant operation have been reported
by many. These films have been suggested as having been beneficial to lead and
copper leaching from plumbing.16"19 One study documented an increase in lead
mobilization apparently resulting from sloughing-off of Al-based pipe scales.19
More recent evidence for a corrosion-reducing aluminum or aluminosilicate film
has been described in a study conducted by the Denver Water Department, where
significant precipitated coatings were inhibiting lead and copper release from
corrosion control pipe loop study test rigs.18 Aluminosilicate minerals clearly are
important naturally-forming solids, and are geochemically plausible for many
drinking waters with a near-neutral to slightly acidic pH.
Passivating Films
Passivating films refer to coatings that result from pipe material itself reacting
with major water quality parameters. Examples include metal- carbonate, hydroxide,

-------
phosphate, and sulfate based solids. Such films can either form naturally or by
imposed by corrosion control treatment by adjusting water quality conditions (ie pH,
DIC, and orthophosphate) to favor the formation of the most stable solid form of the
pipe material.
The DIC concentration and pH of a water have the most significant impacts
on the formation of passivating films and metal solubility. DIC can have either a
positive or negative impact on corrosion control.16,20,21 DIC serves to control the
buffer intensity in most water systems, and therefore, sufficient DIC is necessary
to provide a stable pH throughout the distribution system for corrosion control of
copper and lead.16,22 23 A DIC guideline of 2 mg CIL has been suggested as a
minimum DIC to provide adequate pH buffering and to form protective lead
carbonate basic lead carbonate. Discussion among corrosion researchers has
recently resulted is suggestions favoring pushing that minimum limit upwards to
as much as 5 mg C/L primarily to assure adequate buffering ability. However, it
is well established that increasing amounts of DIC can result in increased lead and
copper solubility.16,22"26
The impact of other water quality parameters on lead and copper solubility,
however, are not as well understood. For example, Schock et al.23'27 showed that
copper levels were consistently above the solubility of copper predicted by cupric
hydroxide or oxide models when elevated levels of sulfate (> 30 mg S042~/L) were
present at pH 8.0 in experimental test systems. The authors found that the sulfate-
containing mineral Cu4(0H)6S04 H20 (posnjakite) was present on the copper pipes.
Edwards, et al.24'25 also showed that sulfate increased copper corrosion rates in
water. Rehring and Edwards28 attributed higher copper corrosion rates in enhanced
coagulated waters to additional sulfate carryover from the alum coagulant and also
showed lower copper corrosion rates from chloride carryover from ferric chloride
enhanced coagulation. The impact of anions such as nitrate, sulfate, and chloride
on lead solubility have been considered by several researchers.29
Considerations and Implications
Investigation of the interior surface of home plumbing will likely show a
complex mix of both surficial coatings and passivating films. The solid materials
exist in chemical equilibrium with the ions in the water either as a result of natural
evolution or through encouragement by the addition of chemicals. Disturbance
of that equilibrium will certainly disrupt that balance and lead to the gradual
destabilization of those films.
When considering applying new treatment processes such as nanofiltration,
a water utility must consider secondary impacts that may arise such as pre- and
post-treatment needs and costs, chemical costs and handling issues, and regulatory
conflicts. An evaluation of the impact of the treatment on the water chemistry
should be made either through a pilot- or bench-scale testing program and through

-------
examination of the existing deposits responsible for corrosion control in the present
distribution system. An assessment of how those changes might effect other issues
such as corrosion control must also be made.
The data presented in this paper demonstrated that nanofiltration clearly
rejected TOC and presumably DBP precursor material at a high rate (Table 2)
which was the intended objective of the evaluation. However, a high rejection rate
of DIC, aluminum, and calcium as well as a major drop in pH were also observed.
Based upon the proceeding introduction, such changes can have a negative impact
on the films that may exist on the pipe and metal solubility. The following sections
present a detailed evaluation of the corrosion impacts of nanofiltered water.
Pipe Loop Results
Although the pipe loop portion of the study began in November, 1995, a
month-long partial government shutdown and holiday season limited the number
of data points collected and represented at the time of the preparation of this
document. Figure 3 shows the copper levels during the first month of the study.
Preliminary results showed that copper levels in the loops subjected to nanofiltered
water are considerably higher than in the control loops. Lead levels were also
higher in the solder loops subjected to the nanofiltered water. Lead levels,
however, were not reproducible, which is a common pattern for lead levels
collected from new soldered pipe during the early stages of a study. Because the
films that form on the surface of pipe can take months to develop in a pipe loop,
accurate assessment of the results cannot be made until they are established.
Other Issues
Although not a direct regulatory compliance issue, degradation of
performance of unlined iron, cement mortar-lined or asbestos-cement pipes can
constitute a tremendous cost to utilities, in terms of replacement for structural
failure, loss in flow efficiency, and additional maintenance programs (such as
routine cleaning, relining, and flushing). Further, deterioration in the aesthetic
quality of water provokes serious consumer responses and lack of confidence in the
water supplier. Consumers that then choose to replace municipal water by bottled
water incur both the cost of the conventional water bills, and the cost of the
commercial product. This is often a neglected impact in cost/benefit evaluations.
Aside from the impetus of the lead and copper regulations, for tens of years many
utilities have practiced corrosion control through water chemistry adjustment and
sometimes corrosion inhibitor addition, for these non-regulatory reasons.
By almost any corrosivity standards, the low ionic strength, the low pH and
somewhat low levels of many potentially protective ionic species constitute serious
aggressive properties towards most standard distribution system and domestic
plumbing materials. Various forms of attack and special problems with imbalanced
ionic ratios (such as sulfate and chloride to bicarbonate) have been discussed for

-------
low ionic strength waters.30 Severe attack and undermining of cementitious
materials is also an inevitable consequence of major demineralization.31
Essentially, water supplies largely processed by softening nanofiltration membranes
are changed from frequently stable or otherwise non-corrosive states to waters
requiring extensive corrosion control treatment for all realistic metallic or
cementitious materials. This treatment is necessary to not only restore the finished
water to a state of passivation or inhibition, but also to overcome the serious water
quality degradation from destabilization of existing pipe scales that usually
accompanies any major changes in water quality. Many utilities that have changed
water sources, or that have introduced blending of different (even relatively non-
corrosive) water sources, have had serious problems in this regard.
Blending
Overview
The previous discussions have been based on non-blended nanofiltered
water which in the ease of corrosion control would be considered a "worst case"
condition. If the goal of implementing nanofiltration is to provide adequate
removal of microbial contaminants such as Cryptosporidium, blending feed and
permeate water will not likely be an option. The worst case scenario will be the
case and appropriate evaluation and action to reconstitute the water and provide
desirable corrosion control. If the goal of nanofiltration is to achieve a desired
level of DBP precursor material, blending feed and permeate waters will be a
function of the influent levels and rejection rate. Blending will provide some
reconstitution of the water and corrosion protection.
Estimating Corrosivity Effects
Limitations on the ability of short-term experiments to predict ultimate
implications for regulatory compliance, aesthetic degradation, and infrastructure
integrity resulting from changes in water treatment processes is well-recognized.
Therefore, because the response of lead and copper solubility to different water
qualities is now becoming more predictable, some modeling of the water quality
impacts from blending nanofiltered and bypassed water was undertaken. The
modeling was done in three stages. First, a check of the general quality of the data
and possible inorganic scaling substances. Second, the composition and pH of
various ratios of unadjusted feed water to permeate were estimated. Finally, for
each mixing ratio, the ionic strength, DIC, pH and sulfate concentration were input
into specialty computer programs for computing estimated equilibrium solubility
assuming control by any of several solid phases.
Lead was assumed to be controlled by well-aged PbC03 (normal lead
carbonate, mineral cerussite) or Pb3(C03)2(0H)2 (basic lead carbonate, mineral
hydrocerussite) depending upon the pH and DIC concentration.16,17 This is a
compromise estimate, between the likely worst-case scenario represented by new

-------
lead pipe in pipe loop or coupon situations where a metastable Pb3(C03)2(0H)2 of
slightly increased solubility forms32, and the erratic and sometimes low
concentrations produced by brass fixtures and soldered joints of varying ages, lead
composition33, and exposed surface lead.
The tendency of copper to form metastable phases that persist for many
months to many years has recently been investigated for drinking water
conditions.23"25,27 In recognition of this phenomenon, three alternative scenarios for
copper were chosen: "aged" pipe condition, represented by established coatings of
Cu2(0H)2C03 (malachite), CuO (tenorite), or both; "young" pipe case, represented
by equilibrium with Cu(OH)2 (cupric hydroxide); and metastable equilibrium with
the basic copper sulfate phase approximately corresponding to Cu4S04(0H)6 H20
(langite, or similar). Evidence for the latter case had been found in an earlier
study with finished Cincinnati tap water with a pH greater than 8, where a similar
hydrated hydroxy-sulfate mineral (posnjakite) was observed, but credible solubility
data has not been found for that specific solid.23,27
To do the mixing calculations, the equilibrium chemical speciation program
PHREEQE (PH/REdox EQuilbrium Equations) was used.34 Some major water
constituents are relatively non-reactive and conservative upon mixing. However,
pH and alkalinity are not simple linear mass functions directly proportional to
dilution.35"37 Furthermore, additional examination was desired for changes in
possible precipitate phases that might be responsible for membrane fouling or
changes in the saturation state of possible diffusion barrier forming substances.
These concerns necessitated the use of a comprehensive water chemistry model.
The two end-member cases for water blending modeling were the feed
water (before acid addition) and the combined permeate water. The assumption
made was that a utility might lower costs or reduce radical distribution system
water quality changes through bypassing the membrane system to some extent. The
average water chemistries over the study period covered in this report (Table 1)
were first checked for gross analytical errors via ion balance and speciation
calculations using the WATEQX equilibrium chemistry computer program38 with
slight modifications.23 The equilibrium constants for major dissolved mineral
species were standardized between the two computer programs to yield consistent
computations and predictions. Though no important discrepancies or errors were
found, there was an internal inconsistency among DIC, pH and total alkalinity far
beyond analytical error. This apparently was caused by the non-equilibrium nature
of the operation of the system, resulting in analytical or sampling biases. No good
estimate of system redox potential (pE or Eh) is available, which would be
necessary to accurately compute iron speciation. However, conservatively
assuming a pE of 10 because the feed water was initially near dissolved oxygen
saturation, saturation index computations23,34,38 Calculations showed considerable
over saturation for various aluminum hydroxide, ferric iron oxyhydroxide, and

-------
aluminosilicate solids.
Because of the aforementioned inconsistency among pH, DIC and
alkalinity, mixing computations were performed using the PHREEQE program on
an individual representative case, using analyzed DIC and observed pH for the feed
and permeate waters. Both end-member waters were precisely charge balanced by
"adding" different small quantities of a relatively non-reactive constituent (sodium)
to the analyzed values, in order to facilitate the most accurate computation of the
predicted pH for each mixing case by a widely-accepted approach.40'41
Computations were done assuming mixtures every 10%, from 100% feed water
(0% permeate) to 100% permeate (0% feed). The accuracy of this approach is
largely dependent upon the accuracy of the analyzed pH values. Because of a lack
of viable thermodynamic data for complex formation and solubility constants at
different temperatures, all modeling was done assuming a temperature of 25°C.
This will not affect the trends of the projections, however.
To compute the predicted equilibrium solubility for each mixing ratio, the
resulting pH, ionic strength, sulfate, DIC, and chloride concentrations were input
into the specialized FORTRAN-language LEADSOL and CU2SOL computer
programs, which were described and used for numerous previous studies.17,23,27,32
Figure 4 shows the predicted response of pH, calcium, sulfate and DIC for
various ratios of blending feed water and permeate. The results are very
predictable, showing the linear decrease in the properties where total mass is
conserved (the system is assumed to be closed so that there is no exchange of C02
gas), and the nonlinear decrease in pH. Only approximately 50% blending in this
system reduces the pH from approximately 7.8 to 7. Drastic reductions in other
metal concentrations, such as calcium and aluminum, would be likely to destabilize
existing distribution system films of which they were a significant component. The
role of DIC in corrosion control, metal solubilization, and passivation is
complicated, and interrelated with pH. Therefore, reducing DIC is not necessarily
bad in all cases. What is of concern, however, is that as blending is shifted to high
percentages of permeate water, buffer intensity of the water is considerably
reduced. This effect would be even more pronounced after any subsequent pH
adjustment for corrosion control, particularly if the target pH needed were around
8 to 8.5.22,42 Because DIC realistically provides almost all of the buffering ability
in drinking water, even in the presence of phosphate-based inhibitors, excessive
removal will lead to pH instability throughout the distribution system and poor
corrosion control performance.
Figure 5 shows the predicted effect on copper(II) and lead(II) solubility
given different assumptions about the ages of the pipe and passivating films
forming on them under different water qualities. Effects on lead leaching from
relatively new brass and exposed solder would follow similar trends. For this

-------
particular water chemistry, degradation of performance for lead-containing
materials would begin for even small amounts of added permeate to the finished
water, without subsequent water chemistry adjustment.
The effects on copper(II) solubility are even more dramatic, with copper
solubility likely increasing more than 10-fold with 80% permeate included in the
blend. Only for old copper pipe with very well-developed malachite and tenorite
films would the blending be relatively non-aggressive in terms of overall solubility.
However, even at the low copper levels in equilibrium with well-passivated pipe,
the relative solubility increase, and hence the potential for undermining and
destabilization of the existing surface films would be great.
Comparison Between Nanofiltration and GAC
The feed water to the nanofiltration unit was also the feed water to the
Miller Plant's GAC filtration system (treats an average of 100 MGD) which
allowed for a comparison between treatment processes and corrosion control issues.
Table 4 is a comparison between major water quality parameters and removal
efficiencies for the pilot nanofiltration system and the GAC system. Aside from
major organic parameters, particulates, and turbidity, GAC treatment creates no
major water quality changes. Nanofiltration achieves better removal of organic
parameters, and also creates major water quality changes as already discussed. In
addition, water recovery of the nanofiltration membrane system averaged only
36%. This affects final water quality, as discussed earlier, and concentrate
disposal issues.
Figure 6 is a slightly expanded-scale version of Figure 5, with the addition
of the line for predicted TOC removal for easy comparison. Also marked is the
point in TOC concentration that is equivalent to the removal observed in practice
by the GAC system described previously. Note that this occurs in this system at
ia final mix of approximately 80% permeate water. Because copper and lead
solubility would be elevated 2 to 6 fold at this point, substantial membrane post-
treatment would be required before the water could be distributed.
The primary operational change upon GAC installation, for the Miller Plant
whose influent was a moderately hard, low- to moderate- alkalinity surface water,
was a change in pH control chemical and adjustment point. Final adjustment for
corrosion control consists of lime addition prior to sand filtration. This provided
an increase in pH to 8.5 or above, and occasionally some inadvertant softening.
Final total alkalinity was usually in the 45 to 55 mg CaC03/L range, providing
adequate buffer intensity for unlined iron mains and other distribution system
materials, and a stable water for cementitious materials. Sodium hydroxide is
added to the GAC filtered water to supplement pH adjustment while still adding
lime prior to sand filtration. Without sodium hydroxide addition, the final pH goal

-------
of 8.5 cannot be consistently met. The addition of sodium hydroxide, after
filtration and GAC adsorption, is necessary because during certain times of the
year, lime addition caused calcium carbonate deposition in the sand and GAC
filters. Therefore lime feed is lowered to avoid this complication and sodium
hydroxide was added after GAC to supplement pH adjustment.
Generally, modifications of existing treatment processes that are otherwise
acceptable from a corrosion control standpoint to accommodate only GAC would
tend to involve changes in application points of corrosion control chemicals. The
impact of the GAC process on the major ionic background of the water has been
shown to be relatively minor. Thus, the kinds of process changes to be made
would be unlikely to produce a water chemistry change that would result in major
opportunities to increase corrosivity or destabilize existing films.
On the other hand, the fundamental nature of nanofiltration or other
membrane processes is to largely demineralize the water. To reconstitute even a
blend of the treated source water with membrane filter product to a similar water
quality state as has been established for the distribution system materials over many
years of use, would be a complicated and costly task. In addition to pH
adjustment, DIC, calcium and possibly other constituents may have to be re-
introduced in the form of additional feed of bulk chemicals. Subsequent mixing,
settling or filtration may be needed, as well as any other corrosion inhibitor
chemicals. These factors introduce both capital construction costs, as well as
continuting maintenance and operation cost increases.
Discussion
The results of this study demonstrated that nanofiltration was effective at
rejecting large percentages of organic material as well as a number of inorganic
parameters. Nanofiltered water, however, was determined to be corrosive to lead
and copper due to its low ionic strength, low pH, low DIC, and large removal of
other parameters that may be critical in surface coatings. In addition, the corrosive
water may contribute to the degradation of other distribution system materials and
potentially lead to pipe failures and aesthetic consumer complaints.
When the objective of nanofiltration is to achieve a high degree of
microbiological removal, blending of feed and permeate waters will not be
possible. The resulting water will also represent a worst case scenario requiring
significant post-treatment to achieve corrosion control goals. If the goal of
treatment is to achieve TOC removal, for example, the degree of blending may be
dictated by the amouunt of TOC to be removed. Post-treatment corrosion control
will remain an issue and will have to be evaluated on a case by case basis.
In addition to post-treatment costs, a number of other issues and costs must

-------
be considered prior to implementing nanofiltration. Fouling of the membranes was
a problem in this study. Presumably small amounts of iron and microbiological
foulants were believed to create fouling concerns. This introduces the issue of pre-
treatment to reduce membrane fouling (organic, inorganic, microbiological), which
may consist of iron/manganese oxidation or removal, scaling inhibition (anti-
scalents or acid addition), and disinfection. Because membranes are sensitive to
many oxidants, residual oxidant removal from influent water may also be required.
If pH adjustment is required to prevent calcium carbonate scale build-up, the buffer
intensity of water and original pH will dictate the volume of acid continuously
used. It will also dictate the volume of base chemical used to recover alkalinity.
Chemical safety, storage, and feed equipment issues must also be
considered. And examination of the nanofiltration reject water disposal issues and
increased pumpage demands must be made.
Conclusion
This study demonstrated that nanofiltered-softened water absent of blending,
such as the need for microorganism removal, is a demineralized water of low pH
and is corrosive towards distribution system and domestic plumbing products.
Costs associated with post-treatment of the water to meet lead and copper
requirements, as well as chemical addition, handling, and safety issues must be
addressed before considering nanofiltration. The blending and solubility
calculations presented may provide slightly different results depending primarily
upon the pH and DIC for the normal source and finished waters involved.
However, they still demonstrate that corrosion issues remain a concern when
blending is employed. Pre-treatment costs to prevent membrane fouling can also
be a significant cost.
Nanofiltration was more effective at removing organic material than GAC.
However, it produced a far more aggressive finished water. Nanofiltration also
had concentrate disposal issues that GAC lacks. Ironically, nanofiltration may be
more viable for source waters with low hardness, pH and DIC, so that the changes
in parameters most influential to corrosion will be minimized, and scaling
constraints will be reduced.
Acknowledgements
The authors would like to thank Renea Lohmann and Jack DeMarco of the
Cincinnati Water Works for sharing operational data and providing review of the
experimental results and the paper. The authors also appreciate insightful
suggestions and comments provided by Michelle Frey, Black and'Veatch, Inc.
(Aurora, CO). Analytical data, sampling, and technical assistance was provided
by Keith Kelty, James Doerger, James Caldwell, Don Mitchell, Maura Lilly,

-------
Christopher Keil, and Herb Braxton of the Water Supply and Water Resources
Division of the USEPA National Risk Management Research Laboratory, and
Gregory George, Stephen Harmon, Mark Domino and John Dammann of
Dyncorp/Technology Applications Incorporated. Leo Fichter of NRMRL, USEPA
also provided important construction and testing support, and Thomas Sorg and
Jeffrey Adams also of NRMRL, USEPA, provided manuscript reviews.
Disclaimers
Mention of commercial names does not constitute endorsement or
recommendation by the agency. The views expressed in this paper are those of the
authors, and do not necessarily reflect USEPA policy.
References
1.	Filtration and Disinfection: Turbidity, Giardia lamblia, Viruses, Legionella,
and Heterotrophic Bacteria. Final Rule. Fed. Reg. 54:124:27486 (June
29, 1989).
2.	Coffey, B. M., Stewart, M. H. & Wattier, K. L. Evaluation of Microfiltration
for Metropolitan's Small Domestic Water Systems. Proc. AWWA
Membrane Technology Conference, Baltimore, Md., Aug. 1-4, (1993).
3.	SOCs and IOCs. Final Rule. Fed. Reg. 56:20:3526 (Jan. 30, 1992).
4.	SOCs and IOCs. Final Rule. Fed. Reg. 57:138:31776 (July 17, 1992).
5.	Aldicarb, Aldicarb Sulfoxide, and Aldicarb Sulfone, Pentachlorophenol, and
Barium. Final Rule. Fed. Reg. 56:126:30266 (July 1, 1992).
6.	Sulfate. Proposed Rule. Fed. Reg. 59:243:65578 (Dec. 20, 1994).
7.	Radionuclides. Proposed Rule. Fed. Reg. 56:138:33050 (July 18, 1991).
8.	Speth, T. F., et al. Evaluating Bench-Scale Nanofiltration Studies for
- Predicting Full-Scale Performance. To be published in the Proc. AWWA
GAC and Membrane Workshop, Cincinnati, OH, March 4-6, (1996).
9.	Lytle, D. A., Schock, M. R. & Tackett, S. Metal Corrosion Coupon Study
Contamination, Design, and Interpretation Problems. Proc. AWWA Water
Quality Technology Conference, Toronto, ON, Nov. 15-19, (1992).

-------
10.	U.S. Environmental Protection Agency. "Handbook for Analytical Quality
Control in Water and Wastewater Laboratories", Environmental Monitoring
and Support Laboratory, Cincinnati, OH., EPA-600/4-79-019 1979).
11.	U.S. Environmental Protection Agency. "Test Methods for Evaluating Solid
Wastes", Risk Reduction Engineering Laboratory, SW846 (Sept., 1986).
12.	APHA, AWWA & WPCF. Standard Methods for the Examination of Water
and Wastewater. 18th Ed. (1992).
13.	Modified from Methods for the Determination of Inorganic Substances in
Water and Fluvial Sediments, U.S. Geological Survey Open-File Report,
(85-495). (1985).
14.	Alpkem Research Inc. Clackamas, OR. (1995).
15.	U.S. Environmental Protection Agency. "Methods for Chemical Analysis of
Water and Wastes", EPA-60014-79-020 1983).
16.	Schock, M. R. "Understanding Corrosion Control Strategies for Lead".
AWWA, 81:7:88 (1989).
17.	Schock, M. R. & Wagner, I. "The Corrosion and Solubility of Lead in
Drinking Water". Ch. 4 In: Internal Corrosion of Water Distribution
Systems. AWWA Research Foundation/DVGW Forschungsstelle, Denver,
CO, (1985).
18.	Lauer, W. C. & Lohman, S. R. Non-Calcium Carbonate Protective Film
Lowers Lead Values. Proc. AWWA Water Quality Technology Conference,
San Francisco, CA, November 6-10, (1994).
19.	Fuge, R. & Perkins, W. "Aluminum and Heavy Metals in Potable Waters of
the North Ceredigion Area, Mid-Wales". Environ. Geochem. & Health,
13:2:56 (1991).
20.	AWWARF. Lead Control Strategies. AWWA Research Foundation and
AWWA, Denver, CO, (1990).
21.	Gregory, R. & Jackson, P. J. Central Water Treatment to Reduce Lead
Solubility. Proc. AWWA Annual Conference, Dallas, TX, June 10-14,
(1984).

-------
22.	Schock, M. R. "Internal Corrosion and Deposition Control". Ch. 17 In:
Water Quality and Treatment: A Handbook of Community Water Supplies.
McGraw-Hill, Inc., New York, (1990).
23.	Schock, M. R., Lytic, D. A. & Clement, J. A. "Effect of pH, DIC,
Orthophosphate and Sulfate on Drinking Water Cuprosolvency", U. S.
EPA Office of Research and Development, Cincinnati, OH,
EPA/600/R-95/085 1995).
24.	Edwards, M., Meyer, T. E. & Schock, M. R. "Alkalinity, pH and Copper
Corrosion By-Product Release", submitted manuscript 1995).
25.	Edwards, M., Meyer, T. & Rehring, J. P. Effect of Various Anions on
Copper Corrosion Rates. Proc. AWWA Annual Conference, San Antonio,
TX, June 6-10, (1993).
26.	Schock, M. R. & Lytle, D. A. The Importance of Stringent Control of DIC
and pH in Laboratory Corrosion Studies: Theory and Practice. Proc.
AWWA Water Quality Technology Conference, San Francisco, CA,
November 6-10, (1994).
27.	Schock, M. R., Lytle, D. A. & Clement, J. A. Modeling Issues of Copper
Solubility in Drinking Water. Proc. ASCE National Conference on
Environmental Engineering, Boulder, CO, July 11-13, (1994).
28.	Rehring, J. P. & Edwards, M. The Effects of NOM and Coagulation on
Copper Corrosion. Proc. ASCE National Conference on Environmental
Engineering, Boulder, CO, July 11-13, (1994).
29.	Beccaria, A. M., et al. "Corrosion of Lead in Sea Water". Br. Corros. J.,
17:2:87 (1982).
30.	Ferguson, J. F. "Corrosion Arising from Low Alkalinity, Low Hardness, or
High Neutral Salt Content Waters". Ch. C, 8 In: Internal Corrosion of
Water Distribution Systems. American Water Works Association Research
Foundation/DVGW Forschungsstelle, Denver, CO, (1985).
31.	Holtschulte, H. & Schock, M. R. "Asbestos-Cement and
Cement-Mortar-Lined Pipes". Ch. 6 In: Internal Corrosion of Water
Distribution Systems. AWWA Research Foundation/DVGW
Forschungsstelle, Denver, CO, (1985).

-------
32.	Schock, M. R. & Gardels, M. C. "Plumbosolvency Reduction by High pH
and Low Carbonate-Solubility Relationships". Jour. AWWA, 75:2:87
(1983).
33.	Lytle, D. A. & Schock, M. R. Impact of pH and Lead Composition on Metal
Leached from Brass Coupons. Proc. AWWA Annual Conference, San
Antonio, TX, June 6-10, (1993).
34.	Parkhurst, D. L., Thorstenson, D. C. & Plummer, L. N. PHREEQE - A
Computer Program for Geochemical Calculations. Water-Resources
Investigations 80-96, U. S. Geological Survey (1980).
35.	Trussell, R. R. & Thomas, J. F. "A Discussion of the Chemical Character of
Water Mixtures". Jour. AWWA, 63:1:49 (1971).
36.	Wigley, T. M. L. & Plummer, L. N. "Mixing of Carbonate Waters".
Geochim. Cosmochim. Acta, 40:989 (1976).
37.	Jordan, C. "The Mean pH of Mixed Fresh Waters". Water Res., 23:10:1331
(1989).
38.	van Gaans, P. F. M. "WATEQX - A Restructured, Generalized, and Extended
FORTRAN 77 Computer Code and Database Format for the WATEQ
Aqueous Chemical Model for Trace Element Speciation and Mineral
Saturation, for Use on Personal Computers or Mainframes". Comp. &
Geosci., 15:6:843 (1989).
39.	Nordstrom, D. K. & Munoz, J. L. Geochemical Thermodynamics. Blackwell
Scientific Publications, Palo Alto, CA, (1986).
40.	Plummer, L. N. Mixing of Sea Water with Calcium Carbonate Ground Water.
Memoir 142, Geological Society of America (1975).
41.	Wigley, T. M. L. & Plummer, L. N. "Mixing of Carbonate Waters".
Geochim. Cosmochim. Acta, 40:989 (1976).
42.	Snoeyirik, V. L. & Jenkins, D. Water Chemistry. John Wiley and Sons, New
York, (1980).

-------
Concentrate Concentrate
0.34 GPM 0.36 GPM
1" X 12"
Nanofiltration System
Concentrate
to Waste 0.31 GPM
PP-
C3
Acid Feed (6 mL/min)
0.75 NH,SO.
6.0 GPM
PP-P123 TO WASTE
FAA
FBA
Concentrate
4.5 GPM
Concentrate
5.3 GPM
Concentrate
to Waste 3.7 GPM
4" X 40"
Nanofiltration System
FS-
C3
PP-
P123
Unit
Unit
Unit
Unit
Unit
Unit
0.8 GPM 0.7 GPM
0.7 GPM
Combined
Permeate
- Sampling site
Pipe Loops
Storage Reservoir
(450 Gallons)
Figure 1. Schematic of nanofiltration pilot systems.

-------
Percent Feed Rejection for Total System
s
o
• MM
(J
o>
*0?
PC
"9
PQ
+•>
c
v
a
u
o>
Ph
/
Percent Recovery
Figure 2. Feed rejection versus bulk rejection and recovery for a membrane system.

-------
3.0
2.5
2.0
d
en
£
S3 15
a
a.
e
U
1.0
0.5
o.o-
0 10 20 30 40 50 60 70 80 90 100
Time, days
• Copper tubing control
—O— Copper/solder pipe control
-	• - Copper tubing treated
-	O - Copper/solder pipe treated
o-
Figure 3. Copper levels from pipe loop samples taken following 11
hours stagnation time. Control and nanofiltered loops are shown.

-------
10 20 30 40 50 60 70 80 90 100
Percent Permeate
Figure 4. Theoretical effect of blending on some major water
quality parameters.

-------
7.0
Pb (aged)
Cu (langite)
Cu (young)
Cu (aged)
6.0
5.0
0	20	40	60	80 100
Percent Permeate
0
0
0
0
0
0
0
0
0
0
Figure 5. Theoretical effect of blending on lead and copper
solubilities.

-------
Percentage nanoftllered equivalent to full-scale TOC removal by GAC
•O- TOC
Pb (aged)
- -A- * Cu (Iangite)
Cu (young)
• * Cu (aged)
4.0
3.0
2.0
60
100
Percent Permeate
Figure 6. Relationship between theoretical TOC removal
and lead and copper solubility under different mixing scenarios.

-------
Table 1. Water quality charatcteristies at stages of nanufiltration treatment system.

Feed






Feed Water
Feed Water
4"X40'


1"X12"

Before Acid
H2S04Acid
Final
Combined
Final
Combined
Analvte
Addition
Addition
Concentrate
Permeate
Concentrate
Permeate
Location
FBA
FAA
FS-C3
FS-P123
PP-C3
PP-P123
Cations






Aluminum, mg/L
0.1
0.1
0.2
0.0
0.2
0.0
Calcium, mg/L
37.3
38.3
65.3
2.0
63.2
0.7
Iron, mg/L
0.0
0.0
0.0
0.0
0.0
0.0
Potassium, mg/L
2.8
3.0
4.9
0.9
5.0
0.4
Magnesium, mg/L
10.0
10.4
17.9
0.6
17.5
0.2
Sodium, mg/I.
15.8
16.4
28.0
4.0
28.3
2.6
Anions






Chloride, mg/1
16.7
19.0
33.0
4.3
32.1
1.0
Fluoride, mg/L
L0
1.1
na
na
na
11a
Nitrate, mg NO3/L
1.0
1.2
1.6
0.4
1.4
0.3
Phosphate, mg PO4/I.
0.0
0.0
0.0
0.0
0,0
0.0
Silica, mg SiO£
5.0
4.9
7.3
0.8
7.2
0.4
Sulfate, mg SO^/L
69.0
73.0
127.4
2.6
123.7
0.6
Other






Alkalinity, mg CaCOq/L
69.9
70.7
120.8
7.4
117.3
3.8
Total Inorganic Carbon, mg C/L
18.4
18.2
29.0
3.0
na
na
Hardness, mg CaCOjj/L
134.3
138.4
236.7
7.2
229.8
2.3
pH
7.8
7.4
7.7
6.1
7.6
na
Temperature, C
26.6
27.9
29.8
29.8
29.9
29.7
TDS
218.0
223.0
318.0
21.4
311.0
11.6
Ionic Strength
0.0057
na
0.0005
na
na
na
TOC, mg/L
2.2
na
na
0.2
na
na
UV 254, m-1
5.1
na
na
0.1
na
na
na- not analyzed

-------
Table 2. Water quality rejection recovery performance of pilot nanofiltration membranes.

4"X40"
1"X12"
Difference
4"X40"
PX12"
Difference
Analvte
% Rejection
% Refection % Rejection
% Recovery % Recovery % Recovery
Aluminum, mg/I,
93.4
95.5
2.2
113.3
114.8
1.3
Calcium, mg/I,
95.7
98.3
2.6
118.2
98.3
-20.2
Iron, mg/L
91.6
92.5
1.0
41.0
32.1
-27.7
Potassium, mg/L
80.3
87.3
8.0
93.0
106.1
12.4
Magnesium, mg/L
94.8
98.5
3.8
101.6
104.7
3.0
Sodium, mg/L
78.3
85.6
8.5
101.7
104.8
3.0
Chloride, mg/1
83.4
92.8
10.1
109.8
90.6
-21.2
Nitrate, mg NO3/L
36.4
65.3
44.3
90.9
104.8
13.3
Silica, mg Si02
82.7
90.8
8.9
101.4
104.4
2.9
Sulfate, mg SO4/L
96.6
99.3
2.7
102.7
105.0
2.2
Total Inorganic Carbon, mg C/L
84.7
na
na
100.9
na
na
Hardness, mg CaCOjj/L
94.6
98.3
3.8
na
na
na
TOC, mg/I,
91.8
na
na
na
na
na
UV 254, m-1		
97.6
na
na
na
na
na
na- not analyzed

-------
Tabic 3. Membrane cleaning solution analysis.

Concentration in

Feed in

Cleaning Solution,
Weekly Recovery
Cleaning Solution,
Species
mg/L
%
%
Aluminum
18.9
94.6
16.1
Calcium
60.7
101.5
0.1
Iron
13.8
18.3
226.6
Magnesium
8.2
101.1
0.1
Silica
29.1
101.8
0.6
Sodium
11.6
101.9
0.1

-------
Table 4. Comparison between nanofiltration and GAC filtration removal efficiency of selected parameters.
Nanofiltration 			GAC
	 	Analyte			
Influent
FJfluent
% Rejection
.... Influent
F.ffluent,
% Removal
Bromide, ug/L
74.1
12.4
83.3
0.1
0.1
0.9
HPC, CFU/mL
19215.0
525.0
90.9
238.0
104.6
na
SOS I"HM, ug/L
14.2
4.0
96.4
149.5
38.8
71.7
MAX THM, ug/L
282.0
15.3
94.4
na
na
na
Spore - forming Bacteria, CFU/mL
10.0
2.0
80.0
na
na
na
SDSTOX. ug/L
246.0
11.3
95.6
228.5
38.8
83.0
MAX TOX. ug/L
336.0
17.0
94.7
na
na
na
Conductivity, us/cm
418.0
39.2
90.6
426.1
426.9
-2.5
TOC, mg/L
2.2
0.2
91.8
2.0
0.6
68.4
Particles*, units/mL
861.0
5.0
99.4
785.3
273.2
67.8
Turbidity, NTU
0.2
0.1
72.2
0.1
0.1
26.6
UV254,m-1
5.1
0.1
97.6
na
na
na
Calcium, mg/L
37.3
2.0
95.7
44.1
44.0
0.1
Magnesium, mg/I.
10.0
0.6
94,8
9.4
9.8
-11.4
I lardness, mg CaC03
134,6
7.2
94.6
147.1
149.1
-1.7
Alkalinity, mg CaCOg
69.9
7.4
—
77.4
77.3
0.3
Sulfate, mg SO4/I,
69.0
2.6
96.6
81.7
86.9
-6.9
11C, mg C/L
18.4
3.0
84.7
na
na
na
pH, units
7.8
6.1
--
8.0
7.9
—
* Particles greater than 2 uM
na- not analyzed

-------
TECHNICAL REPORT DATA
(Please reed Instructions on the reverse before eompktir
, »t»nST NO. 2,
EPA/600/A-96/016
3. P
4, TITLE AND SUBTITLE
EVALUATION OF CORROSION ISSUES FOR NANOFILTRATION
TREATED WATER
S. REPORT DATE
6. PERFORMING ORGANIZATION CODE
7. autmoris) (l) Darren A- Lytle, Michael R. Schock, and
Thomas F. Speth and (2) Jeffrey Swertfeger
B.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
(1)	OSEPA/NRMRL/WSWRD/TTEB
26 W. Martin Luther King Drive
Cincinnati, OH 45268
(2)	Richard Miller Treatment Plant
Citv of Cinti Dept. of Hater Works rinrinnatif OH
10. PROGRAM ELEMENT NO.
11 CONTRACT/GRANT NO.
In-House
12, SPONSORING AGENCY NAME AND ADDRESS 45228
NATIONAL RISK MANAGEMENT RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
13. TYPE OF REPORT AND P1RIOO COVERED
Conference Paper
14. SPONSORING AGENCY CODE
EPA/600/14
IB. supplementary notes Contact - Darren A. Lytle 513/569-7432
To be presented at the 1996 AWWA/DSEPA Workshop - GAC and Membranes: Bench and Pilot
Scale Evaluations - March 4-6, 1996 - Cincinnati, OH
16. ABSTRACT
Nanofiltration membranes are known to remove a large percentage of naturally
occurring organics. Removals are often higher than 90 percent. This is also
true foi disinfection byproduct (DBP) precursors. Due to the properties, the
Information Collection Rule may require water utilities of a certain size and
water quality to evaluate either membrane treatment or granular activated
carbon treatment. Along with high organic rejection properties,
nanofiltration membranes are also known to remove a high percentage of
hardness, altcaxmity, and a fair amount of monovalent ions. Because of the
rejection of inorganic species, steps must be taken to prevent collection of
inorganic precipitation on the membrane. This often is completed by
depressing the pH of the influent water, or by adding an antiscalent to the
water. The permeate water of nanofiltration membranes is typically a low pH
water that has little buffering capacity. The nanofiltration product water is
therefore very corrosive. Steps must be taken to avoid lead and copper
leaching in plumbing materials found in distribution systems and private
residences. It is the objective of this paper to describe the organic and
inorganic constituents that are rejected by a nanofiltration membrane, to
determine the corrosivity of nanofiltration permeate water by conducting pipe
loop experiments, and to evaluate method for mitigating corrosion problems.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. I DENT I F IE RS/OPE N ENDED TERMS
c. COS ATI Field/Group
Nanofiltration
Corrosion
Copper
Lead


18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report}
UNCLASSIFIED
21. NO. OF PAGES
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (R«v. 4-77) previous edition is obsolete

-------