ENVIRONMENTAL ENGINEERING SCIENCE
Volume 19, Number 6, 2002
© Mary Ann Liebert, Inc.
Micro-Organism Rejection by Membrane Systems
William A. Lovins III, 1 James S. Taylor,2* and S.K. Hong2
1 Boyle
Engineering Corporation
Orlando, FL 32801
2
Civil and Environmental Engineering Department
University of Central Florida
Orlando, FL 32816-2450
ABSTRACT
The removal of micro-organisms by membrane systems was investigated using single-element membranes
and five species of micro-organisms in a plant setting at East St. Louis, MO. Single-element membranes
included a cellulose acetate ultrafilter (UF), a polysulfone microfilter (MF), a cellulose acetate (CA)
nanofilter (NF), and two composite thin-film (CTF) nanofilters. Micro-organism challenge studies were
conducted using raw, alum coagulated-settled, and finished plant water. Model micro-organisms consisted
of Clostridium perfringens (strain 26) spores (,1–5 mm) for bacteria simulation, MS-2 (,0.025 mm), and
PRD-1 (,0.1 mm) phage for virus rejection and Cryptosporidium parvum oocysts (,4–6 mm) and Giardia lamblia cysts (,8–14 mm) for cyst rejection. Sixty-eight observations of micro-organism rejection
were gathered over 1 year of operation in eight separate challenge events where micro-organisms were
spiked separately and as a mixture. The composite thin-film nanofilters provided significantly better disinfection than the cellulose acetate nanofilter. However, a cellulose acetate ultrafilter rejected more
micro-organisms than any membrane tested, indicating disinfection by cellulose acetate membranes is a
function of construction and module configuration rather than membrane film, as both the CA and CTF
membranes were constructed in a spiral wound configuration. Micro-organism log rejection was independent of organism size except for the MF, which passed viruses, and was independent of membrane
material but varied among membranes.
Key words: membranes; nanofiltration; microfiltration; ultrafiltration; disinfection; micro-organism;
Clostridium perfringens; Cryptosporidium parvum oocysts; Giardia lamblia cysts; log rejection
INTRODUCTION
I
supplies are a worldwide public heath concern. Over 100
different pathogenic waterborne organisms including
NADEQUATE AND COMPROMISED DRINKING WATER
protozoa, bacteria, and viruses have been identified in
contaminated drinking water. Cryptosporidiumand Giardia are waterborne protozoa that have been found in some
municipal water systems supplied by surface water or
groundwater under the influence of surface water (Fox
*Corresponding author: Civil and Environmental Engineering Department, University of Central Florida, 4000 Central Florida
Blvd., P.O. Box 162450, Orlando, FL 32816-2450. Phone: 407-823-2785; Fax: 407-823-3315; E-mail: [email protected]
453
454
and Lytle, 1996). These parasites, which are not species
specific, are excreted in a fully infective form by an infected human or animal. Infection by Cryptosporidium
or Giardia is characterized by severe diarrhea and nausea. Immunocompromised individuals are particularly
at risk. No therapies for cryptosporidiosis have yet been
accepted by the Food and Drug Administration. However, some success has been achieved with therapies for
giardiasis. Cryptosporidium is currently recognized as
the most significant cause of waterborne disease in the
United States, where at least 12 waterborne Cryptosporidium outbreaks have been documented. Cryptosporidiosis outbreaks associated with drinking water
also have been documented in the United Kingdom,
Japan, and Holland. The largest U.S. outbreak occurred
in Milwaukee, WI, in 1993, where an estimated 450,000
people became ill and 144 died due to contamination of
the water supply (Hoxie et al. 1997). The occurrence of
Cryptosporidium in surface waters has been reported in
4 to 100% of the samples examined with levels ranging
from 0.1 to 10,000 organisms/100 L, depending on impacts from sewage and animals (Lisle and Rose, 1995).
The average number of samples testing positive for
Cryptosporidium in groundwater have ranged from 9.5
to 22% Hancock et al. (1998).
Groundwater contamination is of particular concern for
viruses because of their submicron size and ability to be
transported through soil structures and contaminate
groundwater. Illnesses most commonly associated with
viral contamination include, but are not limited to, hepatitis, Norwalk, rotavirus, and coxsackie viruses. Gerba
and Rose (1989) reported viruses can survive several
months in groundwater. LeChevallier (1996), reported viral contamination in approximately 20 to 30% of groundwater systems did not correlate with coliform contamination. Lack of correlation between viruses and bacterial
contamination was also shown in a study performed by
Glover (1994), which found no correlation between coliform, heterotrophic plate count, chlorine level, and Mycobacterium avium complex (MAC) measured in drinking water reservoirs. MAC is particularly invasive in
immunocompromised conditions, which are common to
individuals with AIDS, neoplastic disorders, or postheart
transplant (Novick, 1990). Cholera, Helicobacter pylori
and Escherichia coli O157:H7 are additional bacterial
agents responsible for waterborne disease, which have
had historically devastating effects worldwide. Hence,
the removal of known and unknown pathogens is highly
desirable in drinking water treatment. Given limited removal capability by conventional treatment processes; resistance to chlorine disinfection; and the importance of
potable water to public safety, the economy and morale;
more effective barriers to pathogens are needed to pro-
LOVINS ET AL.
tect against both incidental and intentional microbial contaminant threats.
Membranes have been selected for use in part because of their ability to removal many different microorganisms, yet surprisingly little information has been
documented to quantitatively state the microorganism removal capacity of membranes (Urbain and Manhem,
1997). Membrane pilot studies have demonstrated complete or very high micro-organism removal, but few have
used micro-organism challenge studies to demonstrate
the micro-organism removal capacity of membranes (McNamara and Gavin, 1997; Mourato et al., 1997). Visual
inspection and dye testing, vacuum testing, E. coli tracer
studies and particle counting have been compared for assessment of membrane integrity on a field scale. E. coli
tracer studies were superior to other techniques for assessing integrity because of their ability to show less than
one log rejection in damaged elements and greater than
four-log rejection for intact elements. In this same study,
greater than 5- to 6-log removal was demonstrated for
Cryptosporidium parvum oocysts and Giardia Muris
cysts. Greater than 4-log removal was demonstrated for
MS2 phage (Seyde et al., 1999). Hong et al. (1999) demonstrated greater than 6-log rejection for MS2 phage and
aerobic spores, but found only 5-log rejection for C.
parvum oocysts in MF field studies. The difference in organism rejection was attributed to sample collection volumes and organism feed stream concentrations. Similar
results have been found in a laboratory study. Colvin et
al. (1999) found greater than 7-log rejection of MS2
phage in flat sheet studies of RO membrane films and
lack of correlation of microsphere and organism rejection. In a later publication on the field study, Colvin et
al. (2001) found that pressure decay testing, particle, and
turbidity monitoring of damaged and undamaged membranes could detect fiber failure of UF or MF membranes
and demonstrated repair of damaged membranes returned
performance to that at the undamaged state. Kruithof et
al. (2001) in a large scale pilot study in The Netherlands
found that undamaged UF membranes removed greater
than 5.4 log of MS2 phage, and that vacuum testing successfully identified damaged membranes from the manufacturers.
State regulatory issues facing ultrafiltration and microfiltration have been recently documented (Allegier,
2001). As of 2001, 29 states had UF or MF plants.
Twenty-one states awarded 2- to 3-log credit for Giardia. Only six states awarded the 2-log credit for Cryptosporidium that has been recently required by the Interim Enhanced Surface Water Treatment Rule. The
majority of states do not grant any credit for virus inaction to either MF or UF. No such assessment has been
made for nanofiltration or reverse osmosis, which high-
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
lights the need for studies on and regulatory recognition
of membrane rejection of micro-organisms.
This article presents results from a pilot plant investigation that assessed the effectiveness of membranes
for rejection of micro-organisms. Five different microorganisms ranging in size from 0.025 (phage) to 12 mm
(Giardia) were used to challenge five different membranes. Turbidity and particle counts were evaluated as
surrogate parameters for the assessment of micro-organisms. The University of Central Florida in cooperation
with the American Water Works Service Company, Inc.
conducted the project at the Illinois–American East St.
Louis surface water treatment facility. The water treatment plant intake is located on the Mississippi River several miles downstream of the confluence of the Mississippi and Missouri Rivers. This source has varied water
quality throughout the year, but is generally regarded as
a high turbidity, moderate organic source water.
THEORY
A representation of flow through a pressure driven
membrane module is shown in Fig. 1. Mass transfer
through pressure driven membrane modules can be described by either size exclusion or diffusion controlled
mechanism. Models describing solute mass transfer by
size exclusion or diffusion are shown in equations (1)
(Taylor and Wiesner, 1999) and (2) (Mulford et al.,
1999). Other investigators have proposed similar models
(Chellam and Taylor, 2001; Shetty et al., 2002).
(1)
Cp 5 fCf
CF KS
RFW
Cp 5 } ln 1 2 }
2RFW
FW 1 KS
1
2
(2)
where: CF is the feed solute concentration (M/L3), f is
the rejection coefficient; FW is the water flux (L3 /L2t), Ks
is the solute mass transfer coefficient (L/t), CP is the permeate solute concentration (M/L3 ), and R is the fraction
recovery.
455
Size-exclusion and diffuson-controlled mass transfer
are illustrated in Fig. 2. Size-exclusion mass transfer is
only affected by feed stream concentrations, and is independent of flux, and recovery. Diffusion controlled solute
mass transfer is affected by the membrane solute mass
transfer coefficient, flux, and recovery. Assessment of
solute mass transfer mechanisms by varying flux and recovery is well documented and useful for assessing and
modeling potable water treatment applications (Taylor et
al., 1989; Taylor and Jacobs, 1996; Chellam and Taylor,
2001). Solute mass transfer is illustrated in this manner
in Fig. 2. The flat and curved planes show solute passage
for size-exclusion and diffusion-controlled processes, respectively, as measured by log organism rejection, the
method commonly used for calculating microbial removal. Note, a high log rejection corresponds to a low
permeate concentration. Because solute passage for a size
exclusion process only varies with feed concentration,
which in this case was held constant, log rejection for the
size exclusion case does not vary with flux or recovery.
The curved plane in Fig. 2 illustrates diffusion controlled
mass transfer increases with decreasing flux and increasing recovery. Log rejection is highest at the highest
flux and lowest recovery, and lowest for the lowest flux
and highest recovery.
METHODOLOGY
The efficiency of membranes for micro-organism removal was evaluated in the field using continuous injection of stock organisms mixtures for 12 to 15 min into
the feed stream. Samples were taken from the feed and
permeate streams and analyzed for micro-organism concentrations.
Sample collection times
Microbial broths were prepared and injected upstream
of five membrane systems. A protocol was developed
and implemented to assure steady-state conditions were
Figure 1. Representation of a pressure driven membrane module showing flow (Q), concentration (C), pressure (P), water (Kw )
normalize mass transfer coefficients and solute (Ks) mass transfer coefficients.
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002
456
Figure 2.
LOVINS ET AL.
Size exclusion and diffusion controlled log rejection as a function of permeate concentration, flux, and recovery.
achieved during each test. Use of a precision chemical
metering pump provided a continuous microbial feed for
12 to 15 min. As shown in Fig. 3, the time required to
achieve steady state for each system was determined using a sodium chloride tracer test. Based on the tracer
study, micro-organisms were fed to directly to the membranes for 12 to 15 min before collecting replicate feed
and permeate samples. Similar testing confirmed these
sampling times were also appropriate for the MF and UF
systems.
Commercially available membranes
Three different nanofilters—one ultrafilter and one microfilter—were tested. The characteristics of these membranes are shown in Table 1. Three different NF systems
were examined and compared for varying organisms, concentrations, flux, recovery, and temperature. NF membranes included one cellulose acetate (CA) and two composite thin films (CTF). These membranes are identified
as NF1, NF2, and NF3, and correspond to manufacturer
models LFC1, CALP, and DK-C, respectively. The UF
and MF membranes had CA and polypropylene-based
films, respectively. The UF and NF2 membranes were removed and replaced on two occasions for autopsy. The
same NF1 and NF3 membranes were utilized throughout
testing.
Micro-organisms
Bacterial removal was demonstrated using Clostridium
perfringens (strain 26) spores (,1–5 mm), while model
bacterial viruses MS-2 (~0.025 mm) and PRD-1 (,0.1
mm) phage simulated virus retention. Protozoan removal
was demonstrated by challenge with C. parvum oocysts
(,4–6 mm) and Giardia lamblia cysts (,8–14 mm). Seed
levels were targeted to allow evaluation of 5- to 6-log assessment of Cryptosporidium and Giardia and 6 to 8-log
assessment of MS-2, PRD-1, and Clostridium. Membrane
disinfection capability was evaluated using microbial log
removal values (LRVs). Analytical sample sizes were 1
mL for MS2 and PRD1, 200 mL for Clostridium, and
100 mL for Cryptosporidium and Giardia. These aliquots
were taken from a ,2-liter volume sampled during challenge testing.
Challenge tests were conducted on eight occasions. Challenge tests were typically done with a mixture of microorganism, which did not involve all micro-organisms for
every challenge. The challenges were completed in 1 or
2 days, but sometimes required longer times when organisms were individually challenged. Cryptosporidium
challenges were done individually. A total of 68-log rejection values (LRV) were determined for the five different organisms from the eight challenge tests and five
membrane systems. The membrane pilot systems UF,
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
457
Figure 3. Premixer, postmixer and NF permeate tracer test results for pilot systems at American Water Services Water Treatment Plant at East St. Louis, IL.
MF, NF1, NF2, and NF3 were challenged directly before
any pretreatment.
Sample collection
Micro-organisms tested included bacterial viruses MS-2
and PRD-1, C. perfringens (strain 26) spores, Giardia
lamblia cysts, and Cryptosporidium muris oocysts. Samples for E. coli, Clostridium, MS-2, and PRD-1 analysis
were stored at 4°C and analyzed within 48 h after collection. Giardia and Cryptosporidium samples were preserved using formalin (10% formaldehyde final) and
stored at 4°C until analyzed.
Growth and purification of bacteriophages MS-2
and PRD-1 stocks
Male-specific virus MS-2 (ATCC # VR31) and its bacterial host E. coli C3000 (ATCC #15597) were purchased
from ATCC (Rockville, MD). Bacteriophage PRD-1 and
its host Salmonella typhimurium were generously supplied by Dr. Charles Gerba, University of Arizona (Tucson, AZ). Stocks of each bacterial host were produced by
supplementing with glycerol (20% final) 3- to 4-h tryptone broth cultures of each host and freezing 1-mL aliquots at 270°C. Stocks of each phage were produced by
adding approximately 1 3 105 plaque forming units
Table 1. Membrane manufacturer specifications for four membranes employed during pilot testing.
Parameter
Manufacturer
Membrane
MWC-(DA)
Configuration
Flow
Material
Filtration area
Max pressure
UF
NF-1
NF-2
NF-3
Aquasource
UF
100,000
Hollow fiber
Crossflow
CA derivative
155 sf/15 sm
29 psi/2 br
Hydranautics
NF (LFC1)a
100–200
Spiral wound
Crossflow
TFC
7.9 sf/85 sm
Fluid Systems
NF (CALP)
300
Spiral wound
Crossflow
CA
11.4 sf/123 sm
Desal
NF (DK-C)
150–300
Spiral wound
Crossflow
TFC-Polypeprizine
9 sf/97 sm
400 psi/27 bar
a Prototype membrane.
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002
458
LOVINS ET AL.
(pfu)/mL of each phage to a test tube containing 4 mL
of molten (46°C) tryptone top agar and 0.1 mL of a 3-h
culture of the appropriate host. Tube contents were
mixed, poured onto tryptone-bottom agar plates, and incubated at 37°C overnight. Plates with semiconfluent
plaques were flooded with 5-mL volumes of saline-calcium solution (8.5 g NaCl and 0.22 g CaCl2 per 1,000
mL MilliQ water), and incubated at 37°C for 1 h to allow migration of viruses into the buffer. The phage-buffer
mixture was harvested by pipette, centrifuged at 4,000 3
g to remove dislodged agar, and filtered through a 0.22
mm sterile filter. The phage was tittered and stored at 4°C
until used.
freshly steamed media. After 16 h of incubation at 37°C,
the cultures were heat treated at 60°C for 10 min to destroy vegetative cells. The suspension was immediately
cooled on ice, spores harvested by centrifugation at
4,000 3 g for 30 min (4°C), washed twice with ice-cold
buffered water, and suspended in buffered water.
Bacteriophage MS-2 and PRD-1 assay
Giardia and Cryptosporidium assay
Serial 10-fold dilutions of each sample were made in
a saline–calcium solution and kept on ice. Appropriate
dilutions were tested in triplicate by adding 0.1 mL of diluted sample to test tubes containing 3 mL of molten tryptone top agar (46°C) and 0.1 mL of a 3-h culture of E.
coli C3000 or S. typhymurium host. Tube contents were
mixed, immediately poured onto tryptone-bottom agar
plates and incubated overnight at 37°C.
Density gradient purified G. lamblia cysts and C. muris
oocysts were purchased from Waterborne, Inc. (New Orleans, LA). Giardia cysts and Cryptosporidium oocysts
preserved with formalin were identified and counted by
an indirect immunofluorescence antibody (IFA) procedure.
Preparation of bacterial stock suspensions
C. perfringens strain 26 (ATCC #3624) stocks were
stored at 4°C as sporulated cultures in buffered water.
Briefly, C. perfringens was inoculated into thioglycollate
broth (Difco, Detroit, MI) and grown overnight at 37°C.
The overnight culture was transferred into fresh thioglycollate broth, grown at 37°C for 4 h and inoculated into
Clostridium perfringens assay
Clostridium numbers after treatment were determined
by membrane filtration technique using mCP agar. Immediately before analysis samples were heat-treated at
60°C for 10 min to destroy vegetative cells, cooled on
ice, and appropriately diluted in ice-cold buffered water.
Microscopic examination
Total counts for each organism were made from scanning the entire sample membrane, and counts per mL
were calculated from the sample volumes applied to each
membrane examined. Two control slides were examined
along with every IFA assay of study samples. A negative membrane filter was prepared using PBS as the sample, and was examined for cysts and oocysts. The absence of cyst-like and/or oocyst-like objects was used to
Figure 4. NF1 log rejection vs. log organism size for Clostridium perfringens spores (,1–5 mM), MS-2 (,0.025 mM), PRD1 (,0.1 mM) phage, Cryptosporidium parvum oocysts (,4–6 mM), and Giardia lamblia cysts (,8–14 mM) challenge studies
conducted at American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
verify that samples were not contaminated during the IFA
procedure. A positive membrane filter was prepared using Giardia cysts and Cryptosporidium oocysts in PBS
as the sample. The presence of fluorescing cysts and
oocysts was used to verify that the IFA procedure was
properly performed.
Particle counting
Particle counts were measured using the HIAC/Royco
Model 8000A Particle Counter, which utilizes a light
scattering technique and is made by Pacific Scientific Instruments Co. (Silver Springs, MD). Particle counts were
measured upon collection via grab sampling. Values were
reported for bin size .2 mm.
RESULTS AND DISCUSSION
Membrane comparison
The LRV by micro-organism rejection and membrane
process are shown in Figs. 4 through 8. The size of the
organisms was approximated for these plots as follows:
G. lamblia cysts (12 mm), C. parvum oocysts (4 mm), C.
perfringens (3 mm), PRD-1 phage (0.1 mm), and MS-2
phage (0.025 mm). Many instances of complete rejection
are shown in Figs. 4 through 8 and are indicated as clear
rectangles. Although instances of complete rejection are
mathematically defined as infinite log removal, LRVs
459
were calculated for complete rejection assuming one organism passed the membrane. The dark rectangles represent detection of organisms in the permeate. Ideally,
the LRV is infinite until the rejection capacity of the
membrane is exceeded. At that point, the LRV would not
vary with increasing challenge concentrations and would
become constant. For example, a membrane with 4-log
rejection would show no organisms in the permeate until challenged with 4-log or greater micro-organisms.
Challenges of 4-, 5-, and 6-log micro-organisms would
result in 0-, 1-, and 2-log micro-organism concentration
in the permeate. Each LRV would be four. Hence, infinite and finite LRVs can be determined for the same organism and membrane. Finite LRVs can be obtained by
increasing feed concentration and/or taking larger permeate samples. Because micro-organism are so ubiquitous, testing can easily be altered by contamination or
false positives, more so than particles or trace chemicals.
However, these figures show the leveling off of LRV,
which demonstrates rejection capacity.
As shown in Fig. 4, the majority of LRV observations
by NF1 are infinite, which is complete removal of organisms. There were 22 total challenges of phage, bacteria,
and cysts to NF1. For 16 cases, no micro-organisms were
found in the permeate. On six occasions micro-organisms
were found in the permeate. Data shown in Fig. 4 suggest
micro-organism LRV was independent of organism size
and challenge concentration, which indicates a size exclusion mechanism for organism mass transfer. Mathemati-
Figure 5. NF2 log rejection vs. log organism size for Clostridium perfringens spores (,1–5 mM), MS-2 (,0.025 mM), PRD1 (,0.1 mM) phage, Cryptosporidium parvum oocysts (,4–6 mM), and Giardia lamblia cysts (,8–14 mM) challenge studies
conducted at American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002
460
LOVINS ET AL.
Figure 6. NF3 log rejection vs. log organism size for Clostridium perfringens spores (,1–5 mM), MS-2 (,0.025 mM), PRD1 (,0.1 mM) phage, Cryptosporidium parvum oocysts (,4–6 mM), and Giardia lamblia cysts (,8–14 mM) challenge studies
conducted at American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
cally the expected LRV for any of the micro-organisms
shown in Fig. 4 is infinity, as all had one or more observations of infinite rejection. Micro-organisms passed the
membrane only during two challenge events, which suggests operation may affect micro-organism rejection during nanofiltration. The smallest organism, MS-2 phage
passed NF1 only on two of seven occasions. C. perfringens passed NF1 on three of seven challenges, and is approximately 100 times larger than MS-2 phage. As the
typical pore size of NF1 is approximately 100 to 10,000
times smaller than C. perfringens and MS2 phage, other
factors such as the dynamic nature of membranes during
operation or analytical limits could affect micro-organism rejection by NF1 or membranes in general. However,
the data in Fig. 4 indicates NF1, a CTF NF, could reliably achieve 5-log rejection for the organisms tested,
which is 2.5 logs greater than the credits given for conventional coagulation, sedimentation, and filtration.
Figure 7. UF log rejection vs. log organism size for Clostridium perfringens spores (,1–5 mM), MS-2 (,0.025 mM), PRD-1
(,0.1 mM) phage, Cryptosporidium parvum Oocysts (,4–6 mM), and Giardia lamblia cysts (,8–14 mM) challenge studies conducted at American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
461
Figure 8. MF log rejection vs. log organism size for Clostridium perfringens spores (,1–5 mM), MS-2 (,0.025 mM), PRD-1
(,0.1 mM) phage, Cryptosporidium parvum oocysts (,4–6 mM), and Giardia lamblia cysts (,8–14 mM) challenge studies conducted at American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
LRV of micro-organisms for NF2, a cellulose acetate
NF, is shown in Fig. 5. Infinite rejection occurred on three
occasions for NF2, while micro-organisms were found in
the permeate on 19 occasions. Assessing reliable microorganism LR from the data shown in Fig. 5 is difficult.
Operation of the CA NF may have affected rejection as
pH control for continuous operation was at times difficult. However, the LRV does appear independent of organism size, and has a large range for C. perfringens (3
mm), PRD-1 phage (0.1 mm), and MS-2 phage (0.025
mm). For the conditions of these tests, NF2 microorganism rejection was always greater than 0.8 log, which
suggests very little if any micro-organism rejection could
be reliably credited to NF2. The NF2 LRV does appear
to average two for varying organism size and concentration, which again is indicative of a size exclusion mechanism.
Micro-organism log rejection for NF3, a CTF nanofilter, is shown in Fig. 6. Fewer challenges were made to
NF3 because of budget and schedule. Of the 13 NF3 challenges, six resulted in infinite rejection. Micro-organisms
were detected in the remaining seven challenges. The
least rejection was 4.5 log for MS-2 (0.025 mm), which
passed NF3 on three of four challenges. C. perfringens
(3 mm), and PRD-1 phage (0.1 mm) passed NF3 on two
of four occasions. However, the data in Fig. 5 shows the
upper limit of LRV is approximately 6, and that NF3
achieved 4.5 LRV on all occasions, which again suggests
a size exclusion mechanism. Comparing the minimum
LRV of NF3 to NF1, NF3 appears to reject about one
(0.9) logs less than NF1.
UF organism log rejection is shown in Fig. 7 by log organism diameter. Of eight challenges, micro-organisms
were detected in the permeate on two occasions and undetected on six occasions. The UF achieved 7-log for MS-2
phage (0.025 m) and C. perfringens (3 mm). Inspection
of the data in Fig. 7 indicates that UF organism rejection
for the conditions of the tests was reliably 7 log, and was
independent of organism size. Additionally, the UF membrane was made of a CA derivative indicating that CA
membranes are able to achieve a high organism LR. The
different organism LRV of the two CA membranes (NF2
and UF) are due to manufacturing and not material.
MF organism log rejection is shown in Fig. 8. The MF
that was utilized in this work was a small laboratory unit
adapted for field operation. Only one challenge was conducted on the MF. For the limited number of observations, Fig. 8 clearly shows increasing organism log rejection for increasing organism size (log organism
diameter). Given the nominally rated 0.1 mm MF pore
size, MS-2 phage (0.025 m) and PRD-1 (0.1 m) would be
expected to pass and the high 6.2 LRV for C. perfringens (3 mm) is not surprising. The MF results are consistent with results presented by other investigations (e.g.,
Jacangelo et al., 1995).
A summary of the LRVs for the entire study is shown
in Table 2. To avoid confusion regarding infinite rejection, the instances where complete are infinite rejection
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002
462
LOVINS ET AL.
Table 2. Approximate organism diameter, log diameter, and log rejection value (LRV) for membrane challenges conducted
at East St. Louis, IL.
NF1
NF2
NF3
UF
MF
Approximate
diameter
m
Log diameter
1
2
3
0.03
0.10
1.50
4.00
10.00
0.03
0.10
1.50
4.00
10.00
0.03
0.10
1.50
4.00
10.00
0.03
0.10
1.50
4.00
10.00
0.03
0.10
1.50
4.00
10.00
21.60
21.00
0.18
0.60
1.00
21.60
21.00
0.18
0.60
1.00
21.60
21.00
0.18
0.60
1.00
21.60
21.00
0.18
0.60
1.00
21.60
21.00
0.18
0.60
1.00
`3.2
`3
`2.6
5.4
5.7
5.7
`4.2
`6
`6.5
5.4
2.1
`3
2.1
1.4
0.8
1.8
1.6
`3.2
`3
`2.4
4.5
6.1
5.3
`4.1
4
5
6
7
6.8
6.2
6.1
`6.2
`6.4
`5.4
`5.8
`6.1
`5.4
`0.6
2.6
1.7
1.8
2.1
1.6
1.8
4.3
5.6
`4.4
4.6
4
4.4
`0.6
5.5
`6.5
`6.4
6
6.5
6.1
7
`7.9
7
`4.8
`4.7
8
`3.7
`3.8
`3.7
1.5
`6.5
`6.4
`6
2.5
3.5
6.2
`EX 5 Infinite rejection given X log challenge.
occurred are preceded by `. Such that a 4-log challenge,
which resulted in complete rejection is noted as `4.0.
This would read infinite or complete rejection for a 4-log
challenge, and is done to avoid the appearance of totally
infinite rejection. The concentration of micro-organisms
varied as did flux and recovery in these studies. As noted,
to calculate LRV one organism was assigned to all
micro-organism permeate concentrations when complete
rejection occurred. Infinite rejection was observed until
the rejection capacity of the membranes are exceeded.
Once organism passage was observed, the LRV was essentially constant. As shown previously in Fig. 2, lack of
variation for the LRV given varying concentration, flux
and recovery indicates a size exclusion mechanism for
organism rejection. If the mechanism were diffusion controlled, the organism concentration in the permeate would
vary with increasing concentration and changing flux and
recovery.
There are reasons for microbial passage of membranes, which include defects and the inherent nature of
construction of spiral wound and hollow-fiber membrane elements. A defect is defined as a large hole in
the membrane film, glue line, seal, or other parts of the
membrane element. The inherent nature of the membrane is defined as the mass transfer characteristics of
these same parts, not just the film. Although membrane
passage of any organism larger than the nominal pore
size of any membrane may be attributed to a defect, it
is more probably is due to inherent mass transfer characteristics of the membrane parts. Given adequate challenge concentrations or sample volumes, membranes
typically demonstrate very high, but not absolute rejection. Separation of element mass transfer characteristics
from defects is difficult, and is of little consequence to
users as membrane performance is viewed as a whole.
However, it is not justified to assume passage of one
organism per 105 to 108 organisms is due to defects in
the membrane or element. It is more likely that there
are no defects in a element that achieves such high rejection.
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
Particle counting as a surrogate
for microbial disinfection
Turbidity and particles are commonly used as surrogates for monitoring microbial contaminants. These techniques provide an inexpensive and continuous method for
monitoring treatment performance; were evaluated at
East St. Louis. The UF system was fed particle counts of
,105 to 106 #/mL. This range of feed particle concentration is comparable to seed concentrations used for microbial challenge testing. UF particle log rejection ranged
from 3.65- to 4.23-log. Corresponding UF microbial
LRVs ranged from 5 to infinity (7.9 if one organism passage is assume) for all organisms. Cryptosporidium and
Giardia are in the size range measured for particle counts,
.2 mm. UF Cryptosporidium and Giardia LRVs were
`4.7, `4.8, and 2.4. In contrast, particle rejection was not
complete for any observations. Over half of the challenge
tests performed on the UF (9 of 15) resulted in complete
pathogen rejection that includes organisms significantly
smaller than 2 mm. These results show UF microbial rejection did not vary with organism size.
Paired particle count and microbial log rejection data
is shown in Fig. 9 for all membranes. In this figure particle count log removals vary within a narrow range while
microbial LRVs vary to a much greater extent. The apparent discrepancy between particle count and microbial
rejection is significant given that many regard particles
as an effective on-line device for monitoring integrity.
463
This was evaluated directly by Miller and coworkers
(1999), who investigated particle and micro-organisms
rejection by MF. They concluded that particle counting
did not provide adequate sensitivity for detection of compromised filter fibers (Hong et al., 2001). The lack of
agreement between particles and microbials may be explained by contamination due to particle sloughing from
the membrane. Nonetheless, use of particle counts as a
surrogate for microbials is limited.
CONCLUSIONS
The results of this study confirm that CTF NF membrane processes are capable of providing excellent removal of microbes. This is particularly significant as several states do not give log credits to spiral wound
membranes for rejection of regulated micro-organism.
This regulatory policy severely limits utilization of membranes by utilities deprives the consuming public of the
benefits of membrane technology. These results clearly
show that membranes reject micro-organisms far greater
than regulatory requirements.
Significant findings resulting from these investigations
were:
1. The composite thin-film nanofilters tested (NF1 and
NF3) proved highly successful for removal of microbial pathogens, and exhibited clearly superior perfor-
Figure 9. Microbial log removal vs. particle log removal for challenge studies For MF, UF, and CTF membranes conducted at
American Water Works Service Company Water Treatment Plant at East St. Louis, IL.
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002
464
2.
3.
4.
5.
LOVINS ET AL.
mance to the cellulose acetate NF membrane (NF2).
In general, the composite thin-film NF exhibited 4.5to 5.5-log rejection for all organisms with over half
of the challenges resulting in complete rejection. In
contrast, rejection by the cellulose acetate CA was
much lower, and microbes were detected in the NF2
permeate in almost all cases.
Although the composite thin-film NFs demonstrated
excellent micro-organism rejection, micro-organism
did pass the membranes, which shows membranes
(and no process) should not be regarded as an absolute
barrier to pathogens.
The microfiltration membrane was capable of much
greater rejection of bacteria than viruses indicating
variation by size. MF virus removal was 2.5 to 3.5 log
(MS-2 and PRD-1, respectively), and 6.2 log for bacteria (Clostridium).
The cellulose acetate ultrafilter provided disinfection
several orders of magnitude greater than the cellulose
acetate NF suggesting disinfection by cellulose acetate
membranes is a function of construction rather than
membrane film.
NF and UF disinfection capability indicated consistent removal capability regardless of organism type or
size.
ACKNOWLEDGMENTS
The authors would like to acknowledge the American
Water Works Association Research Foundation and PAC
chairman Roy Martinez. The authors also acknowledge
the valuable support provided by the American Water
Works Service Company. Key AWWSC team members
were Dr. Mark LeChevallier, Dr. Morteza Abassadegan,
Kimberly Ajy, Robert Kozic, Esther Dundore and Brent
Gregory.
REFERENCES
ALLGEIER, S. (2001). Overview of Regulatory Issues Facing
Microfiltration and Ultrafiltration. In: Proceedings of the
AWWA Membrane Conference, San Antonio, TX.
CHELLAM, S., and Taylor, J.S. (2001). Simplified analysis of
contaminant rejection during ground- and surface water
nanofiltration under the information collection rule. Water
Res. 35(10), 2460–2474.
COLVIN, C.K., ACKER, C.L., MARINAS, B.J., and LOZIER,
J.C. (1999). Mechanisms Responsible for the Passage of Microorganisms through RO and NF Membranes. In: Proceedings of the AWWA Membrane Conference, Long Beach, CA.
COLVIN, C., BRAUER, R., DINATALE, K., and SCRIBNER,
T. (2001). Comparing laser turbidimetry with conventional
methods for monitoring MF and UF membrane integrity. In:
Proceedings of the AWWA Membrane Conference, San Antonio, TX.
FOX K.R., and LYTLE, R. (1996). The Milwaukee cryptosporidiosis outbreak: Investigation and recommendations,
J. AWWA 88, 87.
GERBA, C.P., and ROSE, J.B. (1989). Viruses in source and
drinking water. In G.A. McFeters, Ed., Advances in Drinking Water Microbiology Research. Madison, WI: Science
Tech, p. 19.
GLOVER, N. (1994). The isolation and identification of Mycobacterium avium complex (MAC) recovered from Los Angeles potable water, a possible source of infection in AIDS
patients. J. Environ. Health Res. 4, 63–72.
HANCOCK, C.M., ROSE, J.B., and CALLAHAN, M. (1998).
Crypto and Giardia in US groundwater. J. AWWA 90(3),
58–61.
HONG S., MILLER, F., and TAYLOR, J. (2001). Assessing
pathogen removal efficiency of microfiltration by monitoring membrane integrity. Water Sci. Technol. Water Supply
1(4), 43–48.
HONG, S.K., TAYLOR, J.S., ROSE, J., GIBBSON, C.,
OWEN, C., and JOHNSON, W. (1999). Removal of microorganisms by MF process: Correlation between membrane integrity test results and microbial removal efficiency.
In: Proceedings of the AWWA Membrane Conference, Long
Beach, CA.
HOXIE, N.J., DAVIS, J.P., VERGERONT, J.M., NASHOLD,
R.D., and BLAIR, K.A. (1997). Cryptosporidiosis-associated
mortality following a massive waterborne outbreak in Milwaukee, Wisconsin. Am. J. Public Health 12, 2032–2035.
JACANGELO, J.G., ADHAM, S.S., and LAINE, J.-M. (1995).
Mechanism of Cryptosporidium, Giardia and MS2 virus removal by MF and UF. J. AWWA 87, 9.
KRUITHOF, J.C., KAMP, P.C., and FOLMER, H.C. (2001).
Membrane integrity monitoring at the UF/RO Heemskerk
plant. In: Proceedings of the AWWA Membrane Conference,
San Antonio, TX.
LECHEVALLIER, M.L. (1996). What do studies of potable
water system groundwater studies tell us? In: Proceedings of
Groundwater Foundation’s 12th Annual Fall Symposium
Under the Microscope: Examining Microbes in Groundwater, Lincoln, NE.
LISLE J.T., and ROSE, J.B. (1995). Cryptosporidium contamination of water in the USA and UK: A mini review. J. Water Supply Res. Technol. 44, 103–105.
MCNAMARA, T.D., and GAVIN T.C. (1997). Microfiltration.
Case study; Marquette, Michgan. In: Proceedings of the
AWWA Membrane Conference, New Orleans, LA.
MILLER, F.A. (1999). Verification of membrane integrity testing for microfiltration: Correlation between integrity tests
MICRO-ORGANISM REJECTION BY MEMBRANE SYSTEMS
and microbial removal efficiency, Master’s Thesis, University of Central Florida, Orlando, FL.
MOURATO, D., NEWCOMBE T., and LESLIE, G. (1997).
Applications of immersed membranes in the drinking water
field. In: Proceedings of the AWWA Membrane Conference,
New Orleans, LA.
MULFORD, L.A., TAYLOR, J.S., NICKERSON, D., and
CHEN, S.S. (1999). Comparison of full- and pilot-scale
nanofiltration on planet performance. J. AWWA 91, 64–
75.
465
SHETTY, G.R., SHARMA, R.R., and CHELLAM, S. (2002).
Scale-up for municipal water nanofiltration based on permeate water quality. Environ. Eng. Sci. 19, 553–570.
TAYLOR, J.S., and JACOBS, E.P. (1996). Nanofiltration and
reverse osmosis. In J. Mallevialle, P.E. Odendaal, and M.R.
Wiesner, Eds., Water Treatment Membrane Processes. New
York, McGraw-Hill, p. 9.1.
TAYLOR, J.S., MULFORD L.A., DURANCEAU, S.J., and
BARRETT W.M. (1989). Cost and performance of a membrane pilot plant. J. AWWA 81(11): 52–60.
NOVICK, R.J. (1990). Nontuberculous mycobacterial infections in heart transplant recipients: A seventeen year experience. J. Heart Transplant, 9, 357–363.
TAYLOR, J.S., and WIESNER, M.W. (1999). Membranes. In
R.D. Letterman, Ed., Water Quality & Treatment. New York,
McGraw-Hill, p. 11.1.
SEYDE, V., CLARK, D., AKIYOSHI, E., KALINSKY A., and
SPANGENBERG, C.W. (1999). Nanofiltration process microbial challenge studies. In: Proceedings of the AWWA
Membrane Conference, Long Beach, CA.
URBAIN, V., and MANHEM, J. (1997). Membrane bioreactor
technologies for disinfection, denitrification and organic matter removal in drinking water. In: Proceedings of the AWWA
Membrane Conference, New Orleans, LA.
ENVIRON ENG SCI, VOL. 19, NO. 6, 2002