You are on page 1of 179

Evaluation of Membrane

Characterization Methods

Web Report #4102


Subject Area: Water Quality

Evaluation of Membrane
Characterization Methods

2012 Water Research Foundation. ALL RIGHTS RESERVED.

About the Water Research Foundation


The Water Research Foundation is a member-supported, international, 501(c)3 nonprofit organization that
sponsors research that enables water utilities, public health agencies, and other professionals to provide
safe and affordable drinking water to consumers.
The Foundations mission is to advance the science of water to improve the quality of life. To achieve this
mission, the Foundation sponsors studies on all aspects of drinking water, including resources, treatment,
and distribution. Nearly 1,000 water utilities, consulting firms, and manufacturers in North America and
abroad contribute subscription payments to support the Foundations work. Additional funding comes
from collaborative partnerships with other national and international organizations and the U.S. federal
government, allowing for resources to be leveraged, expertise to be shared, and broad-based knowledge
to be developed and disseminated.
From its headquarters in Denver, Colorado, the Foundations staff directs and supports the efforts of
more than 800 volunteers who serve on the Board of Trustees and various committees. These volunteers
represent many facets of the water industry, and contribute their expertise to select and monitor research
studies that benefit the entire drinking water community.
Research results are disseminated through a number of channels, including reports, the Website,
Webcasts, workshops, and periodicals.
The Foundation serves as a cooperative program providing subscribers the opportunity to pool their
resources and build upon each others expertise. By applying Foundation research findings, subscribers
can save substantial costs and stay on the leading edge of drinking water science and technology. Since
its inception, the Foundation has supplied the water community with more than $460 million in applied
research value.
More information about the Foundation and how to become a subscriber is available at www.WaterRF.org.

2012 Water Research Foundation. ALL RIGHTS RESERVED.

Evaluation of Membrane
Characterization Methods

Prepared by:
Amy E. Childress
Department of Civil and Environmental Engineering, University of Nevada, Reno, NV 89557
Jonathan A. Brant
Department of Civil and Architectural Engineering, University of Wyoming, Laramie, WY 82071
Pawel Rempala
Department of Civil and Environmental Engineering, University of Nevada, Reno, NV 89557
Donald W. Phipps Jr.
Orange County Water District/Research and Development Department, 10500 Ellis Avenue,
Fountain Valley, CA 92708-8300
and
Pierre Kwan
HDR Engineering, Inc., 500 108th Avenue NE, Suite 1200, Bellevue, WA 98004-5549

Jointly sponsored by:


Water Research Foundation
6666 West Quincy Avenue, Denver, CO 80235
and
U.S. Environmental Protection Agency
Washington DC 20460
Published by:

2012 Water Research Foundation. ALL RIGHTS RESERVED.

DISCLAIMER
This study was funded by the Water Research Foundation (Foundation) and the U.S. Environmental
Protection Agency (USEPA) under Cooperative Agreement No. X-83294801-1. The Foundation
and USEPA assume no responsibility for the content of the research study reported in this
publication or for the opinions or statements of fact expressed in the report. The mention of
trade names for commercial products does not represent or imply the approval or endorsement
of the Foundation or USEPA. This report is presented solely for informational purposes.

Copyright 2012
by Water Research Foundation
ALL RIGHTS RESERVED.
No part of this publication may be copied, reproduced
or otherwise utilized without permission.

2012 Water Research Foundation. ALL RIGHTS RESERVED.

CONTENTS

TABLES ....................................................................................................................................... vii


LIST OF FIGURES ....................................................................................................................... ix
FOREWORD ................................................................................................................................ xv
ACKNOWLEDGMENTS .......................................................................................................... xvii
EXECUTIVE SUMMARY ......................................................................................................... xix
CHAPTER 1: INTRODUCTION AND BACKGROUND ............................................................ 1
Contact Angle Measurements ............................................................................................. 1
Zeta Potential Measurements .............................................................................................. 6
Surface Roughness Atomic Force Microscopy (AFM) ................................................... 8
Transmission Electron Microscopy (TEM) ........................................................................ 9
Scanning Electron Microscopy (SEM) ............................................................................. 10
Chemical Force Microscopy (CFM) ................................................................................. 12
X-Ray Photoelectron Spectroscopy (XPS) ....................................................................... 14
Attenuated Total Reflection Fourier Transform Infrared Spectroscopy (ATR-FTIR) ..... 15
Objectives ......................................................................................................................... 16
CHAPTER 2: MATERIALS AND METHODS .......................................................................... 18
Membrane Samples........................................................................................................... 18
Statistical Analysis ............................................................................................................ 19
Membrane Community Survey On Membrane Characterization Techniques .................. 19
Development Of Standard Techniques For Characterizing Membrane Surfaces ............. 19
AFM Surface Roughness Measurements .................................................................... 20
Contact Angle Measurements ..................................................................................... 21
Streaming Potential Measurements............................................................................. 22
Assessment of Membrane Performance and Fouling ....................................................... 22
Temperature Correction Factors for Normalizing Membrane Permeate Flux ............ 25
Membrane Performance Testing ................................................................................. 28
Overview of Membrane Performance Test Method ................................................... 29
Normalizing Membrane Performance Response ........................................................ 29
CHAPTER 3: RESULTS AND DISCUSSION............................................................................ 31
Method Development For AFM Surface Roughness........................................................ 31
Imaging and Roughness Analysis of Flat Membrane Surfaces .................................. 31
Imaging and Roughness Analysis of Curved (Hollow Fiber) Membrane Surfaces.... 36
Standard Method for AFM Surface Roughness Measurements on Flat-Sheet and
Hollow Fiber Membranes ........................................................................................... 38
Method Development For Contact Angle Measurements ................................................ 38
Contact Angle Analysis Of The Flat-Sheet Teflon Standard Surface ........................ 38
Contact Angle Analysis Of Flat Sheet Membrane Surfaces ....................................... 41
Statistical Analysis Of Contact Angle Results For Flat Surfaces ............................... 44
Contact Angle Analysis Of Curved Surfaces.............................................................. 47
Standard Method For Contact Angle Measurements On Flat-Sheet And Hollow Fiber
Membranes.................................................................................................................. 48

v
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Method Development For Streaming Potential (Zeta Potential) Measurements .............. 48


PMMA Standard Surface ............................................................................................ 48
Streaming Potential (Zeta Potential) Analysis Of Flat Sheet Membrane Surfaces .... 49
Streaming Potential (Zeta Potential) Analysis Of Hollow Fiber Membrane Surfaces 52
Standard Method For Streaming Potential Measurements On Flat-Sheet And Hollow
Fiber Membranes ........................................................................................................ 53
Method Feedback From Partner Utilities.......................................................................... 53
Guidance For Integrity Test Calculations ......................................................................... 61
Evaluating Correlations Between Membrane Characteristics And Performance ............. 65
Multiple Linear Regression (MLR) Methods ............................................................. 70
Assessing the Predictive Ability of MLR Model Inputs by Iteratively Withholding a
Test Exemplar ............................................................................................................. 73
Assessing the Influence of MLR Model Input Parameters ......................................... 75
Modeling Conclusions ................................................................................................ 87
CHAPTER 4: CONCLUSIONS ................................................................................................... 89
Development Of Standard Techniques For Characterizing Membrane Surfaces ............. 89
Correlation Between Membrane Properties And Membrane Fouling ........................ 90
APPENDIX A LIST OF UTILITIES USING MEMBRANES AS A TREATMENT
TECHNOLOGY (AS OF 2010) ........................................................................................92
APPENDIX B SURVEY / QUESTIONNAIRE ON MEMBRANE CHARACTERIZATION
TECHNIQUES AND PROCEDURES ............................................................................117
APPENDIX C INSTRUCTIONS TO PARTICIPATING LABORATORIES FOR
CONDUCTING AFM, CONTACT ANGLE, AND STREAMING POTENTIAL
MEASUREMENTS .........................................................................................................126
APPENDIX D STANDARD PROTOCOLS FOR CHARACTERIZING MEMBRANE
SURFACES: CONTACT ANGLE, ZETA POTENTIAL, SURFACE ROUGHNESS .133
REFERENCES ........................................................................................................................... 152
ABBREVIATIONS .................................................................................................................... 154

vi
2012 Water Research Foundation. ALL RIGHTS RESERVED.

TABLES
Table 1.1. Distribution of adhesion across the HL and SG membrane surfaces as a function of
surface area coverage measured with methyl, carboxyl and hydroxyl functionalized probes (I =
0.01 M NaCl; pH 6; T = 20 C). Taken from Brant et al. (2006). ................................................ 12
Table 1.2 Measured atomic concentration percentages (%) of C1s, O1s, N1s and S2p obtained by Xray photoelectron spectroscopy (XPS) for commercially available and experimental membranes.
Taken from Boussu et al. (2007). ................................................................................................. 15
Table 2.1 Names and select properties of membranes used in this investigation (MFmicrofiltration; UF- ultrafiltration; NF- nanofiltration; RO- reverse osmosis, PP- polypropylene,
PES- polyethersulfone, PA- polyamide, TFC thin film composite). ......................................... 18
Table 2.2 Proposed standard surfaces for atomic force microscopy surface roughness
measurements. ............................................................................................................................... 21
Table 2.3 Proposed standard surfaces for contact angle measurements. ...................................... 21
Table 2.4 Summary of primary system components and accessories for the standard membrane
test unit. ......................................................................................................................................... 24
Table 2.5 Chemistry and composition of the feed solution used for membrane fouling
experiments25
Table 2.6 Summarized temperature correction factors (TCFs) for the NF-270, ESNA1-LF,
ESPA-2, SW30HR, and MUNIRO-400 membranes, which were used in the membrane fouling
experiments. .................................................................................................................................. 27
Table 3.1 Surface roughness statistics generated from atomic force microscope (AFM) imaging
of the ESNA1-LF membrane surface by the OCWD and DU laboratories (resolution = 256
256, measurement mode = tapping, medium = water). Numerals (1, 2, ..) correspond to a sample
site on the membrane. ................................................................................................................... 32
Table 3.2 Surface roughness statistics generated from atomic force microscope (AFM)
measurements of the SW30HR membrane surface from the OCWD and DU laboratories
(resolution = 256 256, measurement mode = tapping, medium = water). Numerals (1, 2, ..)
correspond to a sample site on the membrane. ............................................................................. 33
Table 3.3 Precision statistics for the surface roughness data that was collected for the ESNA1-LF
and SW30HR membranes (cell test results from one laboratory on one material, repeatability
standard deviation (sr), reproducibility standard deviation (sR). ................................................... 34
Table 3.4 Surface roughness statistics as measured using an atomic force microscope (AFM) for
the NF-270, ESNA1-LF, ESPA-2, and SW30HR membranes. All measurements were performed
on flat sheet membrane samples by the Orange County Water District (OCWD) laboratory. All
measurements were performed on wet membrane samples in contact mode and using a liquid
cell. ................................................................................................................................................ 34
Table 3.5 Surface roughness statistics for Teflon tubing, which was being considered as a
standard surface for analyzing hollow fiber membranes using atomic force microscopy
(resolution = 256 256, medium = water, OCWD contact mode, DU tapping mode). ......... 39

vii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.6 Contact angle data for the Teflon standard surface from the three participating
laboratories (Colorado School of Mines - CSM, Duke University DU, and the University of
Nevada, Reno UNR). All contact angles were measured using the captive bubble method and
doubly deionized water as the contact angle probe liquid (T = 20C). ........................................ 39
Table 3.7 Contact angle data for the ESNA1-LF membrane from the three participating
laboratories (Colorado School of Mines - CSM, Duke University DU, and the University of
Nevada, Reno UNR). All contact angles were measured using the captive bubble method and
doubly deionized water as the contact angle probe liquid (T = 20C). ........................................ 41
Table 3.8 Contact angle data for the SW30HR membrane from the three participating
laboratories (Colorado School of Mines - CSM, Duke University DU, and the University of
Nevada, Reno UNR). All contact angles were measured using the captive bubble method and
doubly deionized water as the contact angle probe liquid (T = 20C). ........................................ 42
Table 3.9 Average contact angle values for the PTFE standard surface, ESNA1-LF, and
SW30HR membrane. The associated statistical measures (repeatability and reproducibility) from
the inter-laboratory study are also reported for each test surface. ................................................ 45
Table 3.10 Reported and measured contact angle values for the 2D reference tool. All results
were acquired by the laboratory at the University of Nevada, Reno (UNR). ............................... 47
Table 3.11 Contact angle results for hollow fiber membranes. All results were obtained using the
experimental apparatus that was developed as part of this project. .............................................. 48
Table 3.12 Specific comments on the usefulness and technique for carrying out contact angle
measurements in membrane applications. .................................................................................... 56
Table 3.13 Specific comments on the usefulness and technique for carrying out streaming
potential measurements to calculate the zeta potential of membranes in water and wastewater
treatment applications. .................................................................................................................. 58
Table 3.14 Specific comments on the usefulness and technique for carrying out surface
roughness measurements using an atomic force microscope for membranes in water and
wastewater treatment applications. ............................................................................................... 61
Table 3.15 Multiple linear regression (MLR) model input parameters and values for describing
bentonite clay fouling. .................................................................................................................. 71
Table 3.16 Constants, coefficients and statistics for the best MLR model: J/Jo (at 5g m-2 Load) =
0.67233+ (2.2910-3 Jinitial, L m-2hr-1) (2.1210-3 Rq, nm) + (3.0010-3 , degrees).
....................................................................................................................................................... 72
Table 3.17 Values of K and J/Jo, Plateau obtained by fitting normalized permeate flux (J/Jo) data
and clay loading (g m-2) data for each membrane to equation Eq. 3.3. R-squared values and 95%
asymptotic confidence intervals are reported for each nonlinear regression membrane model80
Table 3.18 Membrane properties selected as independent variables for MLR analyses of K and
J/Jo, Plateau . The zeta potential values were determined at pH=5.3, 2mM KCl saturated with
ambient CO2. Surface roughness statistics were collected by the OCWD, while zeta potential and
water contact angle values were collected by UNR. Note: Rq was slightly cross-correlated with
Jo (r=0.5206) and with (r=0.5084)..81

viii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

LIST OF FIGURES
Figure 1.1 (a) Representative digital image of a liquid droplet on a dry surface in which the
contact angle () is measured according to the sessile drop method. (b) Representative digital
image of an air bubble on a wetted membrane surface (the membrane is immersed in a liquid)
and the contact angle () is measured according to the captive bubble technique. ........................ 2
Figure 1.2 Techniques commonly used for carrying out contact angle measurements: (a) static
sessile drop advancing and receding; (b) static captive bubble advancing and receding; and
(c) dynamic captive bubble. Figure adapted from Drelich (1997). For those illustrations for the
sessile drop technique the term advancing refers to the advancement of the liquid (increase in the
volume of the liquid droplet) over a dry surface, while the term receding refers to the recession
of the liquid droplet (decrease the in the volume of the liquid droplet) over a previously wet
surface. A similar relationship exists for the captive bubble measurements, where the water in
which the surface is immersed either advances or recedes across the surface through a decrease
or increase in the volume of the air bubble. .................................................................................... 3
Figure 1.3 The effect of drop (bubble) size on advancing (Adv.) and receding (Rec.) contact
angles for the air /water/polyethylene film system as obtained with the static sessile-drop (SD),
static captive-bubble (CB), and dynamic captive-bubble (DCB) techniques. Taken from Drelich
et al. (1996). .................................................................................................................................... 4
Figure 1.4 Illustration of the relative locations of the Stern layer and Shear plane from a charged
surface in water and the associated change in surface potential with distance. Taken from
Chapman-Wilbert et al. (1999). ...................................................................................................... 7
Figure 1.5 Schematic illustration of the operating principles associated with the atomic force
microscope (AFM). Taken from Wyart et al. (2008). .................................................................... 8
Figure 1.6 Transmission electron microscope (TEM) image of a graft copolymer film. Taken
from Kim et al. (2008). ................................................................................................................... 9
Figure 1.7 Ordered and disordered co-polymer blend morphologies visualized using TEMT
method. Taken from Jinnai et al. (2006)....................................................................................... 10
Figure 1.8 Scanning electron microscope (SEM) images of cross-sections of polymer blend
membranes. Taken form Li et al. (2008). ..................................................................................... 11
Figure 1.9 Topographical and corresponding CFM images for the HL membrane acquired with a
CH3-, COOH-, and OH-functionalized tip (I = 0.01 M NaCl; pH = 6.1, and T = 20 C). Taken
from Brant et al. (2006). ............................................................................................................... 13
Figure 1.10 Changes in adhesion force of carboxyl-modified microspheres (a) and hydroxylmodified microspheres (b) to PE membranes that were sampled during the pilot-filtration test.
An SEM image of a functionalized polystyrene bead glued to a cantilever tip used in this study is
shown on the right. Taken from Yamammura et al. (2008). ........................................................ 14
Figure 1.11 ATR-FTIR spectra of several commercial and one lab-made membrane, containing
sulfone groups. Ar-SO2-Ar typical frequencies are (S=O asym) = 1325 cm-1, (S=O symm) =
1140 cm-1. Differences in IR spectra stem from C(CH3)2 bridges present between aryl rings in
NTR7450 and Desal 5DL. Taken from Boussu et al. (2007). ...................................................... 16

ix
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 2.1. Process flow diagram of the bench-scale membrane test system used in the
correlation experiments for membrane surface properties and perfromance (P pressure gauge, T
temperature probe, F flow meter). .......................................................................................... 23
Figure 2.2. Normalized permeate flow (25C, flow at t = 0 set to 1) for membranes tested as a
function of time (hours). ............................................................................................................... 29
Figure 3.1. Rendered 3D images of the ESNA1-LF membrane based on atomic force microscope
(AFM) data from (a) OCWD and (b) DU (scan size = 100 m2, measurement mode = tapping,
medium = water). .......................................................................................................................... 32
Figure 3.2. Rendered 3D images of the SW30HR membrane based on atomic force microscopy
data from (a) OCWD and (b) DU (scan size = 100 m2, measurement mode = tapping, medium =
air). Note that the z-scale is different between the rendered 3D images (0.35 m/division for
Figure 3.2a compared to 0.305 m/division for Figure 3.2b). ...................................................... 33
Figure 3.3. Rendered 2D atomic force microscope (AFM) images of the ESNA1-LF, SW30HR,
NF-270, ESPA-2, and MUNIRO-400 membranes. All images were generated by the AFM
OCWD laboratory. ........................................................................................................................ 35
Figure 3.4. Plot of Ra as a function of Rq for the membranes investigated. .................................. 36
Figure 3.5. Box and whiskers plot for Teflon contact angle data: central bar mean, outside bars
minimum/maximum, boxes standard deviation range or confidence interval. The 95%
confidence interval for 6 measurements is very close to the sample standard deviation span, so
99% confidence intervals were plotted on the right. CSM Colorado School of Mines, DU
Duke University, UNR University of Nevada, Reno. All contact angles were measured using
the captive bubble method and doubly deionized water as the contact angle probe liquid (T =
20C). ............................................................................................................................................ 39
Figure 3.6. Digital images of air bubbles on the Teflon standard surface from each of the three
participating laboratories (Colorado School of Mines CSM, Duke University DU, and the
University of Nevada, Reno UNR). In each case the contact angle was measured using the
captive bubble technique in which the Teflon surface was immersed in doubly deionized water
(T = 20C). .................................................................................................................................... 40
Figure 3.7. Digital image of a water droplet (Vdrop = 20 L) on a Teflon standard surface set at an
incline............................................................................................................................................ 41
Figure 3.8. Box and whiskers plot for the ESNA1-LF contact angle data, which was measured
using the captive bubble technique. The 95% confidence interval for 6 measurements is very
close to the sample standard deviation span, so 99% confidence intervals were plotted on the
right. CSM Colorado School of Mines, DU Duke University, UNR University of Nevada,
Reno. ............................................................................................................................................. 42
Figure 3.9. Representative digital pictures of air bubbles from which contact angles were
measured on the ESNA1-LF membrane at each of the three participating laboratories (Colorado
School of Mines - CSM, Duke University DU, and the University of Nevada, Reno UNR). 43
Figure 3.10. Box and whiskers plot for the SW30HR contact angle data that was generated by the
Colorado School of Mines (CSM), Duke University (DU) and the University of Nevada, Reno
(UNR). All contact angles were measured using the captive bubble method and doubly deionized
water as the contact angle probe liquid (T = 20C). ..................................................................... 44
x
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.11. Representative digital images of air bubbles on the SW30HR membrane immersed
in water at the CSM, DU, and UNR laboratories. Each of these air bubbles are representative of
those from which contact angle measurements were performed. ................................................. 44
Figure 3.12. Picture of the 2D certified contact angle fixed-drop calibration reference tool
(Ram-Hart, Netcong, NJ). ........................................................................................................... 46
Figure 3.13. (left) Picture of a 3D contact angle reference tool, which includes the sample surface
stage, flat stainless steel plate, and stainless steel sphere. (right) Schematic illustration and
equations associated with using the 3D contact angle reference tool ( = contact angle measured
through the stainless steel sphere, D = diameter of the stainless steel sphere, H = distance from
the top of the sphere to the top elevation of the stainless steel cross member, and P = distance
from the top elevation of the sample stage to the top elevation of the stainless steel cross
member. ........................................................................................................................................ 46
Figure 3.14. Photographs of the experimental apparatus that was designed and used to measure
contact angles on hollow fiber membranes. (left and middle) Image of the PTFE tubing
sandwiched between two glass slides and secured in the sample stage. (right) Image of an air
bubble that has been placed on the PTFE tubing immersed in water. .......................................... 48
Figure 3.15. Zeta potential data collected for the PMMA control surface from UNR University
of Nevada, Reno (I = 2 mM KCl) and Anton Paar (I = 1 mM KCl). Polynomial fits (cubic) to
both data sets are also provided. . ................................................................................................. 49
Figure 3.16. Zeta potential as a function of solution pH for the ESNA1-LF membrane (I = 2 mM
KCl). Streaming potential measurements were by the UCR University of California Riverside,
UNR University of Nevada, Reno, the FKKT - University of Maribor, and Anton Paar. Those
data points from UCR, FKKT, and Anton Paar represent the average of three separate tests on
three different membarne coupons................................................................................................ 50
Figure 3.17. Zeta potential as a function of solution pH for the SW30HR membrane (I = 2 mM
KCl). Streaming potential measurements were by the UCR University of California Riverside,
UNR University of Nevada, Reno, the FKKT - University of Maribor, and Anton Paar. Those
data points from UCR, FKKT, and Anton Paar represent the average of three separate tests on
three different membrane coupons................................................................................................ 51
Figure 3.18. Zeta potential as a function of solution pH for the inner surface of the Norit X-Flow
hollow fiber membrane (I = 2 mM KCl, n = 3). ........................................................................... 53
Figure 3.19 Distribution of survey responses amongst each of the four respondent categories
(water/wastewater utilities, membrane manufacturers, characterization equipment vendors, and
academia). The total number of survey responses that were received was eleven. ...................... 54
Figure 3.20. Distribution of responses to the question on the usefulness of contact angle
measurements in membrane applications. .................................................................................... 55
Figure 3.21. Distribution of responses to the question on the usefulness of zeta potential values,
acquired through streaming potential measurements, in membrane applications. ....................... 57
Figure 3.22. Distribution of responses to the question on the usefulness of surface roughness
measurements, performed using an atomic force microscope, in membrane applications. .......... 60

xi
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.23 Difference in the calculated values for the minimum direct integrity test pressure
(Ptest) for obtaining a resolution of 3 m assuming a standard error in the contact angle value of
5. The values for Ptest were calculated using Eq. 3.1 ( = 1, = 74.9 dynes/cm, BPmax = 3 psi).
....................................................................................................................................................... 64
Figure 3.24. Results of simple regression analysis for the NF-270 membrane. 100 ml of permeate
corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.996635; R-squared =
99.3281 percent; R-squared (adjusted for d.f.) = 99.1042 percent; Standard Error of Est. =
0.02384; Mean absolute error = 0.015953; Durbin-Watson statistic = 2.77534 (P = 0.6576); Lag
1 residual autocorrelation = -0.434792. The central line represents the regression models, the
inner boundaries represent the 95% confidence limit and the outer boundaries represent the
model prediction limit. .................................................................................................................. 65
Figure 3.25. Results of simple regression analysis for the ESNA1-LF membrane:. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membranemembrane. Correlation Coefficient = 0.999234; R-squared = 99.8469 percent; R-squared (adjusted for d.f.) = 99.7959 percent;
Standard Error of Est. = 0.00762926; Mean absolute error = 0.00468329; Durbin-Watson
statistic = 2.83777 (P=0.7339); Lag 1 residual autocorrelation = -0.62428. The central line
represents the regression models, the inner boundaries represent the 95% confidence limit and
the outer boundaries represent the model prediction limit............................................................ 66
Figure 3.26. Results of simple regression analysis for the ESPA2 membrane. 100 ml of permeate
corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.998841; R-squared =
99.7683 percent; R-squared (adjusted for d.f.) = 99.691 percent; Standard Error of Est. =
0.0134216; Mean absolute error = 0.00948347; Durbin-Watson statistic = 1.50241 (P=0.0550);
Lag 1 residual autocorrelation = 0.0623037. The central line represents the regression models,
the inner boundaries represent the 95% confidence limit and the outer boundaries represent the
model prediction limit. .................................................................................................................. 67
Figure 3.27. Results of simple regression analysis for the SW30HR membrane. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.996067; Rsquared = 99.215 percent; R-squared (adjusted for d.f.) = 98.9534 percent; Standard Error of Est.
= 0.00316966; Mean absolute error = 0.0020736; Durbin-Watson statistic = 2.6779 (P=0.5877);
Lag 1 residual autocorrelation = -0.561779. The central line represents the regression models, the
inner boundaries represent the 95% confidence limit and the outer boundaries represent the
model prediction limit. .................................................................................................................. 68
Figure 3.28. Results of simple regression analysis for the MUNIRO-400 membrane. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.994937; Rsquared = 98.9899 percent; R-squared (adjusted for d.f.) = 98.8215 percent; Standard Error of
Est. = 0.00624946; Mean absolute error = 0.00416953; Durbin-Watson statistic = 2.42804
(P=0.5686); Lag 1 residual autocorrelation = -0.352731. The central line represents the
regression models, the inner boundaries represent the 95% confidence limit and the outer
boundaries represent the model prediction limit. .......................................................................... 69
Figure 3.29. Normalized membrane flux (J/Jo) predicted by the simple regression models as a
function of bentonite clay loading. From this, membrane performance of all five membranes at at
a given clay load can be determined. A 5g m-2 bentonite load was chosen with which to generate
performance data for MLR analysis because this falls within a nearly linear portion of all five

xii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

relationships and membrane performance could be determined in all cases at this point by
interpolation of experimental data. ............................................................................................... 70
Figure 3.30. Results: Best MLR model prediction and UNR membrane data determined from
laboratory measurements. Statistics: R-Squared = 99.2736%; Adj. R-Squared = 97.0943%; Std.
Error of Est. = .00718711; M.A.E = .00251155; P-Value = 0.1069. Horizontal bars = 95%
confidence interval for J/Jo @ 5 g m-2 clay load estimated from laboratory data; vertical bars =
MLR model standard error of the estimate. .................................................................................. 73
Figure 3.31. Comparison of the prediction of membrane behavior by the MLR model constructed
from all the exemplars (filled circles) with prediction of five MLR models that were constructed
using data missing each of the individual membranes indicated (open circles). Horizontal bars =
95% confidence interval for J/Jo @ 5 g m-2 clay load estimated from laboratory data; vertical
bars = MLR model standard error of the estimate. ....................................................................... 75
Figure 3.32. Illustration of the decline in normalized permeate flux rate (J/Jo) with increasing
clay load. The plateau value (J/Jo, Plateau) is taken as a measure of the severity of the impact of
bentonite clay on the permeate flux rate for a membrane and is determined as the point where
J/Jo becomes zero order with respect to clay loaded. .................................................................... 78
Figure 3.31. Intrinsic water flux (J/Jo) as a function of the clay loaded (g m2) for the a) NF-270
and b) MUNIRO-400 membranes. Lines show Eq. 3.3 fitted to the data by nonlinear regression.
....................................................................................................................................................... 79
Figure 3.33. Linearly regressed relationship between J/Jo and K. The intervals are the asymptotic
95% confidence intervals from the individual nonlinear models. ................................................ 80
Figure 3.34. Plot of the observed and predicted K values for each of the different membranes
fouled by bentonite clay. R-squared = 98.82%, adjusted R-squared = 95.27%, p-value = 0.1363.
The value of K determined in the nonlinear regression models for the individual membranes
loaded with bentonite clay could be predicted well from roughness, initial water flux and contact
angle data. Zeta potential data were not required for the prediction (probably due to small
variation amongst the test membranes)......................................................................................... 82
Figure 3.35. Plot of the observed and predicted J/Jo Plateau values for each of the different
membranes fouled by bentonite clay. R-squared = 97.83%, Adjusted R-squared = 91.33%, pvalue = 0.1843. The MLR model was able to describe the majority of the variance in J/Jo Plateau
using roughness, contact angle and initial water flux. As with K, zeta potential data were not
required to explain the variance observed in J/Jo Plateau. ............................................................... 86
Figure 3.36. Mechanistic explanation of fouling by bentonite clay. ............................................ 87

xiii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2012 Water Research Foundation. ALL RIGHTS RESERVED.

FOREWORD
The Water Research Foundation (Foundation) is a nonprofit corporation dedicated to the
development and implementation of scientifically sound research designed to help drinking
water utilities respond to regulatory requirements and address high-priority concerns. The
Foundations research agenda is developed through a process of consultation with Foundation
subscribers and other drinking water professionals. The Foundations Board of Trustees and
other professional volunteers help prioritize and select research projects for funding based upon
current and future industry needs, applicability, and past work. The Foundation sponsors
research projects through the Focus Area, Emerging Opportunities, and Tailored Collaboration
programs, as well as various joint research efforts with organizations such as the U.S.
Environmental Protection Agency and the U.S. Bureau of Reclamation.
This publication is a result of a research project fully funded or funded in part by
Foundation subscribers. The Foundations subscription program provides a cost-effective and
collaborative method for funding research in the public interest. The research investment that
underpins this report will intrinsically increase in value as the findings are applied in
communities throughout the world. Foundation research projects are managed closely from their
inception to the final report by the staff and a large cadre of volunteers who willingly contribute
their time and expertise. The Foundation provides planning, management, and technical
oversight and awards contracts to other institutions such as water utilities, universities, and
engineering firms to conduct the research.
A broad spectrum of water supply issues is addressed by the Foundation's research
agenda, including resources, treatment and operations, distribution and storage, water quality and
analysis, toxicology, economics, and management. The ultimate purpose of the coordinated
effort is to assist water suppliers to provide a reliable supply of safe and affordable drinking
water to consumers. The true benefits of the Foundations research are realized when the results
are implemented at the utility level. The Foundation's staff and Board of Trustees are pleased to
offer this publication as a contribution toward that end.

Roy L. Wolfe, Ph.D.


Chair, Board of Trustees
Water Research Foundation

Robert C. Renner, P.E.


Executive Director
Water Research Foundation

xv
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2012 Water Research Foundation. ALL RIGHTS RESERVED.

ACKNOWLEDGMENTS
The PIs and Key Research Team Members would like to acknowledge the participation
of Dr. Ryan Heck of UNR in confidentiality negotiations with Pall Microza and GE Water. We
would also like to thank OCWD participants, Jana Safarik, and Richard Bold for contact angle
and AFM imaging measurements. We would like to express our gratitude to the Duke
University team Professor Mark Wiesner, Zachary Hendren, and Dr. Soryong Chae for contact
angle and AFM measurements; the Colorado School of Mines team Professor Tzahi Cath and
Matt Bolt for zeta potential and contact angle measurements); the University of California,
Riverside team Professor Sharon Walker, Olgun Zorlu for zeta potential measurements; and the
University of Maribor team Dr. Irena Petrini for zeta potential measurements. Consultation of
Dr. K.C. Khulbe (University of Ottawa) with regard to AFM measurements is also
acknowledged. We greatly appreciate the assistance of Dr. Thomas Luxbacher of Anton Parr in
evaluating the zeta potential results and initiating an additional round robin of measurements.
Assistance from additional UNR personnel, Viktoriya Weirauch and Serife Ozger, is also
acknowledged. Lastly, we would like to acknowledge the WaterRF Project Manager, Jonathan
Cuppett and our Project Advisory Committee members, Michelle Chapman, Jess GarciaAleman, Carl Spangenberg, and Shahram Tabe.

xvii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2012 Water Research Foundation. ALL RIGHTS RESERVED.

EXECUTIVE SUMMARY
INTRODUCTION AND BACKGROUND
A variety of techniques and approaches are routinely used for characterizing the physical
and chemical properties of membrane surfaces. Measuring the contact angle that is formed when
a droplet of water is placed on the membrane surface is used to quantify the
hydrophobicity/hydrophilicity of the membrane surface. Atomic force microscopy is used to both
quantify and visualize the morphology or roughness of membrane surfaces. Streaming potential
measurements are used to calculate a zeta potential, an indicator of surface charge, for membrane
surfaces under variable solution chemistries. While numerous other characteristics may be
measured for membrane surfaces these three aforementioned ones have perhaps received the
greatest interest and use from scientists, engineers, and membrane manufacturers. This strong
interest is attributed to the relative importance of each of these properties in determining
permeate flux rates, solute rejection, and fouling characteristics for microporous (microfiltration,
ultrafiltration) and nonporous (nanofiltration, reverse osmosis) membranes. For example, it is
now widely accepted that the specific flux rate (m3 m-2 day-1 Pa-1) increases with increasing
hydrophilicity of the material that makes up the membrane. Similarly, hydrophilic and smooth
membrane surfaces tend to be less prone to severe fouling events. Contact angle with water, and
other liquids, in addition to zeta potential values are instrumental for describing the chemical
interactions that occur between colloids, microorganisms, and other materials and the membrane
surface. Such descriptions are generally made within the context of the Derjaguin-LandauVerwey-Overbeek (DLVO) model and its various extensions. Using this approach more specific
surface energy parameters, for example the Lewis acid-base values, for membranes may be
calculated using contact angle measurements. Despite the wealth of information that may be
derived from the aforementioned surface characterization techniques a standard method for
applying these techniques to membrane surfaces is lacking.
Surface characterization techniques rely on rather exhaustive sample preparation
procedures and a thorough understanding of the properties of the sample material. In the absence
of either of these requirements it is difficult at best to generate reproducible results. The situation
for polymeric membranes is further complicated by the lack of specific information about
membrane material composition/chemistry and the use of preservatives for long-term storage.
Not surprisingly these complicating factors have resulted in numerous disagreements between
any two labs studying the same membrane under similar operating conditions. Furthermore, in
light of the often-conflicting results it is nearly impossible to derive usable information from
these laboratory studies for utilities and other end users. It is therefore necessary that a standard
method be developed for conducting and interpreting the results from contact angle, streaming
potential and atomic force microscope measurements.
OBJECTIVES
The overall objective of this project was to develop standard methods for both conducting
and interpreting the results from contact angle, streaming potential, and atomic force microscope
measurements for characterizing membrane surfaces. The motivation for this overall objective is
the improvement of our ability to evaluate the role(s) of membrane properties in determining

xix
2012 Water Research Foundation. ALL RIGHTS RESERVED.

membrane performance. The specific research goals that were pursued to achieve the overall
objective of this research project were as follows:
1. Identify three techniques for characterizing the physical and chemical properties of
membrane surfaces,
2. Quantify the variability between data generated using the selected characterization
techniques in order to develop the associated standard methods for their use by the
membrane community,
3. Provide drinking water utilities with information on the value of membrane
characterization methods, a means by which to compare membrane properties on an
equal basis, and guidance for accurate contact angle measurements for integrity test
calculations,
4. Evaluate the ability of those surface characteristics determined using the proposed
standard methods to accurately predict membrane performance and fouling.
APPROACH
The following research tasks were conducted in order to achieve the overall objective of
this project:
Task A: Identification and review of characterization methods
Task B: Identification and selection of membranes for analysis
Task C: Development of standard methods for characterizing membrane surfaces
Task D: Obtain feedback from partner utilities on the proposed standard methods
Task E: Characterize select membrane properties using the proposed standard methods
Task F: Conduct membrane fouling studies
Task G: Evaluate relationships between membrane characteristics and membrane
performance
Task H: Develop guidance for utilities doing integrity test calculations
Following a comprehensive review of the available literature on membrane surface
characteristics and performance the following three characterization techniques were selected for
inclusion in this project: contact angle (surface energy), atomic force microscopy (AFM)
(surface morphology), and streaming potential (zeta potential) measurements. Membrane
samples that included microfiltration (MF), ultrafiltration (UF), nanofiltration (NF), and reverse
osmosis (RO) membranes were selected and acquired based on their use in the water treatment
industry. Standard (control) surfaces were also identified and acquired for each of
aforementioned characterization techniques that were studied as part of this project. Surface
samples (membranes and standard surfaces) were distributed, along with guidance on how to
conduct each type of characterization technique, to three separate laboratories. The participating
laboratories took measurements, assessed precision, and provided input on the proposed standard
techniques. Next, membrane fouling (or more specifically, membrane resistance to fouling) was
evaluated under laboratory conditions using a suspension of bentonite (clay) as the fouling
solution. Membrane properties were then correlated with membrane performance metrics.

xx
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Executive Summary | xxi

RESULTS AND CONCLUSIONS


The following conclusions were reached regarding the development and application of
standard techniques for characterizing membrane surfaces, and specifically for executing surface
characterization using contact angle, surface roughness, and streaming potential measurements:
Operating an AFM in contact mode produces more consistent roughness statistics when
imaging the outside of tubular structures. Therefore, when imaging hollow fiber
membrane surfaces, whether the concave or convex surface, the AFM should be operated
in contact rather than tapping mode.
A linear relationship exists between the root mean square and average roughness statistics.
Therefore, both parameters may be equally applied for drawing correlations between
membrane surface morphology and membrane fouling.
The inherent physical heterogeneity of membrane surfaces requires that greater than three
sites be imaged (i.e., their roughness measured) in order to develop an accurate assessment
of membrane surface roughness.
Surface roughness measurements using the AFM technique is a precise measure of
membrane surface morphology based on the reproducibility and repeatability of surface
roughness statistics between two laboratories. The precision of the measurement is,
however, based on the two laboratories using the standard technique.
Using the standard technique, the captive bubble method produces precise contact angle
results for membrane surfaces with a repeatability and reproducibility between any two
laboratories of approximately 5. This level of precision allows for equal comparison of
relative assessments of membrane surfaces with regards to their hydrophobicity (i.e., is the
membrane hydrophobic or hydrophilic). However, the current level of precision does not
lend itself well to the comparison of surface energy parameters calculated between two
laboratories using contact angles measured at each laboratory. Further study is needed to
determine how variations in contact angle results affect the magnitude and sign of surface
energy parameters determined through contact angle analysis to establish the broader
significance of these values to the membrane community (i.e., are these surface energy
parameters useful in membrane operations?).
The proposed standard method for determining the zeta potential of membrane surfaces
using streaming potential measurements produces relatively reproducible results
independent of the type of electrokinetic analyzer used. However, minor deviations from
the proposed standard method (e.g., differences in sample pretreatment) can result in
substantial errors.
The availability of characterization equipment and operator skill level requirements are the
greatest impediment to membrane characterization tests (contact angle, streaming
potential, surface roughness) being carried out by utilities.
From the perspective of utilities, characterization results, such as contact angle, surface
roughness, and streaming potential, are only useful during membrane selection and lose

xxi
2012 Water Research Foundation. ALL RIGHTS RESERVED.

their usefulness once membranes are put into service. A key challenge that was identified
in this regard is how to determine membrane characteristics without damaging the
membrane modules or elements (i.e., how to determine these characteristics without
conducting a destructive autopsy).
The following conclusions were reached regarding the correlation between membrane
surface properties determined using the proposed standard techniques and membrane fouling (or
more specifically, membrane resistance to fouling):
The permeate flux loss that was observed upon fouling of the five membrane samples with
a bentonite clay suspension at a constant feed pressure could be modeled statistically to
express normalized permeate flux as a function of the clay loading rate (5 g m-2). Based on
these observations a multiple linear regression (MLR) model was developed and found to
be capable of describing >97% of the observed variance in normalized product flux.
Those membrane surface properties that were determined to be necessary for describing
the observed membrane performance included root mean square roughness, contact angle
and initial permeate flux but not zeta potential. This implies that surface roughness and
hydrophobicity have a greater influence on bentonite clay cake density/water permeability
than do surface charge interactions. Overall, better membrane performance (higher
observed normalized flux after 5 g m-2 bentonite loading) was associated with greater
RMS roughness, higher initial water flux and greater contact angle (greater surface
hydrophobicity).
Membrane surface properties (roughness, hydrophobicity and charge) could be used to
predict membrane performance in the presence of foulant materials using numerical
modeling approaches.
APPLICATIONS AND RECOMMENDATIONS
The membrane and standard surface characterization results demonstrate the reproducible
and accurate AFM generated surface roughness, contact angle, and zeta potential results can be
generated independent of the type of characterization equipment used. The generation of such
data is however dependent on the use of standard methods and experimental techniques, which
are presented in this report. The absence of such standard methods and techniques is likely to
result in variability in the aforementioned measures of membrane surface characteristics.
Process Design and Operational Impacts
Improving the accuracy and reliability of membrane characterization results, in particular
contact angle with water (hydrophobicity) and zeta potential (charge) is expected to further
improve the process by which membranes are selected for different applications. While further
research is needed to fully realize the implications of improved characterization results for
predicting membrane performance and predicting membrane fouling, the preliminary results
given in this report suggest that it could substantially improve the accuracy of predictive fouling
models. The development, and ultimate application, of more accurate predictive membrane
performance models is likely to facilitate better membrane selection and operation. For example,

xxii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Executive Summary | xxiii

membranes may be selected whose surface chemistry is suited for minimizing fouling under a
given set of feed water quality conditions.
Data Interpretation
Use of the standard methods developed as part of this project can facilitate the generation
of reproducible contact angle, streaming potential (zeta potential), and AFM produced surface
roughness statistics for membrane surfaces. These findings appear to be independent of the make
and model of the respective characterization equipment used. However, data that is produced in
the absence of these standard methods are open to substantial variability and error. Therefore,
membrane characterization data, and any extensions (interfacial modeling), must be analyzed
and interpreted with great caution. This is particularly important in the absence of any
accompanying detailed information on the methods and techniques used. Based on these findings
it is highly recommended that laboratories carrying out membrane characterization studies
strictly adhere to the standard methods given in this report. Furthermore, consideration should be
given to developing a certification program for those laboratories seeking to carry out membrane
characterization work. Implementing these, and perhaps additional, standardization measures
will greatly improve the reproducibility and accuracy of membrane characterization results
(contact angle with water, zeta potential, and surface roughness statistics).
Regulatory Impacts
Based on the data collected as part of this effort the standard error in the measured
contact angle value for a given surface between any two labs is approximately 5. This
variability in contact angle results has implications for the calculated minimum direct integrity
test pressure for obtaining a resolution of 3 m (Ptest), which is required for direct integrity
testing of hollow fiber membranes. The difference in calculated Ptest values, assuming an error in
the measured contact angle of 5, is dependent on the magnitude of the measured contact angle.
Over a range on contact angle values of 0 to 90 and an error in the contact angle value of 5, the
error in the calculated Ptest value may range from 0.1 to 1.3 psi.

xxiii
2012 Water Research Foundation. ALL RIGHTS RESERVED.

PARTICIPANTS
A list of subcontractors, participating utilities, and other individuals who contributed to
this project are listed below. We gratefully acknowledge the valuable contributions to this project
from all of the participants.
Pierre Kwan, P.E.
HDR Engineering, Inc. (HDR)
Prof. Jonathan A. Brant
University of Wyoming
Don Phipps
The Orange County Water District (OCWD)
Greg Turman
City of Clearwater
Annika M. Bankston, P.E.
Minneapolis Water Works
Subcontractors, Participating
Utilities and Other Participants

Uzi Daniel
West Basin Municipal Water District
Prof. Mark Wiesner
Duke University
Prof. Tzahi Cath
Colorado School of Mines
Prof. Erik Hoek
University of California, Los Angeles
Carl Clegg
Ram-Hart Instruments Company
Thomas Luxbacher, Ph.D.
Anton Paar GmbH

xxiv
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 1
INTRODUCTION AND BACKGROUND
Membrane-based pressure driven processes constitute essential and mature technology
with numerous applications in water treatment and drinking water production. Despite advances
in membrane science, the application of membrane processes continues to be hampered by the
persistent problem of membrane fouling. Membrane fouling results from the deposition,
adsorption, and/or accumulation of rejected species on, or within the membrane, resulting in the
deterioration of water flux and quality.
Membrane processes are playing an ever-increasing role as environmental engineers
continue to seek sustainable methods for water treatment, wastewater reclamation, groundwater
restoration, and pollution prevention. Mitigation of membrane fouling and maximization of
water productivity are essential for optimizing membrane processes. Membrane fouling results
from the attachment, accumulation, or adsorption of substances onto the membrane surface
and/or within the membrane pores; it hinders membrane performance and shortens membrane
life. Physical and chemical interactions between solutes or particles and the membrane
interface substantially affect water productivity and the rate at which membrane fouling
occurs. Previous research has unambiguously linked membrane characteristics to the rate and
extent of membrane fouling, and therefore, to the water productivity and overall performance of
membranes. Contact angle, surface roughness, and streaming potential are three surface
characterization methods that appear to be relatively simple; the measurements appear to be
routine and the physics behind the equations describing these properties appears to be straightforward. However, literature on these surface characterization methods has shown that they are
quite complex and far from being completely understood, especially zeta potential via streaming
potential measurements (Yaroshchuk and Ribitsch, 2002).
CONTACT ANGLE MEASUREMENTS
The contact angle () that is formed at the three-phase interface between solid, liquid, and
gas/vapor phases (Figure 1.1) may be used to elicit information regarding membrane surface
energy properties. There are two accepted techniques for measuring contact angle, sessile drop
and captive bubble, as shown in Figure 1a and 1b, respectively. The information that is gathered
from contact angle analysis may be used to calculate specific surface energy properties (van der
Waals, Lewis acid-base) for detailed interfacial analyses, as well as for qualitatively assessing
the wettability, or hydrophobicity/hydrophilicity, of a membrane surface. This latter application
is perhaps the most common use for contact angle data by utilities and other membrane users.
Perhaps the greatest challenge with goniometric contact angle measurements is contact angle
hysteresis, which is the difference in the measured contact angle depending on whether it is an
advancing or receding measurement. Advancing measurements involve the spreading of a liquid
drop over a dry surface. Conversely, receding measurements involve the liquid droplet shrinking
and thus the contact angle is measured on an already wet surface. Receding measurements are
most commonly associated with the captive bubble technique in which an air bubble is placed on
a surface that is immersed in water. In this way, the water recedes from that volume that is

1
2012 Water Research Foundation. ALL RIGHTS RESERVED.

occupied by the air bubble (Figure 1.1b). The effect can be as large as a 20 difference in
advancing and receding contact angle values, but can also be negligible (Adamson and Gast,
1997; Garbassi et al., 1998). For a brief, recent account of the origin of contact angle hysteresis
see Gao and McCarthy (2006). Although theories of contact angle and contact angle hysteresis
are still controversial, the measurement does have numerous practical applications (e.g.,
calculation of surface energy components and pore-liquid entry pressures).

a)

b)

Figure 1.1 (a) Representative digital image of a liquid droplet on a dry surface in which the
contact angle () is measured according to the sessile drop method. (b) Representative
digital image of an air bubble on a wetted membrane surface (the membrane is immersed
in a liquid) and the contact angle () is measured according to the captive bubble
technique.
Perhaps the greatest advantages of contact angle measurements are the relatively simple
principles of the measurement technique and that the required equipment (goniometer) is
relatively inexpensive. Note that these relative comparisons are all being made to the
requirements for other characterization measurements, such as streaming potential and electron
microscopy techniques. That said, details associated with the algorithms that are used for tangent
line determination in digital images by vendor supplied software can be complicated and are not
usually revealed by the vendors.
From a practical point of view, the captive bubble technique, in which the gaseous phase
is introduced as a bubble under the membrane surface, is more viable than the sessile drop
technique, which requires rigorous drying of the membrane surface. However, with the captive
bubble technique, it is sometimes confusing as to which angle should be reported. As contact
angle is defined as the angle measured through the denser fluid phase, the contact angle through
the liquid phase should always be reported. This means that if the goniometer (because of object
configuration software requirements) measures the angle through the gas phase (*) then the
supplementary angle to the * angle (180- *) should be reported. Interestingly, two references
on this subject cited here (Adamson and Gast, 1997; Garbassi et al., 1998) have figures
suggesting otherwise; also an example in recent literature (Roudman and DiGiano, 2000) is
further indication of confusion on this issue. Another aspect of the contact angle measurement
that can be confusing is the distinction between advancing and receding angles. The use of these
terms is apparent for sessile drop measurements, but is sometimes confused for captive bubble
measurements. As can be seen, in Figure 1.2 attention should not be placed on whether the drop

2
2012 Water Research Foundation. ALL RIGHTS RESERVED.

or bubble is growing or shrinking, but instead on whether the liquid phase is advancing or
receding, because it is the solid-liquid contact angle that is of interest (Drelich, 1997).

a)
Gas
Liquid

Liquid

Advancing
contact angle

Receding
contact angle

b)
Gas

Gas
Liquid

Advancing
contact angle

Receding
contact angle

c)
Gas
Buoyancy force
Liquid
Pressure air

Receding or intermediate contact angle

Figure 1.2 Techniques commonly used for carrying out contact angle measurements: (a)
static sessile drop advancing and receding; (b) static captive bubble advancing and
receding; and (c) dynamic captive bubble. Figure adapted from Drelich (1997). For those
illustrations for the sessile drop technique the term advancing refers to the advancement of
the liquid (increase in the volume of the liquid droplet) over a dry surface, while the term
receding refers to the recession of the liquid droplet (decrease the in the volume of the
liquid droplet) over a previously wet surface. A similar relationship exists for the captive
bubble measurements, where the water in which the surface is immersed either advances
or recedes across the surface through a decrease or increase in the volume of the air
bubble.
Another important issue for contact angle measurement reproducibility is the effect of
drop and bubble size on contact angle. Drelich et al. (1996; 1997) performed a study on a variety
of surfaces (polymers, monolayers on gold, minerals) and made several conclusions. First, for
smooth, homogenous, nearly-ideal surfaces, contact angle does not depend on drop size (Figure
1.3), although hysteresis can remain detectable. Second, dynamic captive bubble results
approach results for receding sessile drop or bubble. Third, receding sessile drop results are most

3
2012 Water Research Foundation. ALL RIGHTS RESERVED.

prone to change with drop size on non-ideal surfaces. The advancing angle should be more
reproducible, as it depends on bubble/drop size to a lesser degree, on both smooth and imperfect
surfaces. However, information on surface quality that can be gained from receding angle/bubble
size relationship will be lost if only advancing angle is measured. Drelich et al. (1996; 1997)
offered explanations for some of these facts based on the concept of line tension. Line tension is
the energy associated with the line where the three phases meet and has dimensions of
energy/length, as opposed to energy/surface units for surface tension. It was concluded that
further discussion and research is necessary to better understand observed phenomena.
Tadmor (2004) used the line tension (energy) concept to express the experimentally
elusive equilibrium contact angle as a function of the advancing and receding contact angles.
However, calculations of the equilibrium value require precise measurement of both the
advancing and receding angles, something that is not practical in non-laboratory applications.

Figure 1.3 The effect of drop (bubble) size on advancing (Adv.) and receding (Rec.) contact
angles for the air /water/polyethylene film system as obtained with the static sessile-drop
(SD), static captive-bubble (CB), and dynamic captive-bubble (DCB) techniques. Source:
Reprinted from Journal of Colloid and Interface Science, 179 (1), Drelich, J., Miller, J.D.,
Good, R.J., The Effect of Drop (Bubble) Size on Advancing and Receding Contact Angles
for Heterogeneous and Rough Solid Surfaces as Observed with Sessile-Drop and CaptiveBubble Techniques,37-50, 1996, with permission from Elsevier.
Another interesting factor is the exposure time during which contact angle is measured.
For special cases, when the solvent alters the surface being characterized (e.g., when swelling
occurs) significant time drift of contact angles (both advancing and receding) can be observed
(Adamson and Gast, 1997). However, use of the captive bubble technique and a conditioned
membrane can eliminate the time drift problem. Measurement temperature is another factor that
may potentially affect the measured contact angle value. For example, Schonhorn (1966)
concluded based on theoretical considerations, that for polar liquids and polypropylene, the
dependence of contact angle on temperature should be minimal ( = 2 for T 100C). He
further concluded that contact angle should decrease as the temperature of the probe liquid
increases. Such a dependence of contact angle on temperature is likely attributed to the

4
2012 Water Research Foundation. ALL RIGHTS RESERVED.

dependence of liquid viscosity on temperature. Surface titration can provide information on the
dependence of contact angle on the chemical interactions of a solution with a solid surface
(Adamson and Gast, 1997). Mnttri et al. (2006) report contact angle changes as function of pH
for several commercial membranes. In the same paper, zeta potentials as functions of pH are also
shown. It is interesting to note that these plots of contact angle as a function of pH correlated
with zeta potentials plotted as absolute values. See also Brant et al. (2006) for further examples
of contact angle titration. Even though performing surface titration bears prohibitive labor cost
for routine analysis (multiple samples and solutions to assure statistically significant data), it
underscores the need for control of pH of the water used in the study.
Boussu et al. (2008) used multiple linear regression (MLR) to study the influence of
surface hydrophobicity, surface roughness, surface charge, molecular weight cut-off,
permeability, and porosity of the top layer on nanofiltration membrane performance. Contact
angle, volume fraction of small pores, and membrane charge were the most significant variables
in predicting relative flux and adsorbed amount as well as retention of dissolved organic
compounds by five commercial NF membranes. Also, contact angle measurements are useful in
membrane integrity testing (if one wishes to use a realistic estimate, rather than a worst case
scenario value of 1 for cos ) as outlined by EPA in Membrane Filtration Guidance Manual
(Alklaibi and Lior, 2005). The need for determination of the pressure necessary for integrity
testing stems from requirements of EPAs Long Term 2 Enhanced Surface Water Treatment Rule
(LT2ESWTR) (Allcock et al., 2006). For example, Cryptosporidium oocysts constitute a public
health risk as they cause gastro-intestinal illness and resist chlorine-based disinfectants. The size
of Cryptosporidium oocysts (spores) is approximately 3 m. The net pressure applied during
direct integrity tests must ensure that any breach large enough to pass Cryptosporidium oocysts
(3 m) would also pass air during the test (EPA, 2005). The minimum direct integrity test
pressure, otherwise known as the pore liquid entry pressure, is a function of both the liquid
characteristics (surface tension) and the membrane characteristics (hydrophobicity and pore
structure):
Ptest = 0.193 cos + BPmax

(1.1)

where Ptest is the minimum direct integrity test pressure (psi), is the pore shape correction
factor (dimensionless), is the surface tension at the air-liquid interface (dynes/cm), is the
liquid-membrane contact angle (degrees), BPmax is the maximum backpressure on the system
during the test (psi), and 0.193 is the constant that includes the defect diameter (i.e., 3-m
resolution requirement) and unit conversion factors (EPA, 2005). This equation appears as
Equation 4.1 in the Membrane Filtration Guidance Manual (EPA, 2005):
The LT2ESWTR does not establish the minimum test pressure to be used during a
pressure-based direct integrity test, but rather only requires that the test achieve a 3m resolution. If a membrane manufacturer has information to support the use of
values other than = 1 and = 0, and these less conservative values are approved
by the State, then Equation 4.1 can be used to calculate the minimum required test
pressure. It is essential that the use of values other than = 1 and = 0 be
scientifically defensible, since the use of inappropriate values could result in the use
of a test pressure that does not meet the resolution criterion established by the rule.

5
2012 Water Research Foundation. ALL RIGHTS RESERVED.

One approach for determining membrane-specific values for and is through


direct experimental evaluation. Because these parameters can have a significant
effect on the required direct integrity test pressure, it is strongly recommended that
States require sufficient justification from a membrane manufacturer prior to
approving the use of values other than = 1 and = 0, such as independent third
party testing results using a method accepted by the scientific community and
demonstrating statistically significant data.
It is clear that in order to reduce the pressure difference necessary for integrity testing and
to achieve the resolution requirement, reliable estimation of membrane contact angle is required.
The linear dependence of pressure drop on cos (Eq. 1.1) implies that for = 30, a pressure
drop 13% less than the worst case scenario estimate ( = 0) can be used. An uncertainty of 1 for
= 30 (i.e., = 30 1) will translate into 1% error bars for the required pressure drop. The
same absolute error bar for will translate into greater absolute and relative cos uncertainty for
more hydrophobic membranes (increasing in degrees: |d cos | = (/180)|sin||d |).
ZETA POTENTIAL MEASUREMENTS
Membrane surface charge is not directly experimentally accessible. Instead, zeta potential
is used as an approximation for both the magnitude and sign (+ or ) of membrane surface
charge. Zeta potential () is the potential difference between the bulk of solution and the sheer
(slipping) plane of the interfacial double layer [Figure 1.4] (Delgado et al., 2005). Zeta potential
is a function of surface and solution chemistry (pH, ionic composition, and ionic strength) at the
solid-liquid interface and is an important membrane characteristic for assessing membrane
fouling potential and developing chemical cleaning protocols. Membrane zeta potential is
typically determined from streaming potential measurements. A streaming potential is generated
when an electrolyte solution flows through a thin channel or porous media (e.g., a sand column)
and is related to zeta potential by the Helmholtz-Smoluchowski equation:

E 0
=
p

(1.2)

where E is the streaming potential due to electrolyte flow through a capillary channel, p is the
applied pressure driving the flow, is the zeta potential, is the electrolyte conductivity, is the
viscosity of the electrolyte solution, is the relative permittivity of the solution (dimensionless),
and 0 is the vacuum permittivity (fundamental constant). Values of E, p, and are measured by
the streaming potential analyzer; while and are calculated based on temperature measurement
(empirical fit functions for pure water data are used).
In experimental investigations of membrane charge, streaming potential measurements
have typically been used to calculate zeta potential. Streaming potential is the potential induced
when an electrolyte solution is pumped across a stationary, charged surface. Streaming potential
can be used to calculate zeta potential using the Helmholtz-Smoluchowski equation, which
relates the pressure dependence of streaming potential to the properties of the solution (i.e.,
conductivity and viscosity). Unlike earlier streaming potential investigations, the majority of
more recent investigations (e.g., (Nystrm et al., 1995; Childress and Elimelech, 1996;

6
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Hagmeyer and Gimbel, 1999; Peeters et al., 1999; Schfer, 1999; Childress and Elimelech, 2000;
Ernst et al., 2000; Tay et al., 2002)) measure streaming potential over a range of solution pH.
Investigating surface charge as a function of pH is crucial for understanding acid-base properties
of membrane surface functional groups.
Although streaming potential measurements are the most frequently used method for
evaluating charge properties, they have also been criticized. Results from prior studies reveal
uncertainty in individual measurements as well as data scatter (Chapman-Wilbert et al., 1999).
Further, the question of overall reproducibility has not been addressed. Other concerns include
the effect of membrane roughness on the measurement, vagueness of the relationship between
the measured zeta potential and the double layer structure, the inherent assumptions of the
Helmholtz-Smoluchowski equation (i.e., laminar flow), and the lack of a calibration standard.

Figure 1.4 Illustration of the relative locations of the Stern layer and Shear plane from a
charged surface in water and the associated change in surface potential with distance.
Source: Reprinted from Journal of Membrane Science, 161 (1-2), Chapman-Wilbert, M.,
Delagah, S., Pellegrino, J., Variance of streaming potential measurements, 247-261, 1999,
with permission from Elsevier.
Although the operating principles and requirements for measuring streaming potential
(e.g., voltage, pressure) are relatively straightforward, commercial instruments are much more
common than lab-made instruments. However, due to the small market for these instruments,
they are relatively expensive. This is an additional disadvantage of the technique. However, zeta
potential from streaming potential measurement is one of few techniques capable of describing
the charge properties of membranes.

7
2012 Water Research Foundation. ALL RIGHTS RESERVED.

SURFACE ROUGHNESS ATOMIC FORCE MICROSCOPY (AFM)


Surface roughness may be measured with atomic-level resolution using atomic force
microscopy (AFM). AFM is a non-optical surface imaging technique which approaches atomic
resolution. In the most common implementation, a sharp tip on a cantilever is scanned at
constant force across a surface to provide a surface profile. The force between the tip and the
surface always has a van der Waals (dispersion) force component, but cantilevers carrying
charge are also available and can be used to image electric potential. A laser beam is reflected off
the back of a cantilever and its position determined using a photo-diode. A surface may be
scanned using any number of operating modes (e.g., contact, tapping, non-contact). Tapping
mode is most commonly used to characterize membrane surfaces as contact mode may result in
damage to the membrane surface. Tapping mode uses a rapidly oscillating cantilever in the
vicinity of the surface, and amplitude damping is used for imaging. Only short, intermittent
contact of the AFM tip with the sample (tapping) occurs, which is especially suitable for
membrane surfaces. Measuring the position and movement of the cantilever as it is scanned over
a membrane surface allows for the direct measurement of surface features.
The membrane surface can be scanned in vacuum, air, or water and no sample
preparation is necessary (Mulder, 1996). An example illustration of an AFM experimental setup
is given in Figure 1.5. Based on the sample profile, h(x, y), it is possible to calculate mean and
RMS roughness (Ra and Rrms, respectively) and 10-point mean roughness (Rz) parameters (Chung
et al., 2002). Chung et al. (2002) determined the relationship between Ra and the water flux and
separation performance of hollow-fiber UF membranes. As Ra decreased, flux decreased and
rejection increased. This example illustrates the relevance of AFM characterization to membrane
performance.

Figure 1.5 Schematic illustration of the operating principles associated with the atomic
force microscope (AFM). Taken from Wyart et al. (2008). Source: Reprinted from Journal
of Membrane Science, 315 (1-2), Wyart, Y., Georges, G., Demie, C., Amra, C., Moulin, P.,
Membrane characterization by microscopic methods: Multiscale structure, 82-92, 2008,
with permission from Elsevier.

8
2012 Water Research Foundation. ALL RIGHTS RESERVED.

TRANSMISSION ELECTRON MICROSCOPY (TEM)


In transmission electron microscopy (TEM) an image is produced by passing electrons
under high vacuum through a sample. Near-atomic resolution is attainable using TEM; however
this technique is limited to thin samples that are at most several tens of nanometers in thickness.
Therefore, TEM is restricted to imaging thin-film samples rather than fully constructed
membranes. An example TEM image of a graft copolymer is shown in Figure 1.6. Kim et al.
(2008) concluded that the differences in electron densities between the two components of this
amphiphilic graft copolymer were sufficient to be observed as image contrast. Dark regions
represent the hydrophobic crystalline regions of the main chains, whereas lighter regions
represent the hydrophilic side chains. In another example, use of sophisticated cutting tools (e.g.,
a diamond knife ultramicrotome) enabled Blom et al. (2003) to study cross-sectional samples of
a proton exchange membrane.

Figure 1.6 Transmission electron microscope (TEM) image of a graft copolymer film.
Taken from Kim et al. (2008). Source: Reprinted from Journal of Membrane Science,
325 (1), Kim, Y., Park, J.T., Koh, J.H., Roh, D.K., Kim, J.H., Anhydrous proton
conducting membranes based on crosslinked graft copolymer electrolytes, 319-325,
2008, with permission from Elsevier.
More recently, tomography capability has been added to the TEM technique (TEMT).
The majority of publications using TEMT pertain to polymers. In TEMT, TEM images are
collected from single specimens at different tilt angles and are processed to generate 3D
representations of nano-structures. In a paper by Jinnai et al. (2006), 200-nm thick co-polymer
samples were studied. Images clearly show the porous structure of the material (Figure 1.7). No
reports have been found pertaining to TEMT for polymeric membranes for water treatment. In
terms of sample preparation, the vacuum environment requires dry samples. Also, coating may

9
2012 Water Research Foundation. ALL RIGHTS RESERVED.

be necessary to prevent sample destruction by the electron beam. These requirements constitute
disadvantages of TEM as they can lead to artifacts.

Figure 1.7 Ordered and disordered co-polymer blend morphologies visualized using TEMT
method. Taken from Jinnai et al. (2006).
SCANNING ELECTRON MICROSCOPY (SEM)
For scanning electron microscopy (SEM) measurements a sample surface is exposed to a
narrow beam of electrons in vacuum. Secondary electrons are liberated form the surface and
detected. These conditions require sample preparation drying and coating with conductive
surface (in most cases). Sample preparation can lead to artifacts in observed images (Mulder,
1996). Resolution of 10 nm or better can be achieved, but varies with equipment available. Pore
size, surface porosity, and pore geometry can be visualized (Mulder, 1996) and porosity and pore
size distribution can be estimated. Asymmetry in composite membranes can also be readily
observed. Li et al. (2008) used SEM to study polysulfone/sulfonated PEEK membranes. Changes
in pore structure resulting from different casting conditions and composition are shown in Figure
1.8.

10
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 1.8 Scanning electron microscope (SEM) images of cross-sections of polymer


blend membranes. Taken form Li et al. (2008). Source: Reprinted from Journal of
Membrane Science, 324, Li, X., De Feyter, S., Vankelecom, I.F.J.,
Poly(sulfone)/sulfonated poly(ether ether ketone) blend membranes: Morphology
study and application in the filtration of alcohol based feeds, 67-75, 2008, with
permission from Elsevier.
Obtaining cross-section samples usually involves razor blades and can result in
compression and tearing. The direct freeze fracture method, where a membrane is made brittle in
liquid nitrogen and is then broken, can also be used. Co-principal investigator Phipps was
involved in developing a new cryo-snap technique, where the membrane is embedded in ice
before a fracture is made. Perhaps the resolving power of SEM measurements is the greatest
advantage of this technique not reaching atomic level, but well beyond optical microscope
resolution, it is perfectly suited for the study of membrane pores. Very rough samples that can
cause a large amount of tip artifacts in AFM (Mller and Mehnert, 1997) are not a problem in
SEM. Also, the preparation of a thin film, as is required in TEM, is not required for SEM. The

11
2012 Water Research Foundation. ALL RIGHTS RESERVED.

only disadvantage is that membrane pore structure can be affected by drying and coating, so
SEM-observed porosity may not exactly correspond to in situ membrane porosity.
CHEMICAL FORCE MICROSCOPY (CFM)
Chemical force microscopy (CFM) is an innovative extension of AFM. CFM uses a
chemically modified AFM tip to investigate specific interactions with, and ultimately the
chemistry of, a membrane surface. Often a microsphere, or a colloidal particle, with specific
functional groups is placed on the tip to create a functionalized probe. For example, a tip covered
with acidic carboxylic groups will exhibit enhanced attraction to polar, rather than hydrophobic,
sites on the surface. CFM experiments on membranes are frequently performed in aqueous
solutions in order to best simulate in situ interactions. There is no need for surface preparation as
in SEM. The distribution of adhesion values that were measured across the HL and SG
membrane surfaces using various (hydrophobic and hydrophilic) probes are reported in Table
1.1. This data may be used to characterize the relative chemical heterogeneity across a membrane
surface (i.e., the distribution of hydrophilic [low adhesion] and hydrophobic [high adhesion]
areas). This data may be further supported using CFM resolved images of the membrane surface
(Figure 1.9), which are similar to the 3D rendered images of membrane surfaces that are
produced using conventional AFM. For example, Brant et al. (2006) used CFM images of two
membrane surfaces in order to understand their associated fouling tendencies.
Table 1.1
Distribution of adhesion across the HL and SG membrane surfaces as a function of surface
area coverage measured with methyl, carboxyl and hydroxyl functionalized probes (I =
0.01 M NaCl; pH 6; T = 20 C).
Probe
High Adhesion
Moderate Adhesion
Low Adhesion
Surface area coverage (%)
HL Membrane
Methyl

5.2

88.0

6.8

Carboxyl

97.2

1.6

1.2

Hydroxyl

7.8

88.9

3.3

Methyl

5.1

22.2

72.7

Carboxyl

93.0

6.7

0.4

Hydroxyl

0.3

21.7

78.0

SG Membrane

Source: Reprinted from Colloids and Surfaces A:- Physicochemical and Engineering Aspects, 280 (1-3),
Brant, J.A., Johnson, K.M., Childress, A.E., Characterizing NF and RO membrane surface heterogeneity
using chemical force microscopy,45-57, 2006, with permission from Elsevier.

12
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 1.9 Topographical and corresponding CFM images for the HL membrane acquired
with a CH3-, COOH-, and OH-functionalized tip (I = 0.01 M NaCl; pH = 6.1, and T = 20
C). Source: Reprinted from Colloids and Surfaces A:- Physicochemical and Engineering
Aspects, 280 (1-3), Brant, J.A., Johnson, K.M., Childress, A.E., Characterizing NF and RO
membrane surface heterogeneity using chemical force microscopy,45-57, 2006, with
permission from Elsevier.
Yamamura et al. (2008) performed CFM experiments on water treatment membranes
using hydroxyl- and carboxyl-modified microspheres. The hydroxyl-coated microspheres
exhibited greater affinity for the membrane surface than the carboxyl-modified microspheres.
Exponential decay of the affinity was observed with time as a result of membrane fouling
(Figure 1.10).

13
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 1.10 Changes in adhesion force of carboxyl-modified microspheres (a) and


hydroxyl-modified microspheres (b) to PE membranes that were sampled during the
pilot-filtration test. An SEM image of a functionalized polystyrene bead glued to a
cantilever tip used in this study is shown on the right. Taken from Yamammura et
al. (2008).
Alsteens et al. (2007) used CFM with hydrophobic tips to measure forces on organic selfassembled monolayers and on bacteria surfaces; contact-mode images and force-distance curves
were reported. A linear relationship was observed between the adhesion force and cos , where
is the macroscopic surface-water contact angle. Li et al. (Li and Elimelech, 2004) used carboxylmodified colloid probes to study NF-270 membranes. Correlations between the measured
adhesion forces and the fouling and cleaning behaviors of the membrane were observed. The
authors used CFM measurements to investigate the role of calcium ions in natural organic matter
fouling and concluded that Ca2+ ions enhance fouling through complexation.
X-RAY PHOTOELECTRON SPECTROSCOPY (XPS)
In X-ray photoelectron spectroscopy (XPS), the sample surface is subjected to X-ray
radiation capable of removing electrons from the inner shells of the atoms (except H and He).
The amount of emitted electrons is recorded as a function of binding energy. Surface depths of
0.5 -10 nm can be probed, depending on incident beam angle (Mulder, 1996). The inner shell
energies are characteristic for a given element, but somewhat sensitive to the external chemical
environment. For carbon, this chemical shift can amount to a few eV: for the carbonyl group
(C=O), the binding energy is 285 eV and for the methylene group (CH2), this binding energy is
14
2012 Water Research Foundation. ALL RIGHTS RESERVED.

287.8 eV (Mulder, 1996). Routine detection limits are 0.1%. From the emitted electron
intensities surface atomic concentrations (empirical formulas) can be deduced. XPS results for a
variety of membranes are reported in Table 1.2. Notably, all three polyether sulfone membranes
(NTR7450, N30F, and NFPES10) showed a prominent percentage of nitrogen, in disagreement
with the composition of the bulk material. Boussu et al. (2007) assigned this discrepancy to the
presence of an additive, polyvinylpyrroli (or solvent dimethylformamide).
Table 1.2
Measured atomic concentration percentages (%) of C1s, O1s, N1s and S2p obtained by X-ray
photoelectron spectroscopy (XPS) for commercially available and experimental
membranes. Taken from Boussu et al. (2007).
NF270 Desal51HL Desal5DL NTR7450 N30F NFPES10 D71
N13
C1s

72.0

75.9

72.4

72.1

79.9

77.3

77.0

79.7

O1s

17.0

12.6

17.6

19.2

14.4

16.8

16.3

14.4

N1s

10.9

10.7

9.5

5.2

1.8

2.4

2.6

2.3

S2p

0.1

0.8

0.5

3.5

3.9

3.5

4.1

3.6

The greatest advantage of XPS is the shallow probing depth that is required (i.e., it
obtains information about the near membrane surface rather than the membrane interior) and its
ability to characterize the elemental composition of the membrane surface. However, chemical
identity of the polymer (or additives) cannot be deduced based on elemental composition alone
[even if supported by atom type assignments (Boussu et al., 2007)].
ATTENUATED TOTAL REFLECTION FOURIER TRANSFORM INFRARED
SPECTROSCOPY (ATR-FTIR)
Attenuated total reflection fourier transform infrared spectroscopy (ATR-FTIR) allows
for infrared (IR) analysis of surfaces. The IR spectrum can provide determination of vibrational
frequencies and transition intensities of most molecules (with the exception of diatomics such as
N2 and O2), including characteristic functional group frequencies. Knowledge of vibrational
frequencies of functional groups (or reference spectra) allows for chemical identification of at
least a class of compounds (e.g., aromatic amides), which may be less apparent in XPS spectra.
Example images of ATR-FTIR spectra for a variety of membranes are given in Figure 1.11. The
IR radiation typically penetrates 1 m into the surface. This is a disadvantage because this is
deeper than the active layer thickness of most composite membranes, so the top layer is not
necessarily isolated (Boussu et al., 2007). The penetration depth can be decreased by careful
selection of crystal and incident angle (for Ge ATR element and 3000 cm-1 wavenumber it can be
estimated as 200 nm (Mller and Mehnert, 1997) at 45 incident angle). Alternatively, it is
possible to perform ATR-FTIR measurements of both sides of an asymmetric membrane and
then to essentially subtract out the support layer spectral regions.

15
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 1.11 ATR-FTIR spectra of several commercial and one lab-made membrane,
containing sulfone groups. Ar-SO2-Ar typical frequencies are (S=O asym) = 1325 cm-1,
(S=O symm) = 1140 cm-1. Differences in IR spectra stem from C(CH3)2 bridges present
between aryl rings in NTR7450 and Desal 5DL. Taken from Boussu et al. (2007).
Operation of ATR-FTIR is based on the total internal reflection phenomenon. A block of
material with greater optical density than the sample is put into close contact with the sample
surface. The incident radiation angle must be greater than the critical angle value. Evanescent
standing waves penetrating the sample surface are present, though no refracted ray appears due
to total internal reflection conditions. The evanescent field decays exponentially into the sample
and constitutes specifics of the method. The IR information about the surface is weighted by the
exponential decay function the layers closest to the surface have the strongest representation in
the acquired spectrum (Mller and Mehnert, 1997).
Interest in ATR-FTIR for membrane characterization has recently increased, but this
technique seems to be underutilized as compared to XPS. Access to the surface vibrational
frequencies, as opposed to bulk material, is an advantage of ATR-FTIR, albeit diminished by the
somewhat large penetration depth. Vibrational frequencies cannot be easily converted into
complete chemical structure of the material, but can provide more molecular structure
information than XPS.
OBJECTIVES
The overall objective of this project was to use a comprehensive approach for
standardizing select membrane characterization methods and to quantitatively evaluate the
combined effects of select membrane properties on membrane performance. The specific goals
of this research were to:

16
2012 Water Research Foundation. ALL RIGHTS RESERVED.

1. Identify and evaluate various available methods for characterizing the physical and
chemical characteristics of membrane surfaces,
2. Quantify the variability between data generated using select characterization techniques
in order to develop a series of recommended standard methods for characterizing
membrane surfaces,
3. Provide drinking water utilities with information on the value of membrane
characterization methods, a means to compare membrane properties on an equal basis,
and guidance for accurate contact angle measurements for integrity test calculations
4. Evaluate how surface characteristics determined using the proposed standard
characterization methods accurately predict membrane performance and fouling.
The focus of this investigation was on pressure-driven membrane processes that are used
for municipal drinking water treatment (i.e., MF, UF, NF, and RO). Results and insights from
this research are expected to benefit industry, academics, and consultants by facilitating and
providing guidance for: comparing published information and data on membrane contact angle,
surface roughness, and streaming potential; improving the precision of contact angle, surface
roughness, and streaming potential measurements; determining the relation of membrane
characteristics to membrane performance; interpreting membrane characteristic information
provided by manufacturers; and using contact angle measurements for integrity test calculations
as required to receive membrane treatment credits under the Long Term 2 Enhanced Surface
Water Treatment Rule (LT2ESWTR).

17
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 2
MATERIALS AND METHODS
MEMBRANE SAMPLES
Membranes from each of the four different membrane categories [microfiltration (MF),
ultrafiltration (UF), nanofiltration (NF) and reverse osmosis (RO)] were selected for evaluation
in this study (Table 2.1). Membrane selection was based on discussions with partner utilities, as
well as on input from other engineering professionals and PAC members. A list of utility
installations where the selected membranes are currently in use is included in Appendix A.
Table 2.1
Names and select properties of membranes used in this investigation (MF- microfiltration;
UF- ultrafiltration; NF- nanofiltration; RO- reverse osmosis, PP- polypropylene, PES
polyethersulfone, PA- polyamide, TFC thin film composite).
Membrane
Category

Selected Membrane

Justification for Selection

MF

Siemens Memcor CMF-S


hollow fiber
PP, outside-in

The CMF membrane series is a very common


membrane. It is used by Orange County Water
District (OCWD) and the West Basin Municipal
Water District, two of the research partners.

UF

Norit X-Flow
hollow fiber
PES, inside-out

Norit X-Flow is also very common. It is used by


Minneapolis Water Works, one of the partner
utilities.

NF

Dow Filmtec NF-270


spiral wound
PA, TFC
Hydranautics ESNA1-LF
spiral wound

RO

The NF-270 and ESNA1-LF membranes are fairly


commonly used in NF applications. The ESNA1LF membrane is used by the City of Hollywood,
FL.

Hydranautics ESPA-2
spiral wound
PA, TFC

The EPSA membrane is a brackish water


membrane used by OCWD.

Dow Filmtec SW30HR-380


spiral wound
PA, TFC

The SW30 and HR-380 membranes are used at the


Tampa Bay seawater desalination facility.

GE Water MUNI-RO-400
spiral wound

The MUNI-RO-400 membrane is used by the City


of Clearwater, FL, one of the research partners

18
2012 Water Research Foundation. ALL RIGHTS RESERVED.

STATISTICAL ANALYSIS
Two statistical measurements were used in evaluating the precision of the
characterization data collected in this investigation, the repeatability standard deviation (sr):

sr =

/p

(2.1)

and the reproducibility standard deviation (sR):

sR =
( sx ) 2 + ( sr ) 2 (n / (n 1))

(2.2)

where p is the number of separate laboratories from which data were collected, s is the cell
standard deviation (standard deviation for results from a single laboratory), sx is the standard
deviation of cell averages, and n is the number of measurements. These two measures of
precision are described in detail in ASTM E-691-09 ((ASTM), 2009).
MEMBRANE COMMUNITY SURVEY ON MEMBRANE CHARACTERIZATION
TECHNIQUES
A questionnaire was sent out to membrane manufacturers, utilities using membrane
processes, academics, and membrane scientists in order to gain their input and thoughts on the
standard membrane characterization methods developed as part of this project. A copy of the
questionnaire that was sent out to all participants may be found in Appendix B. Survey responses
were compiled and interpreted to determine and/or obtain the following information:
1. The usefulness of contact angle, surface roughness, and zeta potential data in
various membrane applications and fields of study.
2. The level of interest by the membrane community for conducting contact angle,
surface roughness, and zeta potential measurements.
3. Acquire feedback on, and suggestions for, improving the proposed standard
techniques for conducting contact angle, AFM surface roughness, and streaming
potential (zeta potential) measurements.
DEVELOPMENT OF STANDARD TECHNIQUES FOR CHARACTERIZING
MEMBRANE SURFACES
Two membranes (ESNA1-LF nanofiltration and SW30HR reverse osmosis) were
distributed to the original three partner laboratories: Colorado School of Mines (CSM), Duke
University (DU), and Orange County Water District (OCWD) and eventually, to the additional
partner laboratories, University of California, Riverside (UCR) and the University of Maribor
(FKKT). Each laboratory was asked to conduct specific characterization measurements (contact
angle, surface roughness using AFM, and streaming potential) using a set of guidelines provided
by UNR. Once the raw data were collected they were returned to UNR for statistical analysis and

19
2012 Water Research Foundation. ALL RIGHTS RESERVED.

interpretation. Based on this data analysis and input from the partner laboratories, standard
methods were developed for each of the characterization measurements (see Appendix C). Once
the membrane characterization standard techniques were developed for this subset of
membranes, the techniques were then applied to all of the acquired flat sheet membrane samples
in Table 2.1.
AFM Surface Roughness Measurements
Data collected by the AFM during imaging can be analyzed to calculate a variety of
surface roughness statistics and used to produce 3D rendered images of the membrane surface.
Those roughness statistics that are most commonly used to characterize membrane surfaces
include average roughness (Ra), root mean square (RMS) roughness (Rq), and surface area
difference (Khulbe et al., 2008). Each of these statistical parameters was used to interpret the
AFM generated data in this investigation. The RMS roughness, Rq, is a population standard
deviation for a set of z values (depth coordinate) collected for a given area, where zi are heights
(z-coordinates) from the AFM scan, N is number of data points (i = 1, 2, , N), and i is the data
index.

Rq =

(z

zavg ) 2
N

(2.3)

The average surface roughness, Ra, is calculated using averaged absolute values of
deviations from the average surface plane:

Ra =

| z

zavg |

(2.4)

Surface roughness data from the DU and OCWD laboratories on two membranes
(ESNA1-LF and SW380HR membranes) were acquired and used to estimate inter-laboratory
precision. In addition to the previously mentioned membranes, the OCWD laboratory also
analyzed three other flat sheet membranes (NF-270, ESPA, MUNIRO-400). Results from
OCWD for all five flat sheet membrane samples were used for developing correlations between
membrane surface properties and performance.
In addition to the membrane samples, standard surfaces (Table 2.2) were sent to each of
the participating laboratories for analysis. The standard surfaces were used for characterizing the
surface roughness of a material with a well-defined surface morphology. Flat, concave (to
represent the inside of a hollow fiber membrane) and convex (to represent the outside of a
hollow fiber membrane) standards were selected. While the flat standard surface (i.e., the z-axis
calibration standard) is not strictly a roughness standard, it does allow for calibrating the AFM
instrument in the z-direction. It was assumed that calibration in the x and y directions (i.e., the
sample plane) were not necessary. The z-axis standard was intended to help correct for
instrument drift in the direction that is most critical for calculating roughness (i.e., the zdirection).

20
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 2.2
Proposed standard surfaces for atomic force microscopy surface roughness measurements.
Measurement \
Sample

A
Flat

AFM Roughness

HS-20MG nanostructure array


(20 nm step height) for z-axis
calibration from
budgetsensors.com, or similar
grid.

B
Concave
Teflon tubing
~1 mm outer
diameter

C
Convex
Teflon tubing
~1 mm outer
diameter

Contact Angle Measurements


This study involved developing standard methods for measuring the contact angle with
water on both flat-sheet and hollow fiber membranes. To accomplish this it was necessary to
identify and evaluate candidate standard surfaces for three separate conditions: flat surfaces,
concave surfaces, and convex surfaces. These three surfaces correspond to a flat membrane, the
outside of a hollow fiber membrane, and the inside of a hollow fiber membrane, respectively.
Those surfaces that were considered for each of these conditions are summarized in Table 2.3.
A similar approach to that which was previously described for the AFM measurements
was followed for developing the standard technique for contact angle measurements. This
approach involved distributing standard surface and membrane samples (the Teflon standard
surface, ESNA1-LF membrane, and SW30HR membrane) to DU and CSM. Contact angle
measurements were also performed at the UNR laboratory. All laboratories were provided with
guidelines for conducting the contact angle measurements (see Appendix C). However,
variations existed in the type of goniometer used by each laboratory for measuring contact angle,
and other laboratory techniques, amongst the three laboratories. Both the DU and CSM
laboratories were only responsible for conducting contact angle measurements on flat-sheet
membrane samples. Therefore, they were not provided samples of the concave and convex
standard surfaces. Developing a standard method for measuring contact angles on hollow fiber
membranes was the sole responsibility of the UNR laboratory.
Table 2.3
Proposed standard surfaces for contact angle measurements.
Measurement \
Sample

A
Flat

Contact Angle

Teflon flat sheet from Ram-Hart


with contact angle range and
material origin certificate.

B
Concave
Cleaned glass
capillary
painted black
on the outside

PMMA poly(methyl methacrylate)

21
2012 Water Research Foundation. ALL RIGHTS RESERVED.

C
Convex
Glass capillary
covered with
PMMA on the
outside

Streaming Potential Measurements


An approach similar to that which was previously described for the surface roughness
and contact angle measurements was followed for developing the standard technique for
streaming potential measurements in order to calculate zeta potential. This approach involved
distributing three different sample surfaces to the DU and CSM laboratories and having them
measure streaming potential as a function of solution pH and ionic strength. The three surfaces
that were distributed included a PMMA standard surface, the ESNA1-LF membrane, and the
SW30HR membrane. Streaming potential measurements were also performed at the UNR
laboratory. All laboratories were provided with guidelines for conducting the streaming potential
measurements (see Appendix C). However, variations existed in the type of electrokinetic
analyzer used by each laboratory for measuring streaming potential, and other laboratory
techniques, amongst the three laboratories.
After a first round of testing, and interpretation of the subsequent streaming potential
results, it was determined that a number of equipment and procedural errors resulted in erroneous
results. Following discussions with the partner laboratories and representatives from Anton Paar
(manufacturer of the SurPASS electrokinetic analyzer) it was determined that the streaming
potential measurements should be repeated. During this second round of testing the membrane
samples were distributed to and analyzed by the following parties: Anton Paar, the University of
Maribor (FKKT), the University of California, Riverside (UCR), and UNR. Only UNR and
Anton Paar provided results for the PMMA standard surface. Both DU and CSM did not conduct
measurements for the second round of streaming potential measurements. For this round robin of
tests the following streaming potential analyzers were used:

University of California, Riverside (UCR) Electro Kinetic Analyzer (EKA) from Anton
Parr (Ashland, VA)

University of Nevada, Reno (UNR) ZetaCAD from Colloidal Dynamics (Ponte Vedra
Beach, FL)

University of Maribor (FKKT) and Anton Paar SurPASS Electrokinetic Analyzer from
Anton Parr (Ashland, VA)

ASSESSMENT OF MEMBRANE PERFORMANCE AND FOULING


Membrane characteristics are often used to estimate or predict membrane performance,
and specifically, the occurrence or severity of membrane fouling. A standard bench-scale
membrane test unit was used to measure the performance of the membranes previously
characterized in this study. A process flow diagram for the standard membrane test unit is given
in Figure 2.1. Process components and their associated properties are summarized in

22
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 2.4. Originally, membrane fouling experiments were planned to be performed at both UNR
and at a Clearwater, FL desalination facility. However, testing using the actual RO feed from the
Clearwater site resulted in no significant permeate flux decline after 5 days of testing. This result
was independent of the membrane being evaluated. Therefore, testing using the RO feed from
the Clearwater plant was discontinued.

Figure 2.1. Process flow diagram of the bench-scale membrane test system used in the
correlation experiments for membrane surface properties and perfromance (P pressure
gauge, T temperature probe, F flow meter).

23
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 2.4
Summary of primary system components and accessories for the standard membrane test
unit.
Item
Quantity
Notes

120 Volt Water Pumps

Membrane Test Cells

Tubing

Test Cell Inlet Assembly

Test Cell Outlet


Assembly

One pump supplies feedwater to two test cells (flow


is split between two test cells)
Fluid-O-Tech Model 311, Stainless steel vane
pumps, 101 GPH at 200 psi, 120 VAC. Pumps are
equipped with on/off switches on power cords.

Stainless steel cells that house flat-sheet membrane


samples operating in tangential flow configuration
Each cell consists of two plates, a Teflon gasket, 10
bolts, and 10 washers
Each cell is labeled using a number from 7 to 10
Feed channel dimensions (determined by Teflon
washer): 146 mm 95 mm 0.8 mm.
Membrane active area (determined by stainless
steel frit area): 127 mm 81 mm

17 sections +
1 spare
section

Feed water pressure tubing (reservoir to pump,


pump to cell, flow meter to waste): Polyethylene
3/8 outer diameter, total length ~ 36 ft + spare.
Fittings: stainless steel Swagelok for cell inlet
(shipped on tubing), custom flow meter black
plastic fittings (present on tubing), quick connect on
pumps.
Permeate recycling: polyethylene (1/4), total
length ~20 ft, plastic compression fittings (shipped
on tubing)
Permeate collection: Black rubber tubing (4x~1 ft)
inserted on stainless steel adapter of each valve.
Spare tubing (3/8) is also supplied in the event
that it is required

Inlet assembly consists of a pressure gauge and a


needle valve
Fittings included in the inlet assembly are 3/8
stainless steel compression (Swagelok)

Outlet assembly consists of a needle valve and a


flowmeter
Fittings included in the outlet assembly are 3/8
stainless steel Swagelok compression fittings (cellassembly) and plastic custom compression fitting

24
2012 Water Research Foundation. ALL RIGHTS RESERVED.

for flow meter 3/8tubing (shipped on tubing)

Thermocouple

To be installed in each cell inlet channel brass


3/8 NPT fitting, thermocouple installed inside
needle with epoxy resin connecting to male brass
fitting

Digital Thermometer

Calibrated to be used with thermocouples supplied

One template for cutting the membrane samples to


the correct dimensions for the test cells
One template for cutting the membrane feed spacer
mesh to the correct dimensions for the test cells

Templates

The chemistry and composition of the feed solution used in all membrane fouling
experiments is summarized in Table 2.5.
Table 2.5
Chemistry and composition of the feed solution used for membrane fouling experiments.
Concentration
Parameter
mg/L
Sodium Silicate (Na2SiO39H2O)
118.3
Calcium chloride (CaCl22H2O)

190.9

Potassium chloride (KCl)

142.8

Sodium carbonate (Na2CO3)

140.0

Sodium sulfate (Na2SO4)

239.4

Starch1

Hydrochloric acid (HCl)


Bentonite clay
200
1 Starch was added in order to adjust the total organic carbon (TOC) content of the feed
solution.
Temperature Correction Factors for Normalizing Membrane Permeate Flux
Temperature correction factors (TCFs) are needed to compare permeate flux values
measured at different temperatures. For example, a temperature increase of 1 C will result in a
permeate flux increase of approximately 3% (with all other factors, especially pressure,
remaining the same). In order to standardize the permeate flux values acquired as part of these
experiments the measured permeate fluxes were corrected and reported at a temperature of 25C
using the following expression:
(2.5)

25
2012 Water Research Foundation. ALL RIGHTS RESERVED.

where JT is the permeate flux measured at some temperature (T), TCF is the temperature
correction factor, and J25 is the permeate flux at a temperature of 25 C. Using an Arrhenius-type
functional dependence for water viscosity on temperature, one could expect the following
dependence for the inverse of the pure water permeate flux at a constant feed pressure:
(2.6)
where T is the temperature (C), E is the viscosity activation energy, and R is the universal gas
constant. Letting U = E/R, then Eq. 7 may be linearized, if the inverse of T is taken as an
independent variable:
(2.7)
The values of lnA (and A) and U can be determined using linear regression for ln(q) vs
the inverse of absolute temperature. The resulting TCF should have the following form:
(2.8)
The regression parameters (A, U) were determined by measuring the pure water permeate
flux as a function of feed pressure (3 to 4 runs, each consisting of 4-5 data points) for each
membrane. The regression parameters were then averaged and used to calculate the TCFs for
each membrane (Table 2.5).

26
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 2.6
Summarized temperature correction factors (TCFs) for the NF-270, ESNA1-LF, ESPA-2,
SW30HR, and MUNIRO-400 membranes, which were used in the membrane fouling
experiments.
T [OC]

NF270

ESNA1-LF

ESPA-2

SW30HR

MUNIRO-400

15

1.459

1.384

1.417

1.512

1.408

Dow
Brochure1
1.422

16

1.403

1.339

1.367

1.449

1.359

1.371

17

1.350

1.295

1.319

1.389

1.312

1.323

18

1.299

1.253

1.273

1.332

1.267

1.276

19

1.251

1.212

1.229

1.277

1.224

1.232

20

1.204

1.173

1.187

1.225

1.183

1.189

21

1.160

1.136

1.146

1.176

1.143

1.148

22

1.117

1.100

1.108

1.129

1.105

1.109

23

1.076

1.065

1.070

1.084

1.069

1.071

24

1.037

1.032

1.034

1.041

1.034

1.035

25

1.000

1.000

1.000

1.000

1.000

1.000

26

0.964

0.969

0.967

0.961

0.968

0.971

27

0.930

0.939

0.935

0.924

0.936

0.943

28

0.897

0.911

0.905

0.888

0.906

0.915

29

0.866

0.883

0.875

0.854

0.878

0.889

29.9

0.839

0.859

0.850

0.825

0.853

0.866

1.86E-06

6.43E-06

7.01E-06

9.74E-06

1.03E-05

3246

2794

2996

3552

2938

0.020

0.017

0.0066

0.044

0.0083

560

63

124

346

171

A
[min/mL]
U
[K]
RMSE2
SU
[K]
1

- Dow, Form No. 609-02129-804, Tech Manual Excerpt, FILMTEC Membranes, Addendum:
Temperature Correction Factor.
- RMSE: root mean square error with respect to Dow data: RMS=[(1/N)
Sum(Dow value

predicted)2]1/2

A value for U is needed to determine the TCFA, and is called the membrane specific
manufacturer-supplied constant (EPA Membrane Filtration Guidance Manual, EPA 815-R-06009, 2005). A best fit to the data that was supplied by Dow Filmtec gave a value for U of 2,982
K. The experimental standard deviation, SU, is the difference between the Dow Filmtec value (U
= 2,982 K) and that obtained experimentally for any given membrane. Establishing more precise
U values for each membrane would require more data points. Nevertheless the experimental
values of U for TCF can be used:
(2.9)

27
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Alternatively, substituting the universal expression with UM = 2,982 K, which is a convenient


formula to substitute for Dow tabulated data, would be also acceptable. In this study, the
universal correction was applied.
Membrane Performance Testing
Membranes were installed with mesh feed spacers, as some membranes were susceptible
to damage without spacers. After compacting membranes with DI water overnight at 690 kPa,
tests were run using a bentonite suspension as the feed source (see Table 2.5). Permeate flow was
monitored at 690 kPa for a short period of time (several minutes), cell pressure was then adjusted
to maintain an initial (starting) permeate flow rate of 4 mL min-1. Once the feed solution was
switched to the bentonite suspension the system was operated in a constant pressure, variable
flux configuration. The feed flow rate was maintained at 379 mL min-1. Temperature was
recorded at the beginning and at the end of permeate flow measurement. The average water
temperature was used for calculating the TCF for each system. Water temperature was
maintained between 22 and 25C. The SW30HR membrane was an exception to the 4 mL min-1
target flow, as a sufficient feed pressure could not be attained due to low permeability of this
membrane. The normalized permeate flow as function of time for each membrane evaluated here
is reported in Figure 2.2.

28
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 2.2. Normalized permeate flow (25C, flow at t = 0 set to 1) for membranes tested as
a function of time (hours).
Overview of Membrane Performance Test Method
The experimental setup described above was chosen for its simplicity; however, it is important to
note that since the membranes were operated at constant pressure as opposed to constant water
flux, and all had differing initial permeability, the mass transport of foulant to each membrane
differed significantly as a function of time, making it impossible to use the relationship between
observed water flux and run time as a means of assessing membrane performance.
Normalizing Membrane Performance Response
In order to compare the influence of processing a constant mass of foulant at each
membranes water flux, the mass of clay that was carried to the membrane surface was first
determined based on the permeate passing the membrane. Regression analysis using Statgraphics
Centurion XV (Statpoint Inc., Herndon, VA, USA) software was used for each membrane to
establish the relationship between the total amount of permeate volume recovered and the
resulting membrane performance (as normalized water flux, J/Jo). Permeate volume was then
converted to bentonite clay loaded per square meter of membrane surface based on the
concentration of bentonite clay in experimental suspensions (200 mg L-1) and the active area of

29
2012 Water Research Foundation. ALL RIGHTS RESERVED.

membrane swatches (0.0103 m2). From these models, normalized membrane fluxes at exact
bentonite clay loads of 2.5, 5, 10, and 20 g m-2 were determined for each membrane. Plotting
these results for each membrane revealed that initially normalized water flux decreased with clay
load in a more or less linear fashion, but that as clay continued to be loaded, water flux seemed
to reach a plateau. A bentonite mass loading value used to normalize membrane performance
data was chosen from the region of the curves where 1) the relationship between normalized
water flux and bentonite loading appeared most linear and 2) was within the range of measured
experimental data for all five membrane experiments.
Once this was accomplished, the intrinsic water flux predicted by the simple regression
models at that chosen constant clay load was used as the dependent variable in a multiple linear
regression (MLR) study relating J/Jo to membrane contact angle, zeta potential, roughness, and
initial flux. Each membrane type provided an exemplar to construct this MLR model.

30
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 3
RESULTS AND DISCUSSION
METHOD DEVELOPMENT FOR AFM SURFACE ROUGHNESS MEASUREMENTS
Imaging and Roughness Analysis of Flat Membrane Surfaces
ESNA1-LF Membrane
Surface roughness statistics for the ESNA1-LF membrane are summarized in Table 3.1.
Representative 3-D rendered AFM images of the ESNA1-LF membrane surface are shown in
Figure 3.1. Relative standard deviations for repeated measurements at different locations on the
membrane are approximately 15% for the OCWD results, and 30% for the DU results. Results
overlap within 95% confidence limits, but only as a result of the large spread in the DU data set.
There is a stark difference in the surface roughness statistics Rq and Ra that were generated for
the same membrane by the two laboratories. Considering both parameters, the data from OCWD
suggests a smoother membrane surface than that from DU. The source for this disagreement in
the data is unclear. Both laboratories analyzed membrane coupons from the same sample and
used the same measurement mode for the AFM (i.e., tapping mode). A noticeable trend in the
DU data is that both surface roughness statistics varied considerably from one sample site to the
next (Rq ~ 15 to 42 nm). In contrast, the variability from one site to another for the OCWD data
was less substantial (Rq ~ 10 nm). The small data set precludes the development of hard
conclusions for this single membrane; however, there are several possibilities for explaining the
observed large variation in the DU surface roughness data, including: i) the sites selected for
analysis by DU were more heterogeneous than those selected for analysis by OCWD, ii) some
unknown experimental protocol used by DU resulted in greater variation in the roughness data,
and iii) the sample analyzed by DU became partially contaminated, resulting in a higher surface
roughness being measured at two of the three sites that were analyzed. Speculation aside, it is
clear that collecting data from multiple sites (> 3) on the membrane surface is necessary in order
to account for the natural variation in surface morphology between surface sites.
In Figure 3.1, it is difficult to compare the rendered images of the ESNA1-LF membrane
from the two laboratories because the images were developed using different z-scales. This
observation brings an important conclusion for developing a standard technique for comparing
AFM-generated images for membranes, which is that both images must utilize the same x-, y-,
and z-scales in order to facilitate easier visual comparison. That said, the roughness statistics (Rq,
Ra) should be relied on much more than visual comparison.
One reason for the differences in AFM data from the two laboratories could be
differences in AFM tips used. It is known that probe tip diameter and composition profoundly
influence the AFM image x/y resolution. The OCWD laboratory used a silicon nitride MSNCMT-B tip (nominal tip radius 10 nm, maximum tip radius 40 nm; Bruker AFM Probes,
Camarillo, CA) and the DU team used a silicon nitride ORC8-10 tip (nominal tip radius 15 nm,
maximum tip radius 20 nm; Bruker AFM Probes, Camarillo, CA). While the tips are made of the

31
2012 Water Research Foundation. ALL RIGHTS RESERVED.

same material and their geometric features are similar, they are not identical, particularly when
the variability in tip radius (nominal vs. maximum value) is considered.
Table 3.1
Surface roughness statistics generated from atomic force microscope (AFM) imaging of the
ESNA1-LF membrane surface by the OCWD and DU laboratories (resolution = 256 256,
measurement mode = tapping, medium = water). Numerals (1, 2, ..) correspond to a sample
site on the membrane.
Stan.
Laboratory /
Avg.
1
2
3
4
5
6
Dev.
Roughness Statistic
[nm]
[nm]
Rq [nm]
35.6
37.5
33.7
30.3
36.1
46.1
36.6
5.3
OCWD Ra [nm]

DU

28.2

29.7

26.4

23.7

Size [m]

55

55

55

55

Rq [nm]

96.0

80.9

Ra [nm]

73.8

63.7

Size [m]

28.3

34.9

28.5

3.7

53.7

76.9

21.4

41.1

59.5

16.7

10 10 10 10

10 10 10 10 10 10

(a)

(b)

Figure 3.1. Rendered 3D images of the ESNA1-LF membrane based on atomic force
microscope (AFM) data from (a) OCWD and (b) DU (scan size = 100 m2, measurement
mode = tapping, medium = water).
SW30HR Membrane
Surface roughness statistics for the SW30HR membrane are summarized in Table 3.2.
Representative 3D rendered AFM images of the SW30HR membrane surface are given in Figure
3.2. Overall, the SW30HR membrane may be characterized as a relatively rough membrane
based on the data from both laboratories. This observation is supported by the 3D rendered

32
2012 Water Research Foundation. ALL RIGHTS RESERVED.

image of the membrane surface (Figure 3.2a). The agreement between the roughness statistics
for the SW30HR membrane was considerably better than that for the ESNA1LF membrane.
Relative standard deviations are approximately 8% and 5% for Rq and Ra by the DU and OCWD
laboratories, respectively.
Table 3.2
Surface roughness statistics generated from atomic force microscope (AFM) measurements
of the SW30HR membrane surface from the OCWD and DU laboratories (resolution = 256
256, measurement mode = tapping, medium = water). Numerals (1, 2, ..) correspond to a
sample site on the membrane.
Stan.
Laboratory /
Avg.
95%
1
2
3
4
5
Dev.
Roughness Statistic
[nm]
Conf.
[nm]
Rq [nm]
86.8
89.6
91.8
77.0
77.5
84.5
6.9
8.5
OCWD Ra [nm]
Size [m]
DU

69.4

71.6

72.4

10 10 10 10 10 10

61.3

61.6

55

55

67.3

5.4

6.8

Rq [nm]

87.1

87.7

95.1

90.0

4.5

12

Ra [nm]

68.5

68.3

75.9

70.9

4.3

11

Size [m]

10 10 10 10 10 10

(a)

(b)

Figure 3.2. Rendered 3D images of the SW30HR membrane based on atomic force
microscopy data from (a) OCWD and (b) DU (scan size = 100 m2, measurement mode =
tapping, medium = air). Note that the z-scale is different between the rendered 3D images
(0.35 m/division for Figure 3.2a compared to 0.305 m/division for Figure 3.2b).
Repeatability and Reproducibility
The surface roughness statistics for the ESNA1-LF and SW30HR membranes that were
measured by the OCWD and DU laboratories were further analyzed to assess the precision of the
measurements between the two laboratories (i.e., an inter-laboratory comparison). Precision

33
2012 Water Research Foundation. ALL RIGHTS RESERVED.

statistics for both Rq and Ra are summarized in Table 3.3. Despite differences in average values
determined between the two laboratories for the SW30 membrane, the repeatability and
reproducibility of the two roughness statistics are reasonably good. The higher repeatability (sr)
and reproducibility (sR) standard deviations for the ESNA1-LF membrane may be attributed due
to greater surface heterogeneity of this membrane. Interestingly, however the repeatability and
reproducibility of the statistics was better for the rougher membrane surface.
Table 3.3
Precision statistics for the surface roughness data that was collected for the ESNA1-LF and
SW30HR membranes (cell test results from one laboratory on one material, repeatability
standard deviation (sr), reproducibility standard deviation (sR).
Rq Roughness Precision Data
Average of Cell
Repeatability (sr)
Reproducibility (sR)
Material
Averages, [nm]
[nm]
[nm]
ESNA1-LF
56.7
16
31
SW30HR

86.0

Ra Roughness Precision Data


Average of Cell
Material
Averages, [nm]
ESNA1-LF
44.0
SW30HR

5.8

5.2

Repeatability
[nm]
12

Reproducibility
[nm]
24

4.9

2.6

69.1

Surface roughness statistics for all five of the membranes that were analyzed by OCWD are
summarized in Table 3.4. Representative 2D images of the five membranes are shown in Figure
3.3. Results are discussed here for each of the five membranes in order to provide a basis for
analyzing the membrane fouling results that are presented later in this report.
Table 3.4
Surface roughness statistics as measured using an atomic force microscope (AFM) for the
NF-270, ESNA1-LF, ESPA-2, and SW30HR membranes. All measurements were
performed on flat sheet membrane samples by the Orange County Water District (OCWD)
laboratory. All measurements were performed on wet membrane samples in contact mode
and using a liquid cell.
No. of
Rq
St Dev
95% conf.
Ra
St Dev
95% conf.
Scans
[nm]
[nm]
[nm]
[nm]
[nm]
[nm]
8
3.8
1.9
1.8
2.5
0.9
0.8
NF-270
ESNA1-LF

36.6

5.3

5.6

28.5

3.7

3.9

ESPA-2

38.8

12.9

11.1

30.8

10.3

8.8

SW30HR
MUNIRO400

84.5

6.9

8.5

67.3

5.4

6.8

40.7

4.6

4.5

31.8

3.1

3.0

34
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ESNA1-LF

SW30HR

ESPA-2

MUNIRO-400

NF-270

Figure 3.3. Rendered 2D atomic force microscope (AFM) images of the ESNA1-LF,
SW30HR, NF-270, ESPA-2, and MUNIRO-400 membranes. All images were generated by
the AFM OCWD laboratory.
Of the five membranes analyzed by OCWD, the NF-270 membrane was characterized as
having the smoothest surface. Both the Rq and Ra for the NF-270 membrane were smaller by an
order of magnitude than those for the four other membranes. This observation is further
supported by the 2D image of the NF-270 surface, which lacks the grain structures that are seen
in the images for the other membrane surfaces. The SW30HR membrane was determined to have
the roughest surface, with the remaining membranes having similar Rq values of approximately
30 nm. The standard deviation in the values for Rq and Ra for all of the membranes was relatively
low (< 9 nm), with the exception of that for the ESPA-2 membrane.
The Ra value for all of the membranes was always smaller than the Rq values. Plotting Ra
as a function of Rq revealed that the relationship between these two parameters is linear (Figure
3.4). A possible explanation for this relationship is that distributions of populations of zi
coordinates that describe membranes exhibit some mathematical regularity that is not dependent
on the range of the z variable. Also, this indicates that either value could be used for correlation
of membrane properties with membrane performance.

35
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.4. Plot of Ra as a function of Rq for the membranes investigated.


Imaging and Roughness Analysis of Curved (Hollow Fiber) Membrane Surfaces
Teflon (PTFE) Standard Surface
Surface roughness statistics for the polytetrafluoroethylene (PTFE), or Teflon, tubing,
which was investigated for its suitability as a standard surface for AFM imaging of hollow fiber
membranes, are reported in Table 3.5. The OCWD laboratory imaged only the outside (convex
portion) of the Teflon tubing, while the DU laboratory imaged both the inside (concave portion)
and outside of the Teflon tubing. It is important to note that both laboratories analyzed the same
Teflon tubing. When seeking out a standard flat surface for the AFM surface roughness
measurements Teflon was considered to be an ideal surface as a result of its various
characteristics. The nature of the carbon-fluorine bond makes Teflon a non-stick (low adhesion)
material, which would facilitate more intimate characterization of surface features without
interference from interfacial chemical interactions.
The two laboratories produced dramatically different roughness statistics for the outside
of the Teflon tubing, with the data from OCWD suggesting a smooth surface while that from DU
indicated a rough surface. Of note is the substantial standard deviation in the results from DU for
the outside of the Teflon tubing (Table 3.5), which suggests some error in the measurement. The
source of this disagreement is unclear; however, one would expect the Teflon surface to in fact
be smooth as suggested by the OCWD data. For this reason, the differences in measurement
modes used by the two laboratories should be considered. OCWD utilized contact mode, while

36
2012 Water Research Foundation. ALL RIGHTS RESERVED.

DU used tapping mode. The need for contact mode at the OCWD laboratory was attributed to the
small size of the AFM tip (10 nm radius) and the weak interactions that occurred between the tip
and the Teflon surface. The weak interactions between the tip and surface resulted in the tip
crashing into the surface on approach. This problem could be resolved in the future by using
AFM tips with larger tip diameters, which would result in an increase in the overall force that is
detected by the tip on approach to the sample surface (interaction force scales with the size of the
AFM tip). In the absence of larger tips, it may also be possible that tapping mode is not suitable
for imaging curved surfaces like hollow fiber membranes. This point will however require
further research to determine the ability of either mode to accurately analyze curved surfaces.
DU roughness statistics for the inside of the tubing indicate that it is smoother than the outside. If
true (no comparative data from OCWD was available), then it will be important to consider the
flow orientation for hollow fiber membranes (outside-in or inside-out) when selecting which
surface to image. This is an important note as even for a symmetric material such as the Teflon
tubing studied here, significant differences in the surface morphology from the outside to the
inside were observed.
Inspection of 2D rendered AFM images from the OCWD laboratory (data not shown)
found that there was a great variability in the morphology of the Teflon surface. This was
attributed to extrusion nozzle imperfection during Teflon manufacture. This observation, coupled
with the substantial inter-laboratory discrepancies, suggests that additional testing is required to
demonstrate that Teflon tubing is a suitable standard surface for the imaging of hollow fiber
membranes.
Table 3.5
Surface roughness statistics for Teflon tubing, which was being considered as a standard
surface for analyzing hollow fiber membranes using atomic force microscopy (resolution =
256 256, medium = water, OCWD contact mode, DU tapping mode).
1

Avg.
[nm]

OCWD-outside Rq [nm]

22.0

9.9

12.5

18.1

24.7

28.6

19.3

Stan.
Dev.
[nm]
7.2

Ra [nm]

15.5

7.3

9.7

13.4

16.3

20.8

13.8

4.8

5.1

Size [m]

55

55

55

55 1010 1010

Rq [nm]

64.8

166.0

115.4

71.6

643

Ra [nm]

50.9

126.0

88.5

53.1

477

Sample Site # /
Laboratory

DU outside

Size [m]
DU inside

95%
Conf.
[nm]
7.6

1010 1010

Rq [nm]

34.4

79.7

34.0

49.4

26.3

66

Ra [nm]

25.6

66.8

26.5

39.6

23.5

59

Size [m]

1010 1010 1010

37
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Standard Method for AFM Surface Roughness Measurements on Flat-Sheet and Hollow
Fiber Membranes
Based on advice received from Dr. K. Khulbe (an AFM expert from the University of
Ottawa), feedback from participants (particularly OCWD), and information from Agilent
experts, a standard technique for conducting AFM surface roughness measurements on flat-sheet
and hollow fiber membranes was developed and is presented in Appendix C.
METHOD DEVELOPMENT FOR CONTACT ANGLE MEASUREMENTS
A flat-sheet Teflon surface was determined to be an appropriate standard surface for
contact angle measurements (as will be shown in subsequent sections of this report). This
conclusion was based on the consistency in the data that were acquired from each of the
participating laboratories (excluding one outlier, likely resulting from operator error). However,
it should be noted that the contact angle value for which the Teflon is certified is the advancing
sessile drop value and the value recommended for use in the standard technique is the receding
captive bubble value; thus re-certification of the Teflon for the receding value is required. For the
curved standard surface it is recommended that the Teflon tubing be pursued, as the PMMA film
that was investigated here was unstable when exposed to water.
Contact Angle Analysis of the Flat-Sheet Teflon Standard Surface
The contact angle for the Teflon standard surface is reported by the manufacturer to range
between 102.7 and 105.9. Contact angle data for the Teflon standard surface measured by the
participating laboratories are summarized in Table 3.6 and Figure 3.5. The standard deviation
boxes for each laboratory represent the standard deviation of the data collected at each respective
laboratory and are a measure of the internal precision for each participant and sample. The
confidence intervals at 95% confidence (barring systematic error, there is a 0.95 probability that
the true average is within these intervals) overlap with standard deviations within 4% for a given
sample size (6 measurements); thus, 99% confidence intervals are shown in the right panel of
Figure 3.5. The contact angle values from DU and UNR are lower than the reported certified
values for Teflon but are consistent with a hydrophobic surface. The deviation from the certified
values may be attributed to the fact that receding contact angles were measured at both DU and
UNR, while the manufacturer reported the advancing contact angle. However, because Teflon is
strongly hydrophobic, and thus should not wet out, surface hydration would not appear to be a
complete explanation for the observed difference in contact angle results. In contrast to the DU
and UNR results, those from CSM were indicative of a hydrophilic surface ( << 90). The
source for such a low contact angle with water for a known hydrophobic material is unknown,
though surface contamination was ruled out as a possibility.

38
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.6
Contact angle data for the Teflon standard surface from three the participating
laboratories (Colorado School of Mines - CSM, Duke University DU, and the University
of Nevada, Reno UNR). All contact angles were measured using the captive bubble
method and doubly deionized water as the contact angle probe liquid (T = 20C).
Lab / Air Bubble #

CSM

44.8

49.3

46.5

54.3

56.6

56.3

DU

95.2

91.7

95.7

90.5

92.7

87.0

UNR

100.0

100.3

100.3

98.4

100.6

98.2

Figure 3.5. Box and whiskers plot for Teflon contact angle data: central bar mean,
outside bars minimum/maximum, boxes standard deviation range or confidence
interval. The 95% confidence interval for 6 measurements is very close to the sample
standard deviation span, so 99% confidence intervals were plotted on the right. CSM
Colorado School of Mines, DU Duke University, UNR University of Nevada, Reno. All
contact angles were measured using the captive bubble method and doubly deionized water
as the contact angle probe liquid (T = 20C).
Statistical analyses were conducted on the contact angle data from the three laboratories
to determine if both DU and UNR determined the same average contact angle value for the

39
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Teflon standard. Excluding the contact angle data from CSM the P-value for the ANOVA
analysis is 6.4410-5. Because the P-value is less than a significance level of 0.05, the hypothesis
that the DU and UNR results are in agreement is rejected. Therefore, while both the DU and
UNR laboratories produced similar results, they are in fact statistically different. Thus, three
separate laboratories produced statistically different average contact angle values for the Teflon
standard surface. The source of this disagreement is unclear but may result from physical
deformation of the Teflon material (e.g., the angle between the camera axis and the sample plane
is perhaps different than zero as shown in Figure 3.6). In Figure 3.7 the tear-drop shape shows
two contact angles: a larger (advancing) angle in the front toward the sliding direction, and a
smaller (receding) angle in the back. The advancing angle is consistent with the value reported
by the manufacturer (102.7 to 105.9). Conversely, the receding contact angle ( = 90) is
consistent with the contact angle results determined using the captive bubble method, which
measures the receding angle (Table 3.6). Surface contamination is a less likely scenario as the
Teflon surface was cleaned prior to the measurements and because it is a homogeneous material,
thus, impacts associated with surface chemical heterogeneities should be minimal. It is therefore
most plausible that differences in instrumentation and procedural protocols between the
laboratories were the most significant contributors to the observed variations in the reported
contact angle values. The substantial difference between those results from DU and UNR with
that from CSM, in addition to the difference between the results from CSM and those in the
literature for Teflon, suggest that the CSM results likely suffer from some significant procedural
error (e.g., mistaken sample identity).
CSM

DU

UNR

Figure 3.6. Digital images of air bubbles on the Teflon standard surface from each of the
three participating laboratories (Colorado School of Mines CSM, Duke University DU,
and the University of Nevada, Reno UNR). In each case the contact angle was measured
using the captive bubble technique in which the Teflon surface was immersed in doubly
deionized water (T = 20C).

40
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.7. Digital image of a water droplet (Vdrop = 20 L) on a Teflon standard surface set
at an incline.
Contact Angle Analysis of Flat Sheet Membrane Surfaces
Hydranautics ESNA1-LF Membrane
Contact angle data collected from the participating laboratories for the ESNA1-LF
membrane are summarized in Table 3.7 and Figure 3.8. Additionally, representative captive
bubble images for the ESNA1-LF membrane from each laboratory are shown in Figure 11. Both
CSM and UNR produced average contact angle values for the ESNA1-LF membrane that were
statistically the same (P-value 0.14 > 0.05), while DU generated data that was statistically
different (P-value 1.110-5). Considering the entire contact angle data set, the membrane may be
characterized as being hydrophilic, as indicated by the relatively low contact angle with water
that was measured at each laboratory. Therefore, while the specific contact angle value differs
amongst the three laboratories the overall qualitative outcome does not. The standard deviation
for the contact angle data was rather large (> 5), with the exception of that for the CSM data set
(1.5). Because each laboratory studied the same membrane sample it is possible to eliminate
chemical heterogeneity of the membrane surface as the source of the large standard deviation in
the contact angle data from DU and UNR. Thus, the deviation in the data may be attributed to
contamination of the membrane surface, instrument error, and/or procedural error (insufficient
conditioning, improper sample storage). Of these possibilities, the latter two are most likely as
each laboratory was instructed to handle the membrane samples with powder-free gloves and to
clean them with deionized water.
Table 3.7
Contact angle data for the ESNA1-LF membrane from the three participating laboratories
(Colorado School of Mines - CSM, Duke University DU, and the University of Nevada,
Reno UNR). All contact angles were measured using the captive bubble method and
doubly deionized water as the contact angle probe liquid (T = 20C).
Lab/bubble

Avg.

CSM

25.1

23.9

22.4

23.1

22.6

20.8

23.0

DU

34.1

41.2

49.3

40.8

36.1

44.2

41.0

UNR

26.0

25.9

35.5

27.5

19.0

25.4

25.4

41
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.8. Box and whiskers plot for the ESNA1-LF contact angle data, which was
measured using the captive bubble technique. The 95% confidence interval for 6
measurements is very close to the sample standard deviation span, so 99% confidence
intervals were plotted on the right. CSM Colorado School of Mines, DU Duke
University, UNR University of Nevada, Reno.
Inspection of the representative captive bubble images from each laboratory (Figure 3.9)
suggests that inclination of the sample stage and the use of non-optimal lighting may have
contributed to discrepancies in the contact angle data. The lack of sufficient sample lighting is
evident in the image from DU. Inclination of the sample at both CSM and DU is indicated by the
lack of a clear interface between the membrane surface and the water. Both of these specific
error sources are of a critical nature due to the dependence of the bubble shape, and thus, the
contact angle on the sample orientation and interpretation of the three-phase interface by the
digital software. Any deviation of the sample from a planar orientation will result in errors in the
measured contact angle, with the magnitude of the error increasing the farther the sample is from
planar orientation. Mitigating these sample preparation and procedural errors would likely reduce
the internal standard deviation in the data as well as the global deviation in the entire data set. It
is interesting to note here that visual observations of the contact angle measurements are key to
identifying possible errors and discrepancies in the results.

42
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CSM

DU

UNR

Figure 3.9. Representative digital pictures of air bubbles from which contact angles were
measured on the ESNA1-LF membrane at each of the three participating laboratories
(Colorado School of Mines - CSM, Duke University DU, and the University of Nevada,
Reno UNR).
Dow Filmtec SW30HR
Contact angle data from the participating laboratories for the SW30HR membrane (Dow
Filmtec) are summarized in Table 3.8 and Figure 3.10. Overall, the contact angle data indicate
that the SW30HR membrane is hydrophilic, with reasonably good qualitative agreement between
each of the three data sets. Both DU and UNR produced statistically similar results (P-value =
0.64 > 0.05), while that from CSM was different (P-value = 8.810-4 < 0.05). Inspection of
representative digital images of the air bubbles from the three laboratories (Figure 3.11) suggests
that the instruments camera at CSM may have been placed in a way that truncates the bubble
base, which resulted in the measured contact angle being larger than the actual value. With the
exception of the data set from UNR (STDEV > 5), the standard deviation amongst the contact
angle data from any given laboratory was relatively low (STDEV < 1.8). Although the UNR
laboratory measured large standard deviations (STDEV > 5) for the contact angle data for both
membranes, it measured a relatively small standard deviation (STDEV = 1) for the Teflon
standard. It is therefore most likely that the source of the high standard deviation is due to
techniques or procedures specific to the membranes, rather than an inherent instrument error.
Table 3.8
Contact angle data for the SW30HR membrane from the three participating laboratories
(Colorado School of Mines - CSM, Duke University DU, and the University of Nevada,
Reno UNR). All contact angles were measured using the captive bubble method and
doubly deionized water as the contact angle probe liquid (T = 20C).
Lab/bubble

Avg.

CSM

40.3

40.0

39.7

42.3

40.4

44.1

41.1

DU

34.2

33.4

33.7

32.5

31.3

32.3

32.9

UNR

30.0

29.2

29.0

35.8

39.9

39.7

39.7

43
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.10. Box and whiskers plot for the SW30HR contact angle data that was generated
by the Colorado School of Mines (CSM), Duke University (DU) and the University of
Nevada, Reno (UNR). All contact angles were measured using the captive bubble method
and doubly deionized water as the contact angle probe liquid (T = 20C).
CSM

DU

UNR

Figure 3.11. Representative digital images of air bubbles on the SW30HR membrane
immersed in water at the CSM, DU, and UNR laboratories. Each of these air bubbles are
representative of those from which contact angle measurements were performed.
Statistical Analysis of Contact Angle Results for Flat Surfaces
Average values, in addition to reproducibility and repeatability statistics, for the contact
angle data that was collected from the three participating laboratories is summarized in Table
3.9. The repeatability was highest for the Teflon standard and lowest for the ESNA1-LF
44
2012 Water Research Foundation. ALL RIGHTS RESERVED.

membrane. Reproducibility was highest for the SW30HR membrane and lowest for the ESNA1LF membrane. Notably the repeatability and reproducibility did not correspond with surface
roughness or chemical heterogeneity. Here, the ESNA1-LF membrane was characterized by a
smoother surface relative to the SW30HR membrane, but had the lowest degree of
reproducibility and repeatability with regards to its contact angle data. Similarly, the Teflon
standard would be thought to have the most chemically homogeneous surface relative to the two
membranes, but had a lower degree of reproducibility compared to the SW30HR membrane.
Table 3.9
Average contact angle values for the PTFE standard surface, ESNA1-LF, and SW30HR
membrane. The associated statistical measures (repeatability and reproducibility)
from the inter-laboratory study are also reported for each test surface.
Average of Cell
Repeatability
Reproducibility
Material
Averages []
[]
[]
95.9
2.4
5.6
PTFE
ESNA1-LF

30.2

4.5

10.4

SW30HR

36.0

3.3

5.2

A certified contact angle fixed-drop calibration reference tool (Figure 3.12) from RamHart (Netcong, NJ) was used at the UNR laboratory to investigate systematic errors in the
proposed standard technique. Contact angle results for the calibration tool were compared to
contact angles measured for five membrane samples at the UNR laboratory (Table 3.10). The
RMS error for the calibration tool contact angle values was 3.8, which is comparable to the
random error that was observed for the membrane contact angle data. This result suggests that
the error associated with the membrane contact angles is likely systematic. User error cannot be
ruled out here, though it is less likely as it would be reasonable to expect user error to be more
random. Nevertheless, additional testing using the calibration tool would be required to
completely rule out user error.
Although the 2D calibration tool is vey convenient, it has a baseline and drop image
occurring at the same plane, which reduces error sensitivity to stage-camera misalignment.
Another 3D contact angle reference tool that is based on gauge balls and gauge blocks with
dimensions traceable to NIST was also evaluated (Figure 3.13). The gauge balls are
characterized by a diameter (D) that can be only slightly larger than the block thickness (P), from
which spherical caps having relatively small contact angles can be produced. It turns out that
front-back inclination of the stage with respect to the camera axis can diminish the appearance of
the spherical cap (smaller contact angle) and even make it disappear ( = 0). This observation
strongly suggests that front-back inclination of the stage with respect to the camera axis can be a
major contributor to the experimental error. Recall that sample inclination was cited as a possible
source of error in the contact angle results from the three laboratories, particularly for the Teflon
standard surface where CSM produced considerably lower results than DU and UNR (Table 3.6).
Considering the importance of sample orientation, it is proposed that the 3D tool shown in Figure
3.13 be used to verify stage alignment prior to contact angle measurement. A possible
stage/sample alignment technique would be to search for that position which results in a
maximum contact angle value in sessile drop mode (or minimum contact angle in captive bubble
mode) using the 3D calibration tool. Alternatively, fabricating a new type of sample holder, that

45
2012 Water Research Foundation. ALL RIGHTS RESERVED.

would include a 3D fixed standard at the edge of the membrane mounting area would provide for
convenient stage alignment verification each time a sample is measured.

Figure 3.12. Picture of the 2D certified contact angle fixed-drop calibration reference tool
(Ram-Hart, Netcong, NJ).

H
DP
=
2=
tan1
2 tan1
DH
P

Figure 3.13. (left) Picture of a 3D contact angle reference tool, which includes the sample
surface stage, flat stainless steel plate, and stainless steel sphere. (right) Schematic
illustration and equations associated with using the 3D contact angle reference tool ( =
contact angle measured through the stainless steel sphere, D = diameter of the stainless
steel sphere, H = distance from the top of the sphere to the top elevation of the stainless
steel cross member, and P = distance from the top elevation of the sample stage to the top
elevation of the stainless steel cross member.

46
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.10
Reported and measured contact angle values for the 2D reference tool. All results were
acquired by the laboratory at the University of Nevada, Reno (UNR).
Actual
Measured
Measured - Actual
Drop
Right []
Left []
Right []
Left []
Right []
Left []
A

31.5

30.0

26.2

25.2

-5.3

-4.8

61.2

60.4

61.3

62.1

0.1

1.7

90.9

89.0

87.7

84.3

-3.2

-4.7

117.2

120.7

115.0

116.0

-2.2

-4.7

Contact Angle Analysis of Curved Surfaces


A new method for measuring contact angles on curved surfaces, such as hollow fiber
membranes, is proposed. The proposed method involves conditioning the membrane and placing
it between two flat pieces cut out of microscope slides that are held in place by a holder with
adjustable screws. The captive bubble accessory from the UNR goniometer setup was
incidentally appropriate for the task. The membrane is stretched gently (to provide a straight
horizontal baseline in the image) and the slides are tightened just enough to hold it in place. The
glass slides should be parallel and enclose empty space below the membrane. The hollow fiber
axis is mounted perpendicular to the camera axis in a reservoir with water and a bubble is
introduced via a U-shaped needle between the slides, where it floats upward and attaches itself to
the membrane (see Figure 3.14).
Contact angle values for hollow fiber membrane samples are reported in Table 3.11.
Results for the Norit-X-Flow membrane ( = 39.5 8.8) are lower than expected based on the
matrix material, PES. However, when comparing to measurements on the matrix material, the
type of measurement (advancing or receding contact angle) would have to be considered as well
as the effects of surface roughness and porosity that would occur for membrane samples.
Furthermore, it is typical for membrane formulations to contain hydrophilic components that
make the membrane more hydrophilic than the matrix materials. A similar contact angle value
was measured for the CMF-S membrane. Again, this value is lower than expected based on the
matrix material.

47
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.14. Photographs of the experimental apparatus that was designed and used to
measure contact angles on hollow fiber membranes. (left and middle) Image of the PTFE
tubing sandwiched between two glass slides and secured in the sample stage. (right) Image
of an air bubble that has been placed on the PTFE tubing immersed in water.
Table 3.11
Contact angle results for hollow fiber membranes. All results were obtained using the
experimental apparatus that was developed as part of this project.
Membrane
Contact Angle,
95% confidence error bar,
Norit X-Flow

39.5

8.8

Siemens Memcor CMF-S

43.4

5.0

Standard Method for Contact Angle Measurements on Flat-Sheet and Hollow Fiber
Membranes
Based on the contact angle analysis carried out by the three participating laboratories, a
standard technique for measuring contact angles on membrane surfaces was developed and is
given in Appendix D. Additional recommendations for measuring contact angles on membrane
surfaces are also given in Appendix D.
METHOD DEVELOPMENT FOR STREAMING POTENTIAL (ZETA POTENTIAL)
MEASUREMENTS
PMMA Standard Surface
Zeta potential results for the PMMA standard surface from UNR and Anton Paar are
given in Figure 3.15. It should be noted that the ionic strength values used by each laboratory
varied. The tests by UNR were at an ionic strength of 2 mM KCl, while those by Anton Paar
were at an ionic strength of 1 mM KCl. The relatively small difference in ionic strengths is not
expected to significantly affect the reported results. The reported data points from UNR represent
averages of three trials on three different membrane coupons. The following observations can be
made regarding the zeta potential data that are presented in Figure 3.15:

Both data sets demonstrate a similar isoelectric point for PMMA (pHiep ~ 4.0 to 4.2), and

48
2012 Water Research Foundation. ALL RIGHTS RESERVED.

The two data sets are nearly identical from pH 3 to pH 7, before diverging. The data from
Anton Paar indicates a slightly more negatively charged surface at pH > 7.

Based on the two data sets reported here it appears that PMMA is an appropriate standard
surface for streaming potential measurements. Furthermore, comparing the zeta potential results
that were produced by the different electrokinetic analyzers that were used at the two laboratories
suggests that uniform results may be achieved regardless of the type of instrument used. In other
words, variations in streaming potential results between labs may be minimized through the use
of standardized test procedures developed in this project.

Figure 3.15. Zeta potential data collected for the PMMA control surface from UNR
University of Nevada, Reno (I = 2 mM KCl) and Anton Paar (I = 1 mM KCl). Polynomial
fits (cubic) to both data sets are also provided. .
Streaming Potential (Zeta Potential) Analysis of Flat Sheet Membrane Surfaces
ESNA1-LF Membrane
Zeta potential data as a function of solution pH for the ESNA1-LF membrane are reported in
Figure 3.16. The following observations may be made for these data:

The zeta potential results between the four laboratories all follow a similar trend and
demonstrate good reproducibility independent of the electrokinetic analyser used,

49
2012 Water Research Foundation. ALL RIGHTS RESERVED.

The results from UCR and Anton Paar overlap over the entire pH range tested here; in
comparison the results from UNR were less negatively charged at pH 4, but more
strongly negatively charged at pH > 5,

Those zeta potential results from FKKT were more strongly negatively charged than
those results from UCR and Anton Paar at pH > 3,

The measured pHiep for the ESNA1-LF membrane based on the reported zeta potential
results is between 3.1 and 3.9, and

The ESNA1-LF membrane is strongly negatively charged over a pH range of 4 to 9.

Figure 3.16. Zeta potential as a function of solution pH for the ESNA1-LF membrane (I = 2
mM KCl). Streaming potential measurements were by the UCR University of California
Riverside, UNR University of Nevada, Reno, the FKKT - University of Maribor, and
Anton Paar. Those data points from UCR, FKKT, and Anton Paar represent the average
of three separate tests on three different membarne coupons.

50
2012 Water Research Foundation. ALL RIGHTS RESERVED.

SW30HR Membrane
Zeta potential as a function of solution pH for the SW30HR membrane is reported in Figure
3.17. The following observations may be made for these data:

The zeta potential results for the SW30HR membrane were less reproducible between all
four laboratories relative to those results for the ESNA1-LF membrane,
With the exception of those results from UCR (higher deviation in results) the individual
results for each laboratory were reproducible,
Results from the UCR, UNR, and FKKT all showed a more negatively charged
membrane surface relative to those zeta potential values measured by Anton Paar at pH >
3,
The measured pHiep for the SW30HR membrane ranged from 2.4 to 3.9, and
The SW30HR membrane is less negatively charged relative to the ESNA1-LF membrane
over the same pH range (4 to 9).

Figure 3.17. Zeta potential as a function of solution pH for the SW30HR membrane (I = 2
mM KCl). Streaming potential measurements were by the UCR University of California
Riverside, UNR University of Nevada, Reno, the FKKT - University of Maribor, and
Anton Paar. Those data points from UCR, FKKT, and Anton Paar represent the average
of three separate tests on three different membrane coupons.

51
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Interpretation of the zeta potential results for the two membranes indicates that relatively
reproducible results may be obtained between laboratories. This observation is independent of
the types of electrokinetic analyzer used, but dependent on the use of a standard method for
conducting the streaming potential measurements. Additional tests by Anton Paar (data not
shown) suggest that the differences in zeta potential results between the laboratories may be
attributed to slight variations in membrane pretreatment conditions (i.e., soaking time and
conditioning solution). For example, the measured zeta potential for both the ESNA1-LF and
SW30HR membrane decreased (became less negative) when the membrane was allowed to soak
in an electrolyte solution for 24 hours prior to conducting the streaming potential measurements.
In other words, analyzing dry versus wet membranes, and specifically the soaking time and
chemistry of the soaking solution, may result in different zeta potential values. Although the
specific mechanisms by which the soaking conditions affect the measured zeta potential remain
unclear at this point, specific membrane pretreatment conditions were addressed in the final
version of the standard method for streaming potential measurements (Appendix D).
Streaming Potential (Zeta Potential) Analysis of Hollow Fiber Membrane Surfaces
Norit X-Flow Membrane
Streaming potential measurements for hollow fiber membranes has in the past proven to
be a challenge. This is due to the fact that the standard measurement cells for electrokinetic
analyzers cannot be used for hollow fibers. This challenge was overcome here by modifying a
streaming potential test cell to accommodate a bundle of hollow fibers, which were potted in
place (see Appendix D for further details). Using this approach, streaming potential
measurements were using the Norit X-flow hollow fiber membrane. In these measurements the
streaming potential was measured for the interior of the hollow fiber; however, this same
technique could be applied to the exterior portion of the fiber as well. The zeta potential results
as a function of solution pH for this hollow fiber membrane are reported in Figure 3.18. The
Norit X-flow membrane was characterized by an isoelectric point (pHiep) of approximately 6. It
was positively charged below pH 6 and increasingly negatively charged as solution pH became
more basic. Overall, it can be said that this membrane is relatively weakly charged ( < 10 mV)
over the pH range tested here (3 to 7). The reported zeta potential results were highly
reproducible as indicated by the small error bars in Figure 3.18. These results indicate that the
proposed method for measuring streaming potential for hollow fiber membranes is viable and
capable of producing reproducible results.

52
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.18. Zeta potential as a function of solution pH for the inner surface of the Norit XFlow hollow fiber membrane (I = 2 mM KCl, n = 3).
Standard Method for Streaming Potential Measurements on Flat-Sheet and Hollow Fiber
Membranes
The reproducibility in zeta potential values between different laboratories using different
electrokinetic analyzers was determined to be good when using the proposed standard method
for streaming potential measurements. The proposed standard method was slightly modified to
highlight the importance of uniform membrane pretreatment conditions prior to carrying out the
streaming potential measurements. The final version of the standard method for conducting
streaming potential measurements on both flat sheet and hollow fiber membrane surfaces is
presented in Appendix D.
METHOD FEEDBACK FROM PARTNER UTILITIES
The distribution of surveys that were returned and incorporated into this report is given in
Figure 3.19. A total of 25 surveys were distributed and 11 were completed and returned. The
largest number of responses to the survey was from academia followed by water/wastewater
utilities that are currently using membranes in their treatment plants.

53
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.19 Distribution of survey responses amongst each of the four respondent
categories (water/wastewater utilities, membrane manufacturers, characterization
equipment vendors, and academia). The total number of survey responses that were
received was eleven.
The distribution of responses to the question Given our description of the technique and
the full test method, would you envision finding the results of the contact angle test useful? are
reported in Figure 3.20. All respondents to this question stated that they thought that contact
angle measurements are useful for some purpose in membrane applications. Specific application
areas are summarized in Table 3.12. In general, academics and researchers felt that contact angle
measurements are useful for characterizing the interfacial properties of membranes, while those
in the industry felt that contact angle measurements are best suited for use in selecting an
appropriate membrane for a given application. In fact, all of the academic respondents stated that
they use contact angle measurements frequently in their respective research efforts. Utility
respondents did not feel that they are equipped to perform contact angle measurements for
various reasons (see Table 3.12). Similarly, the responding utilities do not use contact angle
results in the day-to-day operation of their systems. Instead, it was commented that contact angle
results are only useful during membrane selection. These results suggest that while the
importance of understanding the hydrophobicity of a membrane is important to all those
involved in membrane applications, the ability to perform these measurements, and use of the
subsequent results, is restricted to academics and researchers.

54
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.20. Distribution of responses to the question on the usefulness of contact angle
measurements in membrane applications.
All of the academic respondents felt that they could follow and conduct the contact angle
measurements as detailed in the draft standard procedure. Conversely, all utilities and one out of
two membrane manufacturers responded that they could not conduct the contact angle
measurement. This response is attributed to the fact that the utilities and membrane manufacturer
lacked the necessary equipment (i.e., a goniometer) for conducting these measurements.
Surveying the utilities found that like most installations, with the exception of the Orange
County Water District, they are not equipped with research facilities beyond those basic ones
required for water quality testing. Indeed, a lack of available resources and personnel appear to
be the largest hurdles for utilities to take advantage of contact angle measurements (see Table
3.12).
In response to their thoughts on the precision of the presented contact angle method all of
the respondents (the utilities, membrane manufacturers and equipment vendors, and two from
academia) were unsure and in general preferred to withhold judgment until data from various
independent labs could be compared. Those from academia believed that the method was precise
in the manner in which it was presented, though they could not comment on the data that would
be generated.

55
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.12
Specific comments on the usefulness and technique for carrying out contact angle
measurements in membrane applications.
Question Area
Comments

Would the information that could


be gained from contact angle
measurements be useful to you?
How would you foresee using the
results and how often do you think
you would like to have the contact
angle test conducted?

Academia
Cited uses included: (6 respondents) calculate
interfacial components, (1 respondent) calculate
surface charge / zeta potential.
Membrane Utility (2 respondents) The information obtained from the
tests would be useful to membrane designers when
selecting a membrane for a given application / source
water.
The data gathered from these tests are something seen
as a membrane selection tool and are not really
important to actual treatment operations.
We would require a membrane supplier to provide this
information (contact angle) as certified test results as a
required submittal for membrane procurement. The
test method would be specified.
We would only perform (contact angle measurements)
again if Federal or State regulations required the test
be performed on used membranes after a predetermined time of service. The intent being we would
be required to change the pressure set points of our
full-scale plants membrane integrity test system if the
contact angle significantly changed.
Membrane Manufacturers
Yes, because hydrophilicity is an important material
property for water filtration membranes

Does your laboratory have the


capability of conducting contact
angle measurements?

Academia
(6 respondents) Yes.
Membrane Manufacturers
Yes

The distribution of responses to the question Given our description of the technique and
the full test method, would you envision finding the results of the zeta potential test useful? are
reported in Figure 3.21. No responses from manufacturers of streaming potential instruments
were received and therefore Equipment Vendor is omitted from Figure 3.21. Similar to what
was observed for contact angle measurements all of the respondents felt that zeta potential
measurements are useful in their respective applications (Table 3.13). However, as was also
observed for contact angle measurements, no utilities were capable or equipped to conduct these

56
2012 Water Research Foundation. ALL RIGHTS RESERVED.

measurements (i.e., streaming potential measurements). Coincidently, only those respondents


from academia responded that they use zeta potential measurements on a frequent basis.

Figure 3.21. Distribution of responses to the question on the usefulness of zeta potential
values, acquired through streaming potential measurements, in membrane applications.
Overall, the survey respondents felt that the learning curve for conducting the streaming
potential measurements would be steeper than that for the contact angle measurements.
However, the learning curve was not thought to be insurmountable.

57
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.13
Specific comments on the usefulness and technique for carrying out streaming potential
measurements to calculate the zeta potential of membranes in water and wastewater
treatment applications.
Comments1

Question Area

Would the information that could


be gained from the zeta potential
measurements be useful to you?

Academia
(6 respondents) Yes, zeta potential is useful for
characterizing membrane-solute interactions.
Membrane Utility
It (zeta potential) would be useful but there are some
challenges.
Membrane Manufacturer
Yes because it is undisputed that the charge of the
surface has an influence on the fouling potential and
even the separation performance of a membrane.

How would you foresee using the


results and how often do you think
you would like to have the
streaming potential (zeta potential)
test conducted?

Academia
(3 respondents) They would be used anytime a
new/novel membrane is made/tested.
Zeta potential tests would be highly useful to gather
insight into membrane fouling and rejection
performance. I imagine tests could be performed as
often as once a week.
Yes, the zeta potential can show surface charge
density of a membrane, which can help predict or
explain results of particular tests. The zeta potential
test should be used to characterize different film
coatings on membranes.
Membrane Utility
We would require the membrane supplier to
provide this information as certified test result for
virgin membrane as a required submittal for
membrane procurement. The test method would be
specified.
I would need to read more and understand what
initial virgin measures of zeta potential mean for
a membrane fiber and then understand what relative
changes in the zeta potential mean with respect to
membrane fouling and changes in cleaning regimes.
Assuming such information exists, it would be
useful to perform (or have performed) the zeta
potential analysis to assess change in membrane
fouling characteristics over time.

58
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Membrane Manufacturer
Periodically depending on the project requirements.

Does your laboratory/facility have


the capability of conducting
streaming potential (zeta potential)
measurements?

Membrane Utility
No sample preparation for hollow fiber analysis
seems prone to fiber plugging. We would need
training in preparation of the porous plug that was
described in the procedure.
Membrane Manufacturer
We have an apparatus of our own that could
measure the streaming potential of a membrane.
This apparatus is however still in the development
stage, hence it will be difficult to measure this inhouse on a regular basis.

The distribution of responses to the question Given our description of the technique and the full
test method, would you envision finding the results of the surface roughness (AFM) test useful?
are reported in Figure 3.22. Of the three different types of characterization techniques / results
that were presented in the survey it was the surface roughness measurements that were the least
familiar to the respondents from the membrane utilities. This result was reflected in the specific
comments that were received from the utilities (Table 3.14). Interestingly, even those utilities
that responded that the measurements would be useful to them questioned how the results would
be used in a practical application. Similar to the other characterization results, it appears that the
prevailing thought outside of academia is that membrane characterization data would only be
useful during membrane selection and that acquiring this characterization data becomes
impractical once the membrane is in use. The greatest hurdle to characterizing the membrane
properties once it is in service is the removal of a membrane from the operating element or
module. This brings to question the need to implement sacrificial membranes into full-scale
operations that may be easily removed without hindering the operation of the entire membrane
system.

59
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.22. Distribution of responses to the question on the usefulness of surface


roughness measurements, performed using an atomic force microscope, in membrane
applications.

60
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.14
Specific comments on the usefulness and technique for carrying out surface roughness
measurements using an atomic force microscope for membranes in water and wastewater
treatment applications.
Question Area
Comments

Would the information that could be


gained from the membrane surface
roughness measurements be useful
to you and/or your organization?

Membrane Utility
I am unfamiliar with the correlation between surface
roughness and membrane fouling potential. Surface
roughness is not listed on a typical membrane
suppliers technical data sheet.
Membrane Manufacturer
It would depend on the type/characteristics of the
membrane in question.

How would you foresee using the


results and how often do you think
you would like to have the surface
roughness test conducted?

Academia
Roughness data is important for qualitative
interpretations of surface dependent forces, such as
surface charge, etc. Roughness characterization
should be a standard procedure for all new
materials and should be used as often as a lab
studies new membrane materials.
Membrane Utility
To perform this test after initial membrane
procurement would require a means to harvest
fibers from an encased full-scale module which
has been in service. Would need to either sacrifice
a module (adding to the overall cost of doing the
analysis) or work with membrane supplier on a
way to harvest fibers and patch the module casing
to be able to return the module to service.

Does your laboratory/facility have


the capability of conducting AFM
measurements?

Academia
I believe of all the characterization techniques
discussed, AFM is by far the most difficult to
learn. I think our whole lab group will be capable
of conducting the experiment, but it may take
some of us much longer.

GUIDANCE FOR INTEGRITY TEST CALCULATIONS


Pressure integrity testing is a stipulated requirement in the EPAs Long Term 2 Enhanced
Surface Water Treatment Rule (LT2ESWTR) (EPA, 2006). Integrity tests are required to identify
holes, breaks or other losses of structural integrity for MF and UF membranes. This testing is in
place and is necessary to insure the optimal performance of MF/UF membrane processes. These
61
2012 Water Research Foundation. ALL RIGHTS RESERVED.

tests are primarily geared toward identifying integrity breaches that would allow
Cryptosporidium oocysts (dp = 3 m) to pass through the membrane and into the filtrate. The
basis for pressure integrity tests is that any loss of structural integrity would result in a reduction
in the pressure that would be required to observe air bubbling through the pores of the hollow
fiber membrane. The net pressure applied during membrane integrity tests must ensure that any
breach in the membrane structure that is large enough to pass Cryptosporidium oocysts would
also pass air during the test (EPA, 2005). The minimum direct integrity test pressure for
obtaining a resolution of 3 m for the removal of Cryptosporidium may be calculated according
to Eq. 3.1.
Ptest = 0.193 cos + BPmax

(3.1)

where Ptest is the minimum direct integrity test pressure (psi), is the pore shape correction
factor (= 0 to 1), is the surface tension at the air-liquid interface (dynes/cm), is the liquidmembrane contact angle (= 0 to 90), BPmax is the maximum backpressure on the system during
the test (psi), and 0.193 is the constant that includes the defect diameter (i.e., 3 m resolution
requirement) and unit conversion factors (EPA, 2005). A value of 1 for represents a perfectly
cylindrical pore, while values approaching 0 are indicative of greater deviations from this ideal
structure. Theoretically, cylindrical pores are characterized by higher pore liquid entry pressures
relative to non-cylindrical pore geometries. It is current practice to assume that the membrane is
perfectly hydrophilic (i.e., = 0) and that = 1 so as to produce the most conservative estimate
for Ptest. However, in reality most hollow fiber MF and UF membranes are characterized as being
less than perfectly hydrophilic (i.e., more hydrophobic) and as having non-cylindrical pores. On
this subject the following exert is taken from the Membrane Filtration Guidance Manual (EPA,
2005):
The LT2ESWTR does not establish the minimum test pressure to be used during a
pressure-based direct integrity test, but rather only requires that the test achieve a 3m resolution. If a membrane manufacturer has information to support the use of
values other than = 1 and = 0, and these less conservative values are approved
by the State, then Equation 4.1 can be used to calculate the minimum required test
pressure. It is essential that the use of values other than = 1 and = 0 be
scientifically defensible, since the use of inappropriate values could result in the use
of a test pressure that does not meet the resolution criterion established by the rule.
One approach for determining membrane-specific values for and is through
direct experimental evaluation. Because these parameters can have a significant
effect on the required direct integrity test pressure, it is strongly recommended that
States require sufficient justification from a membrane manufacturer prior to
approving the use of values other than = 1 and = 0, such as independent third
party testing results using a method accepted by the scientific community and
demonstrating statistically significant data.
In order to determine a defensible value for , which may be used to calculate an accurate
value for Ptest it will be necessary to apply a standard method for measuring contact angle with
water on hollow fiber MF/UF membranes. On this subject the following general
recommendations are made:

62
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Contact angle results measurements must be made on the same membrane configuration
for which the results are to be applied. In other words, should the results be intended for
calculating Ptest for a hollow fiber membrane, then the contact angle measurements must
be made on a hollow fiber membrane (contact angle results obtained for a flat-sheet
sample cannot be used for calculating the Ptest value for a hollow fiber membrane).

Contact angle results on a certified sample surface, such as those recommended as part of
this report (e.g., PTFE, 2D and 3D calibration tools) should be made alongside the
laboratory results for contact angle measurements on hollow fiber samples.

For a more conservative contact angle measurement, the standard method for measuring
receding contact angle with water on unsealed hollow fiber membranes (as reported in
Appendix D), should be used to determine for use in Eq. 3.1 to calculate Ptest.

If a tensiometric method (e.g., Wilhelmy Plate technique) is used with sealed fibers, then
the less conservative (higher) advancing contact angle will be determined. Results for
receding and advancing contact angle will not agree for membranes due to surface
roughness and porosity.

Further research is recommended to determine the impact(s) of surface roughness and


porosity on the contact angle results, and hence, on the required Ptest. Application of
existing models, such as the Wenzel and Cassie Baxter models, for correcting the
measured contact angle for membrane surface roughness should be examined with
emphasis on how well these corrections improve the accuracy of the resultant Ptest value.

The repeatability/reproducibility in contact angle values between any two laboratories as


determined in this project was 5. Such a difference in water would result in variations in
the required Ptest as determined by an end user or regulatory agency. The difference in the
calculated Ptest value will vary depending on the magnitude of the contact angle (see
Figure 3.23). As the contact angle increases from 0 to 90 the 5 variation results in an
increase in the error that is associated with the calculated Ptest value. For a membrane
characterized by a contact angle of 80 at one lab and 75 at a separate lab the resulting
difference in Ptest values would be 1.2 psi. Future research and consideration by an
appropriate regulatory agency is required in order to determine if such an error is
acceptable.

63
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Difference in direct integrity test pressure


due to 5 standard error in contact angle
measurement. The middle value
represents the integrity test pressure that
would be calculated from the average
contact angle value.

Figure 3.23 Difference in the calculated values for the minimum direct integrity test
pressure (Ptest) for obtaining a resolution (i.e., minimum defect diameter) of 3 m assuming
a standard error in the contact angle value of 5. The values for Ptest were calculated using
Eq. 3.1 ( = 1, = 74.9 dynes/cm, BPmax = 3 psi).

Membrane surfaces become fouled over time with any number of different materials
(organics, inorganics, biological materials), which subsequently alters the hydrophilic and
hydrophobic properties of the membrane surface. The extent of this alteration will be
system specific, as it will depend on the feed water composition and chemistry, in
addition to the properties of the membranes being used. Therefore, values will change
over time and in turn those values measured on virgin membrane samples are not likely
to be accurate as operation time progresses. It is therefore recommended that if measured
values for be used to determine Ptest then new contact angle measurements be
performed on hollow fiber samples harvested from currently operating modules. In order
to alleviate concerns that are likely to be associated with removing fibers from
functioning modules it is suggested that sacrificial membrane modules be included into
system designs that treat side streams of the feed water. Such a design will allow for the
membranes to become similarly fouled as those in the full-scale process, while allowing
for greater flexibility in harvesting membrane samples for contact angle analysis.
Additional research is needed on this subject given the complex relationship between
fouling and its effects on membrane surface chemistry (hydrophilicity/hydrophobicity).
Experimental data is required to characterize the heterogeneity in changes in membrane

64
2012 Water Research Foundation. ALL RIGHTS RESERVED.

surface chemistry (e.g., heterogeneity resulting from the distribution and varying severity
of foulant deposits in a single membrane module) that results from fouling and the extent
to which this heterogeneity affects the reliability of contact angle measurements on
fouled hollow fiber membranes.
EVALUATING CORRELATIONS BETWEEN MEMBRANE CHARACTERISTICS
AND PERFORMANCE
Figures 3.24-28 show the results of flux model fitting for each of the NF and RO
membranes. In these figures, the central lines represent the regression models, the inner
boundaries represent the 95% confidence limits and the outer boundaries represent the model
prediction limits. Figure 3.29 compares the relationship between predicted normalized membrane
flux and bentonite clay loading for all membranes. A clay loading of 5 g m-2 was chosen as the
load used for relating membrane performance to membrane parameters for MLR modeling.

NF270_NORM_FLUX

1
0.9
0.8
0.7
0.6
0

200

400

600

800

1000

NF270_Perm_Volume
Figure 3.24. Results of simple regression analysis for the NF-270 membrane. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.996635;
R-squared = 99.3281 percent; R-squared (adjusted for d.f.) = 99.1042 percent; Standard
Error of Est. = 0.02384; Mean absolute error = 0.015953; Durbin-Watson statistic =
2.77534 (P = 0.6576); Lag 1 residual autocorrelation = -0.434792. The central line
represents the regression models, the inner boundaries represent the 95% confidence limit
and the outer boundaries represent the model prediction limit.

65
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ESNA_1LF_NORM_FLUX

1
0.9
0.8
0.7
0.6
0

200

400

600

800

1000

ESNA_1LF_Perm_Volume
Figure 3.25. Results of simple regression analysis for the ESNA1-LF membrane:. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membranemembrane. Correlation Coefficient =
-0.999234; R-squared = 99.8469 percent; R-squared (adjusted for d.f.) = 99.7959 percent;
Standard Error of Est. = 0.00762926; Mean absolute error = 0.00468329; Durbin-Watson
statistic = 2.83777 (P=0.7339); Lag 1 residual autocorrelation = -0.62428. The central line
represents the regression models, the inner boundaries represent the 95% confidence limit
and the outer boundaries represent the model prediction limit.

66
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ESPA_2_NORM_FLUX

1
0.9
0.8
0.7
0.6
0

200

400

600

800

1000

ESPA_2_Perm_Volume
Figure 3.26. Results of simple regression analysis for the ESPA2 membrane. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.998841;
R-squared = 99.7683 percent; R-squared (adjusted for d.f.) = 99.691 percent; Standard
Error of Est. = 0.0134216; Mean absolute error = 0.00948347; Durbin-Watson statistic =
1.50241 (P=0.0550); Lag 1 residual autocorrelation = 0.0623037. The central line represents
the regression models, the inner boundaries represent the 95% confidence limit and the
outer boundaries represent the model prediction limit.

67
2012 Water Research Foundation. ALL RIGHTS RESERVED.

SW30HR_NORM_FLUX

1
0.9
0.8
0.7
0.6
0

100

200

300

400

SW30HR_Perm_Volume
Figure 3.27. Results of simple regression analysis for the SW30HR membrane. 100 ml of
permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = -0.996067;
R-squared = 99.215 percent; R-squared (adjusted for d.f.) = 98.9534 percent; Standard
Error of Est. = 0.00316966; Mean absolute error = 0.0020736; Durbin-Watson statistic =
2.6779 (P=0.5877); Lag 1 residual autocorrelation = -0.561779. The central line represents
the regression models, the inner boundaries represent the 95% confidence limit and the
outer boundaries represent the model prediction limit.

68
2012 Water Research Foundation. ALL RIGHTS RESERVED.

MUNIRO_NORM_FLUX

1
0.9
0.8
0.7
0.6
0

500

1000 1500 2000 2500

MUNIRO400_Perm_Volume
Figure 3.28. Results of simple regression analysis for the MUNIRO-400 membrane. 100 ml
of permeate corresponds to 1.94 g loaded m-2 membrane. Correlation Coefficient = 0.994937; R-squared = 98.9899 percent; R-squared (adjusted for d.f.) = 98.8215 percent;
Standard Error of Est. = 0.00624946; Mean absolute error = 0.00416953; Durbin-Watson
statistic = 2.42804 (P=0.5686); Lag 1 residual autocorrelation = -0.352731. The central line
represents the regression models, the inner boundaries represent the 95% confidence limit
and the outer boundaries represent the model prediction limit.

69
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Figure 3.29. Normalized membrane flux (J/Jo) predicted by the simple regression models
as a function of bentonite clay loading. From this, membrane performance of all five
membranes at at a given clay load can be determined. A 5g m-2 bentonite load was chosen
with which to generate performance data for MLR analysis because this falls within a
nearly linear portion of all five relationships and membrane performance could be
determined in all cases at this point by interpolation of experimental data.
Multiple Linear Regression (MLR) Methods
The relationship between normalized flux at 5 g m-2 bentonite clay loading and
membrane properties were investigated using multiple linear regression (MLR) analysis
(Statgraphics Centurion XV, Statpoint Inc., Herndon, VA, USA). Several MLR models were
constructed using combinations of membrane properties in order to discover the best set of
independent input variables. The following selection criteria were considered for including
membrane parameters in the MLR analysis:

Minimizing the number of inputs

Maximizing the R-Squared value (maximizing explanation of dependent variable


variation) and seeking a Mallows Cp index value near to the number of model inputs

Minimizing P-value for each variable in the model ( 0.05 indicating significance in
model 95% confidence level)

70
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Minimizing P-value for overall model ANOVA (0.05 desirable)

Minimizing Pearsons r correlation coefficient between independent variables (< 0.5


desirable)

MLR Model Input Variable Selection


Statgraphics Centurion XV provides a tool for simultaneously constructing and
evaluating several possible MLR models. Unfortunately, the paucity of exemplars available for
model construction limited the application of this tool, necessitating the manual construction and
comparison of clusters of three inputs at a time. Inputs considered for inclusion in the model
included:

average roughness (Ra in nm)

RMS roughness (Rq in nm)

zeta potential (/mV)

zeta potential slope (slope/mV(pH unit)-1)

contact angle ( in )

initial membrane water flux (Jo, Lm-2hr-1)

The MLR dependent variable was always the estimated normalized water flux at 5g m-2
bentonite clay load. Input parameter values that were used in the MLR analysis and
corresponding intrinsic fluxes are reported in Table 3.15.
Table 3.15
Multiple linear regression (MLR) model input parameters and values for describing
bentonite clay fouling.
Initial Water
Roughness
Contact Angle
J/Jo @ 5 g m-2
Flux
J
Membrane
o
Rq nm-1
, degrees
Clay Load
L/m2 hr
NF-270

3.8

57.4

21.34

0.903

ESNA1-LF

36.6

26.5

20.75

0.880

ESPA-2

38.8

37.4

20.64

0.908

SW30HR

84.5

33.9

6.82

0.970

MUNIRO-400

40.7

46.9

31.54

0.973

71
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Selected MLR Model Input Variables


The best MLR model found (the model with the largest R-squared and Mallows Cp
Index with closest match to the number of input variables) utilized RMS roughness, initial water
flux and contact angle, but not zeta potential. Table 3.16 shows the best MLR model parameters
and Figure 3.30 illustrates the model fit (Eq. 3.2) to the experiental data.

) (

) (

J Norm = 0.67233 + 2.29 10 3 J initial + 2.12 10 3 Rq + 3.00 10 3

(3.2)

where JNorm is the normalized permeate flux and Jinitial is the initial or starting permeate flux (L
m-2 hr-1).
Table 3.16
Constants, coefficients and statistics for the best MLR model: J/Jo (at 5g m-2 Load) =
0.67233+ (2.2910-3 Jinitial, L m-2hr-1) + (2.1210-3 Rq, nm) + (3.0010-3 , degrees).
Estimate
Std. Error
T-Statistic
P-Value
Constant

6.723E-01

2.47E-02

27.2511

0.0234

Jo, L m-2 hr-1 Slope

2.29E-03

5.24E-04

4.3748

0.1431

Rq, nm Slope

2.12E-03

1.86E-04

11.3951

0.0557

, degrees Slope

3.00E-03

3.81E-04

7.8996

0.0802

72
2012 Water Research Foundation. ALL RIGHTS RESERVED.

1.05

1
Predicted J/Jo

MUNIRO-400
SW30HR
0.95

ESPA-2

0.9

NF-270
ESNA1-LF

0.85
0.85

0.9

0.95

1.05

Actual J/J o

Figure 3.30. Results: Best MLR model prediction and UNR membrane data determined
from laboratory measurements. Statistics: R-Squared = 99.2736%; Adj. R-Squared =
97.0943%; Std. Error of Est. = .00718711; M.A.E = .00251155; P-Value = 0.1069.
Horizontal bars = 95% confidence interval for J/Jo @ 5 g m-2 clay load estimated from
laboratory data; vertical bars = MLR model standard error of the estimate.
Several of the independent variables are not completely independent, strictly speaking;
they shared some degree of cross-correlation. Typically, with multiple linear regressions it is
desirable to see the Pearson r between the variables < 0.5 (except for the constant). In this case,
initial flux and contact angle were both correlated with RMS roughness > 0.5; however, removal
of either term from the model resulted in serious reduction of the R-squared value.
Assessing the Predictive Ability of MLR Model Inputs by Iteratively Withholding a Test
Exemplar
In order to determine whether or not the MLR model approach has sufficient predictive
ability with this very small exemplar set and degree of input variable intercorrelation, five
additional MLR models were constructed, each built using just four of the five membrane
exemplars as a training set, with one membrane exemplar withheld and then predicted by the
resulting model (as a validation set). The performance of the withheld exemplar was then
compared for each of the MLR models to the results of the model constructed from the complete
data set.

73
2012 Water Research Foundation. ALL RIGHTS RESERVED.

If the original MLR model was not generalizing well (was over-fitting the small data set),
then removal of any one training exemplar would result in serious degradation of the model
ability to predict the validation exemplar. If, however, the MLR model was fitting on more
general principals, the model should have sufficient predictive ability to estimate the value of the
missing exemplar. With a very small training data set like this, it is additionally possible that
removal of one exemplar may result in loss of critical information that will seriously degrade
model performance; however, it was anticipated that at least one or more exemplars should be
redundant enough so loss of their information does not seriously affect model predictability, and
those exemplars when presented as validation examples ought to be predicted well.
Figure 3.31 shows the results of withholding individual exemplars by comparing the
membrane behavior predicted by the MLR model constructed from all the exemplars (filled
circles) with that where the chosen exemplar was held back from model training, and its behavior
predicted by an MLR model constructed with only the remaining four exemplars (open circles).
In the case of the ESPA-2 and ESNA1-LF membranes, omission of the exemplar from
the training data resulted in models still largely capable of predicting membrane behavior,
suggesting that this MLR approach was, overall, not just fitting noise but was in fact responding
to a relationship between the model independent variables and the membrane J/Jo. This lends
credence to the validity of the complete data set MLR model, though it is only statistically
significant at slightly less than the 90% confidence level. In the cases where the other membrane
exemplars were withheld, the deviations were more extreme, suggesting that training data
provided by the NF-270, SW30HR and MUNIRO-400 membranes were all critical to
constructing an MLR model capable of predicting membrane behavior.

74
2012 Water Research Foundation. ALL RIGHTS RESERVED.

1.05

Predicted J/J

S W30HR

MUNIRO-400

0.95
ES PA-2

0.9
NF-270

ES NA-1LF

0.85

0.8
0.8

0.85

0.9

0.95

1.05

Measured J/J o
Model with Exemplar Omitted

Model With All Exemplars

Figure 3.31. Comparison of the prediction of membrane behavior by the MLR model
constructed from all the exemplars (filled circles) with prediction of five MLR models that
were constructed using data missing each of the individual membranes indicated (open
circles). Horizontal bars = 95% confidence interval for J/Jo @ 5 g m-2 clay load estimated
from laboratory data; vertical bars = MLR model standard error of the estimate.
Assessing the Influence of MLR Model Input Parameters
It is possible to evaluate the influence that individual input parameters have on model
output by examination of specific component effects. In this approach, the relative influence that
each input has over the model output is assessed over the range of values of that particular input.
This analysis was performed for all of the model inputs, and the inputs were than classified based
on the magnitude of overall influence in the model (below).
RMS Roughness
RMS roughness was the most influential of the input variables, ranging from -0.0774
(NF-270) to +0.0935 (SW30HR), with a full range of 0.171. This input variable was positively
related to the predicted normalized water flux (J/Jo) after loading 5g m-2 of bentonite clay, which
suggests that more rugose membrane surfaces fared better in dealing with the clay foulant than
smoother ones. Though it has been often observed that smoother membranes have been related to
fouling resistance (at least biological fouling), the geometry of the clay particles may explain this
result. Clay particles are thin plate-like structures that tend to stack in close, organized layers on

75
2012 Water Research Foundation. ALL RIGHTS RESERVED.

smooth surfaces; on rougher surfaces it is possible that the stacking geometry becomes disrupted,
resulting in a cake layer that is less organized. Also, for larger clay particles, access to the
surface of more rugose membranes may be impared as the particles are only able to associate
with the tops of membrane features, with reduced opportunity for adhesion.
Initial Water Flux
Initial water flux (Jo, L m-2 hr-1) was the second most influential input parameter, with a
component effect ranging from -0.0297 (SW30HR) to +0.0271 (MUNIRO-400), with a full
range of 0.0971. Initial water flux was positively related to the predicted membrane normalized
water flux after loading 5 g of bentonite clay. At the moment, the exact mechanism by which
initial water flux may be related to bentonite clay membrane fouling is unclear. As with
roughness, this effect seems a little counterintuitive at the outset, because a higher water flux
suggests a higher force vector toward the membrane surface, which should cause suspended clay
particles to be pressed in tighter. However, higher membrane Jo might also increase the
membrane polarization layer, increasing the localized concentration of nonovalent and divalent
cations. Higher salt levels, especially higher monovalent salt levels, tend to favor increased
binding of clay particles. If the particles are arrayed on the rougher surfaces to begin with, the
tighter binding might prevent clay particles from slipping against one another, and maintain a
more open cake structure as more particles are deposited near the membrane surface.
Contact Angle
Contact angle was the least influencial of the model input variables; however as
previously explained its removal from the MLR model resulted in a significant degradation in
explanatory ability. The component effect of contact angle ranged from -0.0389 (ESNA1-LF) to
+.0522 (NF-270) with a total range of 0.0568. Contact angle was positively related to the
predicted membrane normalized water flux after loading 5 g of bentonite clay, meaning that
water flux through the clay cake tended to improve when the membrane surface was more
hydrophobic. Clay particles are microscopic to nanoscopic plate-like particles carrying a diffuse
negative charge, and effectively attract water molecules that can form a layer from 10 nm to up
to 40 nm over the particle surface. This effectively makes the colloidal clay particles quite
hydrophilic, and as such they would tend not to adhere well to the more hydrophobic membrane
surfaces (surfaces with higher contact angles). This would lead to formation of thinner and less
stable clay cake layers on these more hydrophobic membranes, which should result in higher
observed intrinsic water flux. This may well be the principal reason that contact angle appears
positively related to intrinsic water flux during fouling with bentonite clay.
Use of a Log-Decay Nonlinear Model coupled with MLR to Describe Bentonite Clay
Membrane Fouling in Terms of Membrane Properties
It was observed that, with more extensive bentonite clay exposure, the normalized
permeate flux rate for some membranes being fouled did not always decay to zero. Instead, for
several of the membranes it appeared that with increasing clay loading the normalized permeate
flux decayed in a log-like fashion to some plateau value, below which it did not decrease further
regardless of the additional clay loading. It might be postulated that for these membranes a
critical flux had been reached. It was therefore determined that the normalized permeate flux
(J/Jo) decline data for these membranes might be fitted with a classical flux decline model using

76
2012 Water Research Foundation. ALL RIGHTS RESERVED.

nonlinear methods in order to resolve the governing parameters for the observed membrane
fouling process. The log decay function that was determined to be most appropriate for this
approach is given in Equation 3.3.
3.3
where K is a decay rate coefficient whose value is hypothetically based on the impact of a given
foulant on the normalized permeate flux rate (J/Jo); Clay Load is the mass of clay impinged per
unit area of the membrane (g m-2); and J/Jo Plateau is the minimal value to which the normalized
permeate flux declined to following fouling by the clay particles.
Hypothetical Meaning of K
The K value is postulated to represent the immediate interaction between the clay
particles and the membrane surface that leads to a decline in the normalized permeate flux (J/Jo).
For a given amount of clay impinging on the membrane surface, the larger the value of K the
more rapid the decline in J/Jo. Put another way, system variables that increase the value of K
increase the severity of the initial decline in J/Jo (i.e., greater fouling), while parameters that
reduce K lessen the severity of the initial fouling. It is expected that the K value will be most
sensitive to interactions between the clean membrane surface and clay particles, since it
describes the kinetics of fouling where clay is presumably just beginning to coat the membrane
surface as opposed to building up on layers of previously deposited clay particles. Thus, K can be
a good measure of how resistant the membrane is to bentonite clay fouling. It is also presumed
that K will be especially sensitive to membrane surface parameters (e.g.,
hydrophobicity/hydrophilicity, surface roughness).
Hypothetical Meaning of J/Jo Plateau
The method by which the J/Jo Plateau value in Equation 3.3 was determined is illustrated in
Figure 3.31. This value hypothetically represents equilibrium between the addition of bentonite
clay particles to the cake structure on the membrane surface and the removal of clay particles
from the cake by cross flow shear. It may be expected at first to be less related to membrane
surface properties than to the properties of the cake itself, namely its thickness and permeability.
However, a complicating factor is that because of the nature of bentonite clay particles, the
overall structure of the cake as it accumulates on the membrane surface may be significantly
influenced by the geometry of the initial deposition layer. Additional factors that may affect cake
stability and water permeability may include cake compression, ionic effects of concentration
polarization (i.e., cake enhanced concentration polarization), as well as loss of net hydraulic
pressure caused by increased osmotic pressure at the membrane surface. Thus, many factors may
interactively influence the observed J/Jo Plateau value and as a result, it may relate to membrane
parameters in a very complex way. Nevertheless, the lower the observed value of the J/Jo Plateau
the more prone the membrane is to fouling by the clay.
Because membrane surface properties [e.g., roughness, contact angle with water
(hydrophobicity), and zeta potential] have been shown to determine the initial deposition and
attachment of foulant materials to membrane surfaces, at least to some extent, they are expected

77
2012 Water Research Foundation. ALL RIGHTS RESERVED.

to affect the values of both K and J/Jo Plateau. The relative impact that each of the aforementioned
membrane surface properties has on determining the values of K and J/Jo Plateau was investigated
by using multiple linear regression (MLR) methods to determine how well one or more of these
membrane properties can predict the observed variations in K and J/Jo Plateau for the test
membranes.

Figure 3.32. Illustration of the decline in normalized permeate flux rate (J/Jo) with
increasing clay load. The plateau value (J/Jo, Plateau) is taken as a measure of the severity of
the impact of bentonite clay on the permeate flux rate for a membrane and is determined
as the point where J/Jo becomes zero order with respect to clay loaded.
The specific objectives of this MLR analysis were to first determine values of K and of
J/Jo, Plateau for each of the test membranes by applying the laboratory data to Equation 3.3. Once
this was achieved, the observed variations in K and J/Jo, Plateau were related to the previously
described membrane parameters (initial water flux, roughness, contact angle and zeta potential)
using MLR methods.

78
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Using Nonlinear Regression to Determine K and J/Jo, Plateau for the Test Membranes
For each of the five membranes evaluated here, the membrane performance data
collected at the UNR laboratory were used to determine J/Jo as a function of clay impinging on
the membrane (g m-2), and these data fitted to Equation 3.3 using the Method of Marquardt
(Statgraphics Centurion VX, Statpoint, Inc., Herndon, VA). The normalized permeate flux
decline data that was collected for each membrane generally could be well-fitted to the log decay
equation (Eq. 3.3), with R-squared values ranging from 74 to 98%. Based on the fitted K values
for each membrane, it appears that the ESNA-1LF membrane suffered the most severe initial
fouling from the bentonite clay (i.e., the steepest initial slope in the plot of J/Jo as a function of
the clay loaded), followed by NF-270, ESPA-2 and SW30HR membranes. The MUNIRO-400
appeared the most resistant to initial fouling by the clay particles. This point may be best
illustrated through inspection of Figure 3.31 where the data for the NF-270 membrane shows a
much steeper slope in the initial portion of the curve (Figure 3.31a) when compared to that for
the MUNIRO-400 (Figure 3.31b).

Intrinsic Water Flux

a)

0.96
0.86

Intrinsic Water Flux

1.06

1.06

b)

0.96
0.86

0.76

0.76

0.66

0.66

0.56

0.56

0.46
0

20

40

60

80

ClayLoad

100 120

30

60

90

120

150

ClayLoad

Figure 3.33. Intrinsic water flux (J/Jo) as a function of the clay loaded (g m2) for the a) NF270 and b) MUNIRO-400 membranes. Lines show Eq. 3.3 fitted to the data by nonlinear
regression.
With regards to the resulting J/Jo, Plateau values for the different membranes tested here, it
appeared that the lowest value (i.e., the greatest overall reduction in plateau permeate flux) was
observed for the SW30HR membrane, followed by the MUNIRO-400 and ESPA-2/NF-270
membranes (Table 3.16). Note that the J/Jo, Plateau is a different measure of the impact of
membrane fouling on membrane performance than is the previously discussed K value. The J/Jo,
Plateau corresponds to the overall flux loss encountered by a membrane upon fouling, while the K
value corresponds to the rate of initial flux loss. The least overall reduction in plateau permeate
flux (i.e., the highest J/Jo, Plateau value) was observed for the ESNA-1LF membrane. The resulting
best fit values of K and J/Jo, Plateau for all of the test membranes, along with the 95% asymptotic
confidence intervals for the regression models, are summarized in Table 3.16.

79
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 3.17
Values of K and J/Jo, Plateau obtained by fitting normalized permeate flux (J/Jo) data and clay
loading (g m-2) data for each membrane to equation Eq. 3.3. R-squared values and 95%
asymptotic confidence intervals are reported for each nonlinear regression membrane
model.

J/Jo Plateau

+/-Asymptotic
95% Confidence
Interval

R2
J/Jo vs. Clay
Load g m-2

1.88E-02

0.490

0.040

98.5%

1.64E-01

7.98E-02

0.639

0.107

74.5%

ESPA-2

1.20E-01

2.12E-02

0.490

0.047

98.0%

SW30HR

1.58E-03

5.10E-04

-0.020

0.012

95.3%

MUNIRO-400

7.82E-03

9.74E-04

0.013

0.019

99.0%

+/- Asymptotic
95% Confidence
Interval

NF-270

1.24E-01

ESNA-1LF

Membrane

Relationship Observed between K and J/Jo Plateau and a Hypothetical Explanation


Interestingly, the J/Jo Plateau values are relatively strongly positively and linearly related to
the K values for all the membrane exemplars used in the study (Figure 3.32), (J/Jo Plateau =
3.96*K, R-squared = 99.64%). This suggests that the membranes that initially fouled the
quickest with bentonite clay also tended to maintain greater permeate flux once the clay cake
equilibrated on the membrane surface. This suggests a strong link between rate of loss of water
flux upon clay loading and the observed permeability of the clay cake at equilibrium.

Figure 3.34. Linearly regressed relationship between J/Jo and K. The intervals are the
asymptotic 95% confidence intervals from the individual nonlinear models.

80
2012 Water Research Foundation. ALL RIGHTS RESERVED.

The observed relationship between the K and J/Jo Plateau values could be most readily
explained by a fouling mechanism in which membranes with larger K values tend to more
rapidly accumulate thicker, but more water permeable cakes (clay platelets rapidly accumulate
but in highly haphazard structures), which when equilibrated with cross flow shear result in a
more permeable clay layer on the membrane (greater J/Jo Plateau values). Membranes that exhibit
lower K values, by contrast, would tend to accumulate platelets more slowly, but the cake formed
by these platelets would be less water permeable (the platelets would tend to closely align in
parallel). As a result of this sort of accumulation, the cake resulting at equilibrium would tend to
be far less permeable, even though it grew more slowly. The observed laboratory data readily fit
this bentonite clay fouling model, and it will be considered as a good basis for explaining
membrane surface interactions with bentonite clay.
Relating K and J/Jo, Plateau to Membrane Properties by Multiple Linear Regression
Since the membranes were all operated using a similar test configuration (swatch size,
feed spacer geometry, etc.) with a similar feed water matrix (including the clay concentration), at
a similar pressure and with similar crossflow velocity, it is not unreasonable to presume that
differences observed with respect to K and J/Jo Plateau were thus related solely to differences in
the membranes, and likely could be explained by one or more of the fundamental membrane
properties (initial water flux, roughness, contact angle and zeta potential) by MLR analysis using
the previously determined membrane properties as the independent variables. Due to the paucity
of exemplars for model construction, membrane properties chosen for inclusion in MLR analysis
were pre-screened using simple regression analyses with the K or J/Jo, Plateau value as the
dependent variable, and the best correlated membrane properties were subsequently selected for
MLR modeling (data not shown). The resulting membrane properties chosen for use in MLR
analysis are summarized in Table 3.16.
Table 3.18
Membrane properties selected as independent variables for MLR analyses of K and J/Jo,
Plateau . The zeta potential values were determined at pH=5.3, 2mM KCl saturated with
ambient CO2. Surface roughness statistics were collected by the OCWD, while zeta
potential and water contact angle values were collected by UNR. Note: Rq was slightly
cross-correlated with Jo (r=0.5206) and with (r=0.5084).
Initial Water
Flux, Jo
L m-2 hr-1

Rq
nm

mV

Degrees

NF-270

21.34

3.8 1.6

-29.4 3.3

57.4 4.7

ESNA-1LF

20.75

36.6 5.6

-29.3 2.5

26.5 5.6

ESPA-2

20.64

38.8 9.9

-21.5 1.9

37.4 4.0

SW30HR

6.82

84.5 8.5

-23.4 2.2

33.9 5.4

MUNIRO-400

31.54

40.7 4.3

-25.6 3.5

46.9 6.8

Membrane

81
2012 Water Research Foundation. ALL RIGHTS RESERVED.

K Related to Membrane Properties by MLR Analysis


Relationships between the K values shown in Table 3.16 and the membrane properties
reported in Table 3.16 were investigated using MLR analysis. The best model resulting from this
analysis included as independent variables membrane surface roughness (Rq/nm), initial water
flux (Jo) and contact angle [] (degrees) as shown in Equation 3.4 Note that the K value was
found to be insensitive to membrane zeta potential and was thus not included in Equation 3.4.
K = 0.517426 - (0.00368041 * R q ) - (0.0034793 * J o ) - (0.00527614 * )

3.4

The R-squared value for this MLR model (Eq. 3.4) was 98.82%, the adjusted R-squared
was 95.27% and the p-value was 0.1363 (model was significant only at an 86% confidence
level). The observed and predicted K values for each of the different membranes that were fouled
using the bentonite clay plotted as a function of one another are shown in Figure 3.34.

0.18

observed

0.15

ESPA-2
ESNA-1LF

0.12
0.09

NF-270

0.06

SW30HR

0.03
MUNIRO-400

0
0

0.03 0.06 0.09 0.12 0.15 0.18

predicted
Figure 3.35. Plot of the observed and predicted K values for each of the different
membranes fouled by bentonite clay. R-squared = 98.82%, adjusted R-squared = 95.27%,
p-value = 0.1363. The value of K determined in the nonlinear regression models for the
individual membranes loaded with bentonite clay could be predicted well from roughness,
initial water flux and contact angle data. Zeta potential data were not required for the
prediction (probably due to small variation amongst the test membranes).
Membrane zeta potential was not required as a model input to describe the observed
variations in K. This may not be so surprising, as at the pH with which the clay loading
experiments were performed the zeta potentials of these membranes did not vary over nearly as
great a range as the other membrane parameters (Table 3.16). Therefore, there was likely not
enough variation in the membrane zeta potential values to allow the model to differentiate the
observed permeate flux decline results.

82
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Analyses of component effects revealed that the K value was most influenced by
membrane surface roughness [Rq; -0.163 (SW30HR) to +0.134 (NF-270)], second most affected
by contact angle [, degrees; -0.0923 (NF-270) to +0.0670 (ESNA-1LF)] and least influenced by
the initial permeate flux rate [Jo; -0.0420 (MUNIRO-400) to +0.0444 (SW30HR)]. Membrane
surface roughness was negatively related to the K value, meaning that smoother membranes
experienced the most rapid decline in permeate flux. The contact angle with water () was also
negatively related to the K value, indicating that more hydrophobic membrane surfaces were
characterized by a more gradual declines in permeate flux upon fouling with the bentonite clay
(lower fouling rates). Finally, the initial membrane flux Jo was negatively related to K; indicating
membranes with higher initial flux rates exhibited greater flux decline rates.
Clearly membrane properties could be successfully related to the bentonite clay fouling
coefficient K, albeit at less than a 95% confidence level. Due to the potential for complex particle
interactions in bentonite clay cakes and a lack of additional data regarding the structure of the
cakes on the membrane surfaces, it is difficult to indicate with certainty what mechanism(s) may
be responsible for the relationship. However, if the relationship between K and J/Jo Plateau that
was hypothesized earlier is considered, then some explanations for the relationship between
membrane properties and the observed K values might include the following:

Surface Roughness: Studies with other classes of colloidal foulants (see e.g., Zhu and
Elimelech, 1997) have indicated that rougher membrane surfaces tend to be characterized
by greater foulant accumulation (enhanced fouling rates). In this case, the opposite
scenario appeared to occur. It is important to note that many earlier studies evaluated
fouling in terms of a net permeate flux loss rather than the initial decline in flux (i.e.,
what is discussed here in terms of a K value). Analysis of component effects shows that K
was increased by a decrease in membrane roughness (less loss of permeate water flux).
This relationship between membrane roughness and K is consistent with the findings of
the previous MLR analysis in this report in which J/Jo was modeled at a constant clay
loading rate of 5 g m-2. If, as hypothesized, larger K values correspond to more rapid
accumulation of a porous bentonite clay cake, then apparently the clay platelets more
rapidly attached to, and form organized structures on, the smoother membrane surfaces as
compared to the rougher ones. Bentonite clay platelets are unusual colloidal particles in
that they are extremely thin mineral plates as opposed to more globoid structures. The flat
surfaces are negatively charged, which would be repelled by the membrane, but the edges
actually have a positive charge, which could be attracted to it and allow the platelets to sit
edge-on to the membrane surface if is smooth enough to allow sufficient platelet
perimeter contact to hold the particle in place (this is also a mechanism by which platelets
can orient 90 degrees to each other in clay gels). Possibly with a smoother membrane,
platelets could better approach the surface edge-on; rougher surfaces might hamper the
ability of the thin plate-like clay platelets to bind, and the negative surface charges would
then more effectively repel them. An edge-on orientation would also favor the buildup of
a more randomly ordered (and water permeable) cake structure that would ultimately lead
to greater water permeability at equilibrium seen in the membranes with larger K values.

Contact Angle with Water (Hydrophobicity/Hydrophilicity): Analysis of component


effects suggested that contact angle was negatively related to K, suggesting that more
hydrophilic membranes (smaller contact angles with water) tended to favor more rapid

83
2012 Water Research Foundation. ALL RIGHTS RESERVED.

water flux loss for a given clay load. If the previous hypothesis regarding bentonite clay
cake formation and water permeability are considered, this suggests that more
hydrophilic surfaces tend to also be associated with a more rapid accumulation of clay
platelets. Hydrated bentonite clay platelets have a surface layer of water associated with
them, making them effectively very hydrophilic, and theoretically capable of
participating in hydrogen bonding as well. Thus, the clay platelets would tend to more
quickly and firmly associate with more hydrophilic surfaces and as a result may
accumulate more rapidly than on more hydrophobic membrane surfaces (surfaces with
larger contact angles) for a given clay loading in the feed. The hypothesis regarding clay
platelet accumulation at larger K values also suggests that the water permeability of the
rapidly formed cake should be greater (the cake composed of more randomly oriented
platelets), but unlike with surface roughness, its less clear how the nature of surface
binding alone can influence platelet orientation. One complicating factor is that
membrane roughness is slightly cross-correlated with contact angle, so that some of the
apparent relationship between contact angle and K may be mediated via covariance with
roughness. If, however, contact angle is omitted from the MLR model as an input, the Rsquared value is seriously degraded (data not shown).

Initial Permeate Flux Rate: Analysis of its component effect suggested that K is
positively influenced by lower initial permeate flux (Jo) values. As with contact angle, the
initial water flux was cross-correlated with membrane roughness (r = 0.5206), and so its
possible that some of the apparent influence of initial water flux is complicated by this
relationship. However, removal of water flux from the MLR model resulted in significant
degradation of R-squared (data not shown). If the hypothesis that higher K values
correspond with more rapid accumulation of a more porous cake is considered, this
would suggest that increased deposition of more randomly oriented clay platelets is
associated with lower initial membrane water flux. This relationship at first seems
counterintuitive. A previously made observation for negatively charged colloidal silica
particles suggests that accumulation on a membrane surface is a balance between
transport to the membrane surface by permeate flow (permeation drag) and repulsion at
the surface by the electric double layer (Zhu and Elimelech, 1997). In this case, it appears
that with higher Jo the resulting bentonite cake is either more permeable or it forms more
slowly. One possible explanation would be that if platelets are brought slowly to the
membrane by lower initial water fluxes, they can more readily associate edge-on and
form the foundation for the more porous cake structure. On the other hand, if they are
forced in more rapidly with a higher water flux, the tendency may be to turn face-on,
exposing the negatively charges platelet surface to the membrane, which doesnt as
strongly associate with the surface. Unfortunately, since the geometry of deposition of
bentonite platelets were not elucidated in this study, the apparent effects of water flux on
K will require future more scrutiny for elucidation.

84
2012 Water Research Foundation. ALL RIGHTS RESERVED.

J/Jo Plateau Related to Membrane Properties by MLR Analysis


As previously stated, the plateau value is zero order with respect to clay impinged on the
membrane surface, and appears most likely to represent a final equilibrium between
accumulation of clay at the membrane surface by convective transport and removal of clay from
the surface by cross flow hydrodynamic shear. Also, the water transport properties through the
clay layer at this equilibrium are expected to be important to determining the observed value of
J/Jo Plateau. Since it has been shown that J/Jo Plateau is essentially a linear function of K, it is
anticipated that all of the membrane properties identified by MLR analysis as predictive of K
will also predict J/Jo Plateau.
A similar MLR approach as was used to determine the relationship between membrane
properties and K was also performed with the J/Jo Plateau values identified in the nonlinear
regression membrane models. Since the fouling tests were done using similar test conditions
(cross-flow velocity, electrolyte concentrations/types, and solution pH) the principal differences
observed in J/Jo Plateau were presumed to likely be due to membrane properties alone (the
structure of the resulting cake influenced by the deposition of the first layer on the membrane
surface). As before, preliminary simple regression was used to screen the potential input
variables, resulting in those shown in Table 3.16. Following this, construction of MLR models
using these input variables revealed the best to be as follows:

J
= 2.0847 - (0.0150848 * R q ) - (0.0141576 * J o ) - (0.0212567 * )
J o , Plateau

3.5

The R-squared value for Equation 3.5. was 97.83%, the adjusted R-squared was 91.33%
and the p-value was 0.1843 (the model was significant only at an 82% confidence level). The
observed and predicted J/Jo Plateau values for each of the different membranes that were fouled
using the bentonite clay are shown Figure 3.31. As was anticipated since it was shown that K and
J/Jo Plateau appear to be mathematically linked, the same three membrane properties (Rq/nm
surface roughness, contact angle with water, and Jo) that best explained the variance in the K
value also best explained the variance in the observed J/Jo Plateau value. Zeta potential was again
not required for a good model fit, likely for the same reason that it did not vary enough to play a
role in clay interaction with the chosen test membranes.

85
2012 Water Research Foundation. ALL RIGHTS RESERVED.

observed

0.78
0.58

ESNA-1LF
ESPA-2

0.38

NF-270

0.18
-0.02
-0.02

MUNIRO-400
SW30HR

0.18

0.38

0.58

0.78

predicted
Figure 3.36. Plot of the observed and predicted J/Jo Plateau values for each of the different
membranes fouled by bentonite clay. R-squared = 97.83%, Adjusted R-squared = 91.33%,
p-value = 0.1843. The MLR model was able to describe the majority of the variance in J/Jo
Plateau using roughness, contact angle and initial water flux. As with K, zeta potential data
were not required to explain the variance observed in J/Jo Plateau.
As with K, component effect analysis revealed membrane surface roughness to be the
most influential membrane parameter related to J/Jo Plateau [-0.671 (SW30HR) to +0.545 (NF270)]. Surface roughness was negatively related to J/Jo Plateau (the rougher the membrane, the less
the J/Jo Plateau). Contact angle was the second most influential parameter [-0.376 (NF-270) to
+0.260 (ESNA-1LF)]. Contact angle with water was also negatively related to J/Jo Plateau,
suggesting that more hydrophilic membranes tended to favor higher water flux at clay loading
equilibrium than more hydrophobic ones. The initial permeate water flux (Jo) was observed to be
the least influential component [-0.175 (MUNIRO-400) to +0.177 (SW30HR)] and was also
negatively related to J/Jo Plateau, suggesting that membranes exhibiting higher initial water fluxes
tended to lose intrinsic flux the most when clay loading reached equilibrium, even though they
tended to lose permeate water flux more slowly (lower K values) when initially loaded.
The relationships observed here tend to support the previously discussed hypothesis that
membranes that were observed to lose water flux most rapidly were quickly accumulating a more
porous bentonite clay cake (due to a combination of lesser surface roughness, greater surface
hydrophilicity and lower initial water flux), and with continued clay load these membranes
formed more water porous clay cakes that, when equilibrated with membrane hydrodynamic
factors (cross flow shear) tended to exhibit greater water production (less fouling). On the other
hand, membranes that fouled more slowly with bentonite clay did so by more slowly
accumulating a less water permeable cake, so that when equilibrium was reached, the membrane
water flux was significantly reduced (to near zero in some cases). The hypothesized fouling
mechanisms are shown in Figure 3.36. The edges of the platelets (positively charged), are
attracted to the negatively-charged membrane surface. The flats of the platelets are negatively
charged and are repelled by the membrane surface. With less featured surfaces, the edges are

86
2012 Water Research Foundation. ALL RIGHTS RESERVED.

more likely in contact with the membrane surface, and the slower flux allows the platelets to
remain upright. The upright platelets result in a more open cake layer with greater water
permeability. With rougher surfaces, the likelihood of the edges remaining in contact with the
membrane surface decreases; in the case of platelets that do maintain edge contact, the higher
water flux likely pushes them over. When the flats of the platelet are in contact with the
membrane surface, a more planar cake layer accumulates and has less permeability. It's
simplistic, but it does fit the fouling kinetic observations.

Figure 3.37. Mechanistic explanation of fouling by bentonite clay.


As with the case in which a constant bentonite clay load was investigated, the kinetics describing
initial loss of membrane permeate water flux upon addition of a foulant material (K) and
equilibrium water flux J/Jo Plateau could be nearly completely explained in terms of the membrane
properties, again highlighting the importance of these parameters as predictors of membrane
performance.
Modeling Conclusions
Water treatment membranes can be fouled with a fairly wide variety of materials,
including biological substances (biofilms consisting of whole bacteria and bacterial extracellular
polysaccharides and nanodebris), dissolved and colloidal organic substances (e.g., humic and
fulvic acids) and mineral materials (colloidal clay nanoparticles and mineral precipitates). These
foulant materials associate with membrane surfaces by a variety of mechanisms including
hydrodynamic forces, friction, surface hydrophilic/hydrophobic interactions, surface ionic
charge, and van der Walls interactions. Bentonite clay represents a surrogate for one class of
membrane foulant. It is ideal for laboratory study as it is easy to handle and membrane exposure
is easy to quantify, also it can be used to demonstrate in principal how accumulation of foulant,
and the related loss of membrane performance, can be related to fundamental membrane surface
properties. This study does successfully demonstrate how, under a defined set of matrix

87
2012 Water Research Foundation. ALL RIGHTS RESERVED.

conditions, membrane surface properties can be used to predict membrane performance in


presence of a foulant material.

88
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 4
CONCLUSIONS

DEVELOPMENT OF STANDARD TECHNIQUES FOR CHARACTERIZING


MEMBRANE SURFACES
The following conclusions were reached regarding the development and application of
standard techniques for characterizing membrane surfaces, and specifically for executing surface
characterization using contact angle, surface roughness, and streaming potential measurements:

Contact mode produces more consistent roughness statistics when imaging the outside of
tubular structures. Therefore, when imaging hollow fiber membrane surfaces, whether the
concave or convex surface, the AFM should be operated in contact rather than tapping
mode.

A linear relationship exists between the root mean square and average roughness
statistics. Therefore, both parameters may be equally applied for drawing correlations
between membrane surface morphology and membrane fouling.

The inherent physical heterogeneity of membrane surfaces requires that greater than three
sites be imaged (i.e., their roughness measured) in order to develop an accurate
assessment of membrane surface roughness.

Surface roughness measurements using the AFM technique is a precise measure of


membrane surface morphology based on the reproducibility and repeatability of surface
roughness statistics between two laboratories. The precision of the measurement is,
however, dependent on the two laboratories using the standard technique.

Using the standard technique, the captive bubble method produces precise contact angle
results for membrane surfaces with a repeatability and reproducibility between any two
laboratories of approximately 5. This level of precision allows for equal comparison of
relative assessments of membrane surfaces with regards to their hydrophobicity (i.e., is
the membrane hydrophobic or hydrophilic). However, the current level of precision does
not lend itself well to the comparison of surface energy parameters calculated between
two laboratories using contact angles measured at each laboratory. Further study is
needed to determine how variations in contact angle results affect the magnitude and sign
of surface energy parameters determined through contact angle analysis to establish the
broader significance of these values to the membrane community (i.e., are these surface
energy parameters useful in membrane operations?).

The proposed standard method for determining the zeta potential of membrane surfaces
using streaming potential measurements produces relatively reproducible results
independent of the type of electrokinetic analyzer used. However, minor deviations from
the proposed standard method (e.g., differences in sample pretreatment) can result in
substantial errors.

89
2012 Water Research Foundation. ALL RIGHTS RESERVED.

The availability of characterization equipment and operator skill level requirements are
the greatest impediment to membrane characterization tests being performed by utilities.

From the perspective of utilities, characterization results, such as contact angle, surface
roughness, and streaming potential, are only useful during membrane selection and lose
their usefulness once membranes are put into service. A key challenge that was identified
in this regard is how to determine membrane characteristics without damaging the
membrane modules or elements (i.e., how to determine these characteristics without
conducting a destructive autopsy).

CORRELATION BETWEEN MEMBRANE PROPERTIES AND MEMBRANE


FOULING

The loss of normalized permeate water flux (J/Jo) when the five test membranes were
fouled by bentonite clay particles (representative of negatively charged plate-like
colloidal particles) at constant pressure could be modeled statistically using linear
regression methods; the resulting statistical models could then be used to express
normalized permeate water flux as a function of constant clay exposure. The decline in
normalized permeate flux as a function of clay loading followed a log-decay function.
Furthermore, the normalized permeate flux declined to some minimum, but steady-state,
value with respect to clay loading.

It was possible to construct a multiple linear regression (MLR) model capable of


describing >97% of the observed variance in normalized permeate water flux.
Additionally, a nonlinear regression analysis using a log-decay function found that
membrane performance could be successfully described in terms of a decay parameter
(K) and a minimum normalized permeate flux (J/Jo,Plateau) parameter. It was hypothesized
that the K value is related to the rate of accumulation of clay platelets at the membrane
surface, while the J/Jo Plateau value represents a state of equilibrium between clay
accumulation and removal by cross flow shearing. Because the K value describes the
initial fouling of the membrane surface it was anticipated to be significantly influenced
by membrane surface properties.

Membrane properties (Rq and contact angle with water) and the initial permeate water
flux alone were capable of describing the observed fouling behavior of the five different
membranes by bentonite clay. Zeta potential was not required; however, the variance of
zeta potential among the test membranes was the least of the membrane parameters, and
its likely that its influence on clay interactions with the test membranes was thus not
well represented in the study. A similar relationship was observed for the K value, which
was shown by MLR analysis to be largely a function of membrane surface roughness,
contact angle with water (hydrophobicity/hydrophilicity), and initial permeate water flux
(in order of greatest influence to least influence). This implies that membrane surface
roughness and hydrophobicity have a greater influence on bentonite clay cake
density/water permeability than does membrane surface charge.

It was not possible to eliminate significant inter-correlation of input variables and the
resulting model is only significant at ~90% confidence interval; however, validation tests

90
2012 Water Research Foundation. ALL RIGHTS RESERVED.

performed by iteratively withholding data for one membrane suggest this MLR approach
is capable of predicting membrane behavior.

Overall, better performance (higher observed normalized flux after 5 g/m2 bentonite
loading) was associated with greater RMS roughness and higher initial water flux and
greater contact angle (greater membrane hydrophobicity).

Although bentonite clay only represents one class of membrane foulant, the MLR
modeling effort illustrates how measurements of fundamental membrane surface
properties (roughness, hydrophobicity and surface charge) could be used to predict
membrane performance in the presence of any type of foulant whose
adhesion/accumulation depends on these membrane properties.

91
2012 Water Research Foundation. ALL RIGHTS RESERVED.

APPENDIX A
List of Utilities Using Membranes as a Treatment Technology (as of 2010)

92
2012 Water Research Foundation. ALL RIGHTS RESERVED.

MF Membrane Installations
source: Pall Installation
Start-up List
Jan2008.pdf

Pall Microza

Project

Water
Being
Treated

Capacity
(MGD)

Start-Up
Date

Fridley Membrane
Filtration Plant
Minneapolis, MN

Municipal WTP Microza


MF

Surface Water

95

TBD

Johnson County,
KS

Municipal WTP Microza


MF

Ground Water

31

Apr-09

City of North Bay,


Ontario

Municipal WTP
Microza MF

Lake Water

21

Dec. 2007

California Water Service Company


Bakersfield, CA

Municipal WTP Microza


MF

River Water

20

March 2003

Stones River WTP City of Murfreesboro, TN

Municipal WTP Microza


MF

Lake Water

20

Apr-08

City Of Clovis,
CA

Municipal WTP Microza


MF

River Water

15

Sept 2004

Saratoga County
Saratoga, NY

Municipal WTP Microza


MF

River Water

14

Feb-08

City of Olathe,
KS

Municipal WTP Microza


MF

Ground Water under the


influence

13

Dec 2005

City of Westminster
Westminster, CO

Municipal WTP Microza


MF

Lake Water

12 (Peak)

Oct. 2001

Tuscaloosa,
AL

Municipal WTP Microza


MF

Lake Water

12

Est. 6/08

Owner & Location

93
2012 Water Research Foundation. ALL RIGHTS RESERVED.

McCrosky Island WTP


Sevierville,TN

Municipal WTP Microza


MF

River Water

12

Dec-06

Yucaipa VWD
Yucaipa CA

Municipal WTP Microza


MF

Lake Water

12

Apr-07

Guadalupe-Blanco River Authority


GBRA/Western Canyon Water,
New Braunfels, TX

Municipal WTP Microza


MF

River Water

10

Dec 2005

City of Temple
Temple, TX

Municipal WTP Microza


MF

Raw Lake
Water

10

Oct. 2003

South Blount Utility District


Maryville, TN

Municipal WTP Microza


MF

Raw Lake
Water

10

June 2004

City Of Los Angeles Dept Water & Power


Encino, CA

Municipal WTP Microza


MF

Reservoir Water

9.80

Sept 2005

City of Abilene
Abilene, TX

Municipal WTP Microza


MF

Direct
Coagulated
Lake Water

8.0

June 2002

Travis County # 17
Austin, TX

Municipal WTP Microza


MF

Lake Water

8.0

June 2002

San Patricio Municipal W. D.


Ingleside, TX

Municipal WTP Microza


MF

Coagulated,
Clarified
Surface Water

7.8

Jan. 2000

Brushy Creek Municipal Utility District,


Bushy Creek, TX

Municipal WTP Microza


MF

Lake Water

6.00

Oct 2005

94
2012 Water Research Foundation. ALL RIGHTS RESERVED.

City of St. Helens,


OR

Municipal WTP Microza


MF

River/Creek
Water

5.95

March 2006

City of Salmon
ID

Municipal WTP Microza


MF

River/Creek
Water

5.2

Feb 2006

Upper Eagle Regional Water


Avon, CO

Municipal WTP Microza


MF

River / Stream

5.0

Nov. 2002

Town Of Orleans
MA

Municipal WTP Microza


MF

Ground WaterIron &


manganese Removal

4.5

April 2005

Alamitos Barrier,
CA

Municipal WTP Microza


MF

Secondary
Effluent
Pre-ro

3.5

June 2001

City of Brady
TX

Municipal WTP Microza


MF

Surface Water

3.30

Oct 2005

Town of Petrolia
ONT

Municipal WTP Microza


MF

Raw Lake
Water

3.17

Feb 2005

Sonoma County Water Agency


Santa Rosa, CA

Municipal WTP Microza


MF

Secondary
Effluent

3.0

May 2002

Travis County # 18
Austin, TX

Municipal WTP Microza


MF

Coagulated
Settled Water

3.0

Nov. 2003

Washington City WTP


Washington City,
UT

Municipal WTP Microza


MF

Reservoir Water

3.0

July 2003

Amador Water Agency, Buckhorn WTP


Pioneer, CA

Municipal WTP Microza


MF

Surface Water

3.0

April 2005

95
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Holliday Water Co.


Holladay, UT

Municipal WTP Microza


MF

Spring Water

2.5

March 2002

Town of Newton
Newton, NJ

Municipal WTP Microza


MF

Lake Water

2.25

March 2005

Lost Creek
UT

Municipal WTP Microza


MF

Ground Water

2.0

Jan 2005

Town of Byrdstown
Byrdstown, TN

Municipal WTP Microza


MF

Flocculated
Raw Surface
Water

2.0

December
2005

Garden City, UT

Municipal WTP Microza


MF

GWUDI

2.0

November2006

Fountain Hills Sanitation District


Fountain Hills, AZ

Municipal WTP Microza


MF

Tertiary Effluent for


Reuse

2.0

March 2001

City of Chandler
Chandler, AZ

Municipal WTP Microza


MF

Semiconductor
Wastewater
Reclamation

1.75

Nov. 97

Solano Irrig. Distr.


Vacaville, CA

Municipal WTP Microza


MF

River
Water

1.4

April 2001

City of Flowery Branch,


GA

Municipal WWTP
Microza MF

Secondary
Effluent

1.3

Nov 2005

Town Of Whitecourt,
AB

Municipal WTP Microza


MF

Coagulated Settled
Surface Water

1.3

Jan 2005

Crested Butte WTP


Crested Butte, CO

Municipal WTP Microza


MF

Creek Water

1.25

March 2002

96
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Chalmette Refining LLC,


Chalmette, LA

Industrial
Microza MF

Mustang Utility District.


CO

River Water

1.15

Feb 2005

Municipal WTP Microza


MF

Direct
Coagulated Raw
Water

1.15

Dec 2004

Lee County
VA

Municipal WTP Microza


MF

Surface Water

1.04

Feb 2004

City of Parsons
Parsons, KS

Municipal WTP Microza


MF

River Water

1.0

May 2003

Bedford Hills Correctional Facility


Bedford Hills, NY

Municipal WTP Microza


MF

Wastewater

1.0

Sept. 2002

City of San Marcos


San Marcos, TX

Municipal WTP Septra


CB

Groundwater Under the


Influence

1.0

July 2002

Lakeview Utility District,


Rogersville, TN

Municipal WTP Microza


MF

Groundwater
Under the
Influence

1.0

December
2005

Sherkston Shores Resort


Pt. Colborne, ONT

Municipal WTP Microza


MF

Surface Water

1.0

Apr. 2003

Rutherford County
Murfreesboro, TN

Municipal WWTP
Microza MF

Backwash
Water

1.0

Dec 2005

City of Caney
KS

Municipal WTP Microza


MF

Surface Water

0.72

Apr 2005

Hantsport
Nova Scotia

Municipal WTP
Microza MF

Coagulated
Water

0.72

March 2004

Upper Surface Creek,


CO

Municipal WTP Microza


MF

0.72

Jan 2005

97
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Forestville County Sanitation District


Forestville, CA

Municipal WTP Microza


MF

Secondary
Effluent

0.58

Sept. 2003

Hoopa Valley,
CA

Municipal WTP Microza


MF

Surface Water

0.57

May 2005

Bedford PSA,
VA

Municipal WTP Microza


MF

Ground water

0.548

Aug 2005

Raccoon Creek,
PA

Municipal WTP
Microza MF

Lake Water

0.5

Dec 2004

California Water Service Company


Kernville, CA

Municipal WTP Microza


MF

Surface Water

0.5

April 2002

Wheelabrator
Pinellas, FL

Industrial Waste
Microza MF

Secondary
Effluent

0.5

June 2002

Youngs River /Lewis & Clark District


Astoria, OR

Municipal WTP Microza


MF

River Water

0.5

Nov. 2001

Town of Basalt WTP


Basalt, CO

Municipal WTP Microza


MF

Spring Water

0.5

May 2002

City of Red Boiling Springs


TN

Municipal WTP Microza


MF

Ground Water

0.5

Jan 2005

BoWater WTP
Calhoun TN

Plant
Drinkwater
Microza MF

Surface

0.5

Dec 2004

Village of Alix
Alix, AB

Municipal WTP Microza


MF

River Water

0.5

Oct. 2002

City of Meeteetse
Meeteetse, WY

Municipal WTP Microza


MF

Reservoir Water

0.3 / 0.6

April 2001

98
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Village of Hobart
Hobart, NY

Municipal WTP Microza


MF

Wastewater

0.47

May 2002

Bruce Mines WTP


Bruce Mines, ON

Municipal WTP Microza


MF

Lake Water

0.44

March 2003

Village of Groton
NY

Municipal WTP
Microza MF

Ground Water

0.43

Jan 2006

Stoney Creek Municipal Authority


Basye, VA

Municipal WTP Microza


MF

Ground Water Under the


Influence

0.36

Nov. 2002

Coast Springs
CA

Municipal WTP Microza


MF

Ground Water

0.36

June 2004

Manitoba Hydro
ONT

Microza MF

Surface Water

0.36

Feb 2004

No. Slope Borough


Wainwright, AK

Municipal WTP Microza


MF

Lake Water

0.35

June 1999

No. Slope Borough


Nuiqsut, AK

Municipal WTP Microza


MF

Lake Water

0.35

Aug. 2001

No. Slope Borough


Point Hope, AK

Municipal WTP Microza


MF

Lake Water

0.35

June 1999

Westover
Westover, PA

Municipal WTP Microza


MF

Surface Water

0.25

April 2003

Mt. Rainier National Park Srvc.


Ashford, WA

Municipal WTP Microza


MF

Surface Water

0.144

April 2003

Questa La Honda,
CA

Municipal WTP Microza


MF

Surface Water

0.14

Feb 2005

No. Slope Borough


Kaktuvik, AK

Municipal WTP Microza


MF

Lake Water

0.12

Aug. 2003

99
2012 Water Research Foundation. ALL RIGHTS RESERVED.

No. Slope Borough


Point Lay, AK

Municipal WTP Microza


MF

Lake Water

0.12

Aug. 2000

No. Slope Borough


Atqasuk, AK

Municipal WTP Microza


MF

Lake Water

0.12

Aug. 2001

Hot Springs,
AR

Municipal WTP Microza


MF

Spring Water for


Bottling

0.05

Aug 2005

Masonville
NY

Municipal WTP Microza


MF

Secondary
Effluent

0.01

Aug 2004

Hite Marina
Lake Powell, UT

Municipal WTP Microza


MF

Surface Water

0.1

April 2002

Oregon Parks & Recreation Dept.


Beverly Beach, OR

Municipal WTP Microza


MF

Creek Water

0.1

Oct. 1999

Oregon Parks
Bullards Beach, OR

Municipal WTP Microza


MF

Creek Water

0.1

Jan. 2000

Russian River Util.


Forestville, CA

Municipal WTP Microza


MF

Surface Water

0.1

Nov. 2000

Jim BridgerPoint of Rocks, WY

Municipal WTP Microza


MF

River Water

0.1

Feb. 2002

Kennecott Greens Creek Mining Co.


Anchorage, AK

Municipal WTP Microza


MF

Drinking Water

0.04

Nov. 2002

Burleigh Island Lodge


Burleigh Falls, ON

Municipal WTP Microza


MF

Lake Water

0.04

June 2002

Camp LMan Achai WWTP


NY

Municipal WTP Microza


MF

Secondary
Effluent

0.00080

June 2004

100
2012 Water Research Foundation. ALL RIGHTS RESERVED.

SIEMENS Water Technologies Corp.


GLOBAL MEMCOR PRODUCTS
INSTALLATION LIST
Project Name

Country

City

State

Feed Source

BlueScope Steel - No. 2 Blower Station

AUS

NSW

Town Mains

BlueScope Steel - No. 2 Blower Station

AUS

NSW

Town Mains

Cadbury Schweppes, Liverpool

AUS

NSW

Town Mains

Cerestar

FRA

ROW

Surface Water

Crackenback

AUS

NSW

Surface Water

Dunedin

NZL

Dunedin

Durango LaPlata Airport

USA

Durango

CO

Surface Water

Estero Mutual Water Co. (Dillon Beach)

USA

Dillon Beach, CA

CA

Surface Water

Green Spring Valley

USA

WI

Surface Water

Kansas City Power & Light

USA

Clinton, MO

MO

Surface Water

Kwinana Water Reclamation Plant

AUS

Kwinana

WA

Secondary Effluent

LG Phillips P6

SKO

Paju

Metropolitan Water Dsitrict

USA

Yuma

AZ

Surface Water

Oklahoma Gas & Electric

USA

Konawa, OK

OK

Surface Water

Sears Point Raceway (Infineon Raceway)

USA

Sonoma , CA

CA

Surface Water

101
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Surface Water

ROW

Industrial Process

Severn Trent, Homesford WTW

UK

Derby

ROW

Ground Water

ROW
QLD
CO
ROW
CA
QLD

Ground Water
Industrial Process
Surface Water
Secondary Effluent
Surface Water
Secondary Effluent

Southern Water, Arundel WTW


Tarong PS
Walden
Xiang Yang Lu
Bolinas Community Public Utility District
Bundamba 1A

UK
AUS
USA
CHN
USA
AUS

Arundel
Kingaroy
Walden
Tianjin
Bolinas, CA
Brisbane

Butano Canyon Water Company

USA

Pescadero, CA

CA

Surface Water

California Department of Parks (Gaviota State


Park)

USA

Goleta, CA

CA

Surface Water

Carmichael Water District

USA

Carmichael

CA

Surface Water

Caseville WTP, Village of

USA

Caseville, MI

MI

Surface Water

Coca Cola - West Memphis

USA

AR

AR

Cucamonga Water District

USA

Rancho Cucamonga, CA

CA

Surface Water

Duckmaloi

AUS

NSW

Surface Water

Gore Valley (Eagle River)

USA

Vail, CO

CO

Surface Water

Hasting Council - Comboyne WTP

AUS

Comboyne

NSW

Surface Water

Hasting Council - Long FLat WTP

AUS

Long Flat

NSW

Surface Water

Hasting Council - Port Macquarie

AUS

Port Macqurie

NSW

Secondary Effluent

Hasting Council - Telegraph Point WTP

AUS

Telegraph Point

NSW

Surface Water

102
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Hasting Council - Wauchope WTP

AUS

Wauchope

NSW

Surface Water

Idaho Springs, City of

USA

Idaho Springs, CO

CO

Surface Water

Isle Royale National Park

USA

Mott Island

MI

Surface Water

Linwood Metropolitan Water District

USA

Linwood, MI

MI

Surface Water

Los Alamos Radioactive Treatment Facility

USA

Los Alamos, NM

NM

Other

Lower Brule Sioux Tribe

USA

Lower Brule, SD

SD

Surface Water

Mackinac Island

USA

Mackinac Island, MI

MI

Surface Water

Marquette, City of

USA

Marquette, MI

MI

Surface Water

Metropolitan Water District LaVerne

USA

La Verne, CA

CA

Surface Water

Oamaru

NZL

Oamaru

Portsmouth Water, Farlington WTW

UK

Portsmouth

ROW

Surface Water

Portsmouth Water, Farlington WTW

UK

Portsmouth

ROW

Surface Water

Portsmouth Water, Lovedean WTW

UK

Portsmouth

ROW

Ground Water

Portsmouth Water, Lovedean WTW

UK

Portsmouth

ROW

Ground Water

River Mountain WTF

USA

Henderson, NV

NV

Surface Water

Rojana Power

THL

Scotish Water, Invercannie WTW

UK

Aberdeen

ROW

Surface Water

Thames Water, Addington WTW


Willaura

UK
AUS

Croydon

ROW
VIC

Ground Water
Surface Water

Carbondale, Town of

USA

Carbondale

CO

Ground Water

Surface Water

ROW

103
2012 Water Research Foundation. ALL RIGHTS RESERVED.

County of Maui (Olinda WTF)

USA

Maui

HI

Surface Water

Algonquin, Village of
Carter Lake

USA
USA

Algonquin, IL
Berthoud, CO

IL
CO

Ground Water
Surface Water

Erie, Town of

USA

Erie, CO

CO

Surface Water

Kennewick WTP

USA

Kennewick, WA

WA

Surface Water

Manitowoc Public Utilities

USA

Manitowoc, WI

WI

Surface Water

St. Johns - New Foundland

CAN

New Foundland

NF

Carmichael Water District

USA

Carmichael

CA

Surface Water

Delta Centrifical

USA

Temple

TX

Surface Water

Indian Hills

USA

Indian Hills

CO

Surface Water

Kaneka

BEL

ROW

Surface Water

Newell-Warnock, CO

USA

Loveland

CO

Surface Water

Point Cabrillio

USA

Ft Brag, CA

CA

Surface Water

Ravenna (Greenwood Village)

USA

Ravenna

CO

Surface Water

Colorado City

USA

Longmont, CO

CO

Surface Water

DeBeque, Town of

USA

DeBeque, CO

CO

Surface Water

Eatonville, Town of
Homestead

USA
USA

Eatonville
Hot Springs, VA

WA
VA

Surface Water
Ground Water

Lake Santee WTP

USA

Lake Santee

IN

Surface Water

Oak Creek WTP

USA

Oak Creek

CO

Surface Water

104
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Odana Hills
Parachute

USA
USA

Madison
Parachute, CO

WI
CO

Surface Water
Surface Water

Pine Brook Water District

USA

Boulder, CO

CO

Surface Water

Silt, Town of

USA

Silt

CO

Surface Water

Albany-Millersburg
Amcor

USA
PHL

Albany/ Millersburg, OR

Amkor Phillipines

PHL

Andes WWTP

USA

Andes, NY

Appalachian State Univ.


Beenliegh
Blackmores
Boddington Gold
Bogong

USA
AUS
AUS
AUS
AUS

Boone

Brewster (I-684 Restop)

USA

Brewster, NY

NY

Brewster Heights

USA

Brewster, NY

NY

Briarcliff, Village of

USA

Briarcliff, TX

TX

Surface Water

Bristol Water, Banwell


Bristol Water, Banwell

UK
UK

Weston Super Mare


Weston Super Mare

ROW
ROW

Surface Water
Surface Water

Byesville WTP

USA

Byesville, OH

OH

Ground Water

Cadbury Schweppes, Brisbane

AUS

Camp Fitch
Clear Spring WTP
Coca Cola - Brisbane
Columbia WTP

USA
USA
AUS
USA

North Springfield
MD

Cookson WTP

USA

Cookson, OK

Corrumbin Eco Village

AUS

Corrumbin

Boise

105
2012 Water Research Foundation. ALL RIGHTS RESERVED.

OR
ROW

Surface Water
Industrial Process

ROW

Industrial Process

NY
NC
QLD
NSW
WA
NSW

Surface Water
Surface Water
Ground Water
Surface Water

QLD

Town Mains

PA
MD
QLD
ID

Ground Water
Town Mains

OK

Surface Water

QLD

Secondary Effluent

Cortez, City of
Cottees

USA
AUS

Cortez
Liverpool

CO
NSW

Surface Water
Town Mains

Crosshouse Hospital, Glasgow

UK

Glasgow

ROW

Potable Water

Crow Creek Casino WWTP


Crusta
Detour, Village of
Dunedin Airport

USA
AUS
USA
NZL

Canyonville, OR

OR
NSW
MI

Surface Water
Industrial Process
Surface Water
Secondary Effluent

Environmental Management

USA

Lemont, IL

IL

Surface Water

Fort Bidwell

USA

Ft. Bidwell, CA

CA

Surface Water

Fort Ross
Freshpac
Fuxin Jinshan
Geochang

USA
AUS
CHN
SKO

Fort Ross, CA
Tatura

CA
VIC
ROW
ROW

Hallsdale-Powell Utility District

USA

Knoxville, TN

TN

Surface Water

Healdsburg

USA

Healdsburg

CA

Surface Water

Huron Regional Water Authority

USA

Port Austin, MI

MI

Surface Water

Indiana County
InVista - DuPont
Joines Rd

USA
USA
USA

Indiana, PA
TX
Brownsville, TX

PA
TX
TX

Industrial Process
Surface Water

June Lake WTP

USA

June Lake

CA

Surface Water

Lake Alpine Water Company

USA

Lake Alpine

CA

Surface Water

Latrobe

USA

Latrobe

PA

Laura WTP, Cook Shire

AUS

Laura

LeRoy WTP

USA

Leroy, IL

DeTour, MI
Dunedin

106
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Surface Water
Surface Water
Secondary Effluent
Surface Water

QLD

Surface Water

IL

Ground Water

Lexington, Village of

USA

Lexington, MI

LG Micron P3

SKO

ROW

Industrial Process

Lochgilphead Hospital, Lochgilphead

UK

ROW

Potable water

Logan Todd Regional Water Commission

USA

Russellville, KY

KY

Surface Water

London Power Company, Cottam Power Station

UK

Notts

ROW

Surface Water

Loxley TRN
Madisonburg

THL
USA

Madisonburg

ROW
PA

Marulan WTP

AUS

Marulan

NSW

Surface Water

McConnellsburg
Monroe WTP
Mt Piper
Neskowin
Nestle Water

USA
USA
AUS
USA
USA

McConnellsburg
Monroe, OR

PA
OR
NSW
OR
FL

Surface Water
Industrial Process
Surface Water
Surface Water

Nestle Water

USA

Red Boiling Springs

TN

Industrial Process

Noramco

USA

Ridley Creek

PA

North Clackamas County (Sunrise)

USA

Sunrise, OR

OR

Surface Water

Nugans

AUS

NSW

Surface Water

Possum Valley Municipal Water

USA

Bendersville

PA

Pottawatomie, City of

USA

Pottawatomie

OK

Surface Water

Rix/Munemasa Shuzo
Rix/Munemasa Shuzo
Scottish Water, Invercannie WTW

JPN
JPN
UK

Aberdeen

ROW
ROW
ROW

Surface Water

Severn Trent, Homesford WTW

UK

Derby

ROW

Ground Water

Neskowin
Lee

107
2012 Water Research Foundation. ALL RIGHTS RESERVED.

MI

Surface Water

Siemens

THL

ROW

Siemens

THL

ROW

Simms Mesa

USA

Navajo Dam, NM

NM

St. Johns - New Foundland

CAN

New Foundland

NF

Standish, City of

USA

Standish, MI

MI

Tabulam 2nd chance facility

AUS

NSW

Taiyuan PS No. 1
Thames Water, Taplow WTW
Thames Water, Taplow WTW

CHN
UK
UK

London
London

ROW
ROW
ROW

Secondary Effluent
Ground Water
Ground Water

Three Valleys Water, Batchworth WTW

UK

London

ROW

Ground Water

Three Valleys Water, Batchworth WTW

UK

London

ROW

Ground Water

Three Valleys Water, Chertsey WTW

UK

London

ROW

Ground Water

Three Valleys Water, West Hyde/Mill End WTW

UK

London

ROW

Ground Water

Three Valleys Water, West Hyde/Mill End WTW

UK

London

ROW

Ground Water

Toshiba/Toyobo/Yamanaka

JPN

ROW

Tyers
Victor

AUS
USA

Victor

VIC
CO

Surface Water
Surface Water

Waurika, City of

USA

Waurika

OK

Surface Water

WEPCO - Point Beach

USA

Point Beach, WI

WI

Surface Water

Wessex Water, Tucking Mill

UK

Bath

ROW

Surface Water

Wolf Creek

USA

OR

OR

Surface Water

108
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Surface Water

Surface Water

Ameren Coffeen
Arnotts
Barwon Heads
Barwon Water
Blackberry WWTP
Bootawa Dam
Bundamba 1B
Central Highlands Water
Dagang Oil
DaHua
Dongshen
Edinburg, Town of
Encampment
Ennis Co Council, Ennis
Fairfield
Fargo
Flying J Truck Stop
Gascoyne Junction
Gibson Island
Guofeng North
Guofeng South
IBM Silicon
LPE
Milton
Nice Mutual Water
Prairie Pride

USA
AUS
AUS
AUS
USA
AUS
AUS
AUS
CHN
CHN
CHN
USA
USA
IRE
USA
USA
USA
AUS
AUS
CHN
CHN
CHN
THL
NZL
USA
USA

Rend Lake Conservancy District

USA

Benton, IL

Scottish Water, Badentinan


Scottish Water, Fort William
Scottish Water, Howden

UK
UK
UK

Scottish Water, Kirkmichael


Severn Trent, Washgreen WTW

Sydney

Brisbane
Tianjin

Edinburg, VA
Encampment, WY
Ennis
Fairfield, IL
Fargo, ND
Whiteford, MI
Brisbane
Guofeng
Guofeng
Shanghai
Milton
Nice, CA
Deerfield, MO

IL
NSW
VIC
VIC

Surface Water
Industrial Process
Surface Water
Secondary Effluent

NSW
QLD
VIC
ROW
ROW
ROW
VA
WY
ROW
IL
ND
MI
WA
QLD
ROW
ROW
ROW
ROW

Surface Water
Secondary Effluent
Industrial Process
Sea Water
Secondary Effluent
Ground Water
Surface Water
Surface Water
Surface Water
Surface Water
Ground Water
Ground Water
Secondary Effluent
Secondary Effluent
Secondary Effluent

CA
MO

Surface Water

IL

Surface Water

Elgin
Fort William
Melrose

ROW
ROW
ROW

Ground Water
Ground Water
Ground Water

UK

Perth

ROW

Surface Water

UK

Derby

ROW

Ground Water

109
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Sime Oleander (Coca Cola)


Sourdough Home Owners
Stobhill Hospital, Glasgow
Thames Water, Deephams
Ugum
Victoria Hospital, Glasgow
West Cumberland Utility
Yorktown Heights

MLA
USA
UK
UK
USA
UK
USA
USA

Ajax Chemicals, Smithfield

AUS

Auspine
Autaugaville

AUS
CAN

Cadbury Schweppes, Perth

AUS

Cadomin WTP (Yellow Head County)

CAN

Carlton United Breweries

AUS

Chatham-Kent WTP (Ontario)

CAN

Chrysler De Mexico
Exxon, Olympia
Gordon, City of
Hatillo-Camuy WTP
IDS
In Ha University

MEX
RUS
USA
PTR
SKO
SKO

Gordon, TX
Hato Rey

La Laguna

MEX

La Laguna

Mt. Pleasant

AUS

Sandpoint, ID
Glasgow
London
Glasgow
Crossville, TN

ROW
ID
ROW
ROW
ROW
TN
NY

Surface Water
Town Mains
Secondary Effluent
Surface Water
Town Mains

NSW
Industrial Process
WA

Alberta

Town Mains

AB
VIC

Erie Beach

ON
ROW
ROW
TX

Surface Water
Surface Water

ROW
ROW

110
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CA

VIC

Surface Water

Nestle Water

CAN

Gelph, Ontario

PetroBras- Reman

BRZ

Brazil

ROW

Quatar Petroleum
Sana Muerto

QTR
PTR

Sana Muerto

ROW
ROW

Seoul Metropolitan Government

SKO

ROW

Timbercorp

AUS

VIC

Surface Water

Wagga Wagga Base Hospital

AUS

NSW

NSW

Town Mains

Wasagaming

CAN

Manitoba

Wyndham
Xiao Hong Men

AUS
CHN

111
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ON

Surface Water

MB
WA
ROW

Secondary Effluent

UF Membrane Installations
Norit X-Flow
Facility
CHMFP

Commission
Date
2005

Municipal Effluent to Reuse

100

2004

X-Flow S225 UFCM5


X-Flow S225 UFCM5

Conventional Backwash to Drinking water


Conventional Backwash to Drinking water
Conventional Backwash to Drinking water
Conventional Backwash to Drinking water

n/a
n/a
0.36
0.25

1996
2000
2003
2001

X-Flow S225 UFCM5

Municipal Effluent to reuse

5.6

1999

Membrane Type
X-Flow S225 UFCM5

Treatment type
Surface Water to Drinking water

Sulaibiya, Kuwait

X-Flow S225 UFCM5

Eindhoven, Netherland
Macharen, Netherland
Helmond, Netherland
Annen, Netherland

X-Flow S225 UFCM5

Windhoek, Namibia

Aalsterweg

Goreangab

Flow Rate
(MGD)
70

Location
Minneapolis, USA

112
2012 Water Research Foundation. ALL RIGHTS RESERVED.

NF Membrane Installations
Dow Filmtec NF-270
Facility

Membrane Type

Treatment type

Flow Rate (MGD)

Commission Date

Overton Beach Marina, Nevada

Filmtec NF 270-400

Lofsdalen, Sweden

Filmtec NF 270-401

Treatment of lake water

0.19

1998

Filmtec NF 270

Concentrate recovery

0.40

2007

Membrane Type

Treatment type

Flow Rate (MGD)

Commission Date

Hollywood, Florida

ESNA1-LF

Desalination of Brackish water

18

1996

Ft. Lauderdale, FL*

ESPA4 /ESPA4 /
ESNA1-LF2

Desalination of Brackish water

12

2006

Boca Raton, Florida

ESNA

Desalination of Brackish water

40

Deerfield Beach, Florida

ESNA

Desalination of Brackish water

120

Pompano Beach, Florida

ESPA1/ESNA1 365

Desalination of Brackish water

10

2002

North Miami Beach Florida, FL

ESNA1-FL

Desalination of Brackish water

2005

Orange Tree Utilities, FL

ESNA1-LF

Desalination of Brackish water

ESNA1-LF2/ESNA1LF3

Desalination of Brackish water

0.35

ESNA1-LF

Desalination of Brackish water

40.9

HYDRACoRe

Desalination of Brackish water

2002

ESNA1

Desalination of Brackish water

7.35

2003

ESNA1/ ESPA3

Desalination of Brackish water

0.86

2002

Collier County, FL

ESNA

Desalination of Brackish water

12

1993

Dunedin, FL

ESNA

Desalination of Brackish water

1992

ESNA/ESPA

Desalination of Brackish water

12

1990

ESNA

Desalination of Brackish water

1980

Irvine Ranch Water District

0.1

Hydranautics ESNA1-LF
Facility

City of Boca Raton, FL


Deerfield Beach, FL
Irvine Ranch, CA
Glendale, Maryland
Moody Air force Base-Georgia

Fort Myers, FL
St. Lucie West, FL

113
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2005
2004/2005

RO Membrane Installations
Hydranautics ESPA
Facility
Ulu Pandan, Singapore
Orange County Water District, CA
Shedgum/Saramco KSA
Libya
Wichita Falls, Texas
Tropical Farms WTP, Stuart FL
North Miami Beach, FL
West Basin Municipal Water District, California
Orange County Water District, CA
Southmost.Regional, Tx (Brownsville)
Alameda County Water District, CA
Tampa Bay Water, FL
West Basin
Glouster County, VA
Aledo, IL
Jupiter, Florida
Mexicana de Cobre, Mexico
Pt St. Lucie

Membrane Type
ESPA2+
ESPA2+
ESPA2
ESPA2
ESPA2/ESPA4
ESPA2
ESPA2/ ESPA1
ESPA2/ ESPA1
ESPA2
ESPA2
ESPA1/ESPA2
ESPA2/SWC4
ESPA2
CPA3/ESPA2
ESPA2/ESPA1
ESPA2
ESPA2
ESPA2

Treatment type

Flow Rate (MGD)


44
70
3
3
12
8
6
6
5
6
8
5
5
10
6
6
5
4

Commission Date
2006
2006
2006
2006
2006
2006
2005
2005
2004
2003
2002
2002
2001
2001
2001
2000
2000
1999

Dow Filmtec SW30HR-380


Facility

Membrane Type

Flow Rate (MGD)

Commission Date

Mar-03

Weihai City, China

SW30HR-380

Canary Islands, Spain

SW30HR-380

Treatment type
Treats seawater for process
water and drinking water for
plant
Desal

Membrane Type

Treatment type

Flow Rate (MGD)

Commission Date

Clearwater, Florida

Muni-RO-400

Treatment for drinking water


to reduce arsenic levels

2003

Gwinner, ND

MUNI-RO-300

TDS, iron, hardness

0.32

1990

Toluca, IL

MUNI-RO-350

Radium

0.29

1992

GE Water MUNI-RO-400
Facility

114
2012 Water Research Foundation. ALL RIGHTS RESERVED.

0.006

Daupin, AL

MUNI-RO-350

TDS, iron

0.24

1996

MUNI-LERO-400

Nitrates

1999

Pascagoula, MS

MUNI-LERO-FF-365

Color, TDS, H2S

5.7

1999

Torrington, WY

MUNI-LERO-FF-365

Nitrate removal

2.2

2000

Horizon City, TX

MUNI-LERO-FF-365

TDS

2.25

2000

Ottawa, IL

MUNI-LERO-FF-350

Radium

2001

Chelsea, MI

MUNI-LERO-FF-365

Softening

1.2

2001

Wabaunsee, KS

MUNI-LERO-FF-365

TDS

0.22

2001

Pottawatomie, Kansa

Muni-LERO-FF-365

TDS reduction surface water.

0.22

2001

Lake Granbury, TX

MUNI-RO-350

TDS and chlorides.

2002

Maud, OK

MUNI-RO-350

TDS

0.07

2002

Muni-LERO-350

Nitrate

0.43

2002

Inglis, FL

Muni-RO-400

THM/HAA precursors

0.65

2002

Todd Creek, CO

Muni-RO-400

TDS

0.43

2002

Great Rock, CO

Muni-RO-400

TDS

0.14

2002

Village of Golf, FL

Muni-RO-400

Softening / THM

0.87

2002

Hennessey, OK

Muni-RO-400

TDS

0.22

2002

Garden City, KS

MUNI-RO-2.5

TDS

2002

Gila Bend, AZ

MUNI-RO-150

TDS

0.65

2002

Purrysburg, SC

MUNI-RO-300

Arsenic

0.43

2003

Possum Kingdom, TX

MUNI-RO-350

TDS

2003

Box Elder, CO

MUNI-LERO-350

TDS/fluoride

0.14

2003

Milliken, CO

MUNI-LERO-350

TDS

0.65

2003

Hudson, CO

MUNI-LERO-350

TDS

0.22

2003

Windmere, TX

MUNI-LERO-400

TDS

2003

Pipeston, MN

Clara City, MN

115
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Bradford, IL

MUNI-LERO-400

Radium

0.29

2003

MUNI-RO-400

TDS

2004

Grand Blanc, MI

MUNI-LERO-350

TDS

2004

Macomb, IL

MUNI-LERO-400

Radium

2004

Spring Valley, IL

MUNI-LERO-400

Radium

1.4

2004

Bushmell, IL

MUNI-LERO400-

Radium

0.43

2004

El Paso, TX

MUNI-LERO-400

TDS

2004

Meander River, OK

MUNI-LERO-400

TDS

0.14

2004

Goodyear, AZ

MUNI-LERO-400

TDS

3.5

2004

Saline, MI

MUNI-LERO350-

Softening

2004

Davison, MI

MUNI-LERO-350

Arsenic

0.86

2004

San Juan Capistrano, CA

MUNI-LERO-365

TDS

2005

MUNI-RO-400

TDS

Tama, IA

MUNI-LERO-400

TDS, Fe, Mn

0.58

2006

St. John Parish, LA

MUNI-LERO-400

TDS, color, TTHM

2007

Pilot Knob WD Farminton, MO

MUNI-LERO-400

TDS, color, TTHM

0.29

2006

Le Center, MN

MUNI-LERO-400

TDS, hardness

1.3

2007

Pryor Lake, MN

MUNI-LERO-400

TDS, hardness

1.9

2008

Moss Point, MS

MUNI-LERO-400

TDS, color, TTHM

2008

Bonita Springs, FL

Schuylerille, NY

116
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2005

APPENDIX B
Survey / Questionnaire on Membrane Characterization Techniques and Procedures

117
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Date:
First and Last Name:
Name of Organization:
Role at Organization:
Level of Expertise in Surface/Membrane Characterization (circle one):
(Novice, Moderate, or Expert)

CONTACT ANGLE CHARACTERIZATION TECHNIQUE


The contact angle test measures the angle formed by the edge of a bubble on a submerged
membranes surface called the dynamic captive bubble technique. (See below).

Fig. 1 Captive bubble contact angle


In the dynamic captive bubble technique proposed, an air bubble is generated on the tip of a Ushaped needle. The needle is immersed in a quartz cuvette, which contains the probe liquid of
interest (e.g., water). The air bubble is detached from the needle and it rises through the liquid
until it comes into contact and subsequently attaches to the membrane surface. The contact angle
would then be measured by a Gonimeter.
The contact angle data gathered may be used to calculate specific surface energy properties (van
der Waals, Lewis acid-base) for detailed interfacial analyses, as well as for qualitatively
assessing the wettability, or hydrophobicity / hydrophilicity, of a membrane surface.
1. Given our description of the technique and the full test method, would you envision
finding the results of the contact angle test useful?

2. If the previous answer is a yes, how would you foresee using the results and often do you
think you would like to have the contact angle test conducted?

3. The contact angle test would include the following major pieces of equipment:
a. Goniometer (example manufactures: Ram-Hart or Krss Scientific).
Given this list, would you have the capacity of conducting the test in-house on a regular
basis?

118
2012 Water Research Foundation. ALL RIGHTS RESERVED.

If not, what piece(s) of equipment would you require so that you could conduct the test?

4. The contact angle test would require the following chemicals


a. ASTM Type 1 water, and
b. Saturated NaOH in isopropanol (for hollow fiber membrane)
Do you currently have or could readily order these chemicals?
If not, which chemical(s) would need to be specially ordered to conduct the test?

5. The contact angle test would require the following materials


a. PTFE magnetic stirrer,
b. Borosilicate glass vessel,
c. Powder-free gloves, and
d. Scissors.
Do you currently have or could readily order these materials?
If not, which material(s) would need to be specially ordered to conduct the test?

6. Do you believe that your current laboratory and/or analytical staff would have the
capability of conducting the contact angle experiment?

7. If the previous answer is a no, what aspect of the training would the staff require (such as
equipment training or specific procedure training)?

8. Any new test requires the staff to overcome a learning curve before consistent, accurate
results can be obtained. Given our description, how steep would you think this learning
curve would be for your staff to learn the contact angle test?

9. We believe that the contact angle test would take 6 hours to complete (2 hours for each
bubble measurement) once the process of understanding and conducting the test has been
established. Do you believe our time estimate is accurate?

119
2012 Water Research Foundation. ALL RIGHTS RESERVED.

10. Would the contact angle test have a significant impact on other lab activities by factors
such as lead time or the complexity of the test?

11. Do you see this procedure as an accurate and precise method for measuring the contact
angle?

12. Please provide any addition comments regarding the contact angle test.

120
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ZETA POTENTIAL CHARACTERIZATION TECHNIQUE


Zeta potential () is the potential difference between the bulk of solution and the sheer (slipping)
plane of the interfacial double layer along the membrane (A. V. Delgado, F. Gonzles-Caballero,
R.J. Hunter, L. K. Koopal and J. Lyklema, Measurement and Interpretation of Electrokinetic
Phenomena (IUPAC technical Report) Pure and Applied Chemistry, 77 (2005) p. 1753-1805).
Zeta potential, also referred to as streaming potential, is a function of surface and solution
chemistry (pH, ionic composition, and ionic strength) at the solid-liquid interface.
The zeta potential is measured by a streaming potential analyzer. The membrane is prepared
depending on type (flat sheet or hollow fiber) and placed in the analyzers flow cell. Then the
flow of electrolyte is started at constant pressure and changing directions until steaming potential
is stabilized. The measurement is recorded after the zeta potential is stabilized.
Zeta potential is an important membrane characteristic when assessing membrane fouling
potential and in developing chemical cleaning protocols.
1. Given our description of the technique and the full test method, would you envision
finding the results of the zeta potential test useful?

2. If the previous answer is a yes, how would you foresee using the results and often do you
think you would like to have the zeta potential test conducted?

3. The zeta potential test would include the following major pieces of equipment:
a. Streaming Potential Analyzer (by either CAD Instruments or Brookhaven
Instruments Corporation).
Given this list, would you have the capacity of conducting the test in-house on a regular
basis?
If not, what piece(s) of equipment would you require so that you could conduct the test?

4. The zeta potential test would require the following chemicals


a. ASTM Type 1 water,
b. Potassium Chloride (KCl),
c. Hydrochloric Acid (HCl), and
d. Potassium Hydroxide (KOH).
Do you currently have or could readily order these chemicals?
If not, which chemical(s) would need to be specially ordered to conduct the test?

121
2012 Water Research Foundation. ALL RIGHTS RESERVED.

5. The zeta potential test would require the following materials


a. PTFE magnetic stirrer,
b. Borosilicate glass vessel,
c. Polyethylene bottle,
d. Scissors,
e. Dessicator,
f. Play-Doh or similar,
g. Marine grade epoxy mix, and
h. Tube cutter.
Do you currently have or could readily order these materials?
If not, which material(s) would need to be specially ordered to conduct the test?

6. Do you believe that your current laboratory and/or analytical staff would have the
capability of conducting the zeta potential experiment?

7. If the previous answer is a no, what aspect of the training would the staff require (such as
equipment training or specific procedure training)?

8. Any new test requires the staff to overcome a learning curve before consistent, accurate
results can be obtained. Given our description, how steep would you think this learning
curve would be for your staff to learn the zeta potential test?

9. We believe that the zeta potential test would take 12 hours for a single ionic strength at 3
different pH values to complete once the process of understanding and conducting the
test has been established. Do you believe our time estimate is accurate?

10. Would the zeta potential test have a significant impact on other lab activities by factors
such as lead time or the complexity of the test?

122
2012 Water Research Foundation. ALL RIGHTS RESERVED.

11. Do you see this procedure as an accurate and precise method for measuring the zeta
potential?

12. Please provide any addition comments regarding the zeta potential test.

123
2012 Water Research Foundation. ALL RIGHTS RESERVED.

SURFACE ROUGHNESS CHARACTERIZATION TECHNIQUE


Surface roughness is simply the measure of the texture on the membrane surface and the test
would be with atomic level resolution using an atomic force microscope (AFM).
The membrane is first prepared depending on type (flat sheet or hollow fiber) before the surface
roughness is measured. The AFM uses a fine tip that exists on the end of the cantilever and is
scanned over the membrane surface using a tapping method. The tapping mode uses a rapidly
oscillating cantilever in the vicinity of the surface, and amplitude damping is used for imaging.
Only short intermittent contact of AFM tip with the sample (tapping) occurs, which is especially
suitable membrane surfaces. Direct measurement of surface features is by measuring the position
and movement of the cantilever as it is scanned over a membrane surface.
Surface roughness is an important property in determining membrane fouling propensity.
1. Given our description of the technique and the full test method, would you envision
finding the results of the surface roughness test useful?

2. If the previous answer is a yes, how would you foresee using the results and often do you
think you would like to have the surface roughness tested?

3. The surface roughness test would require the following major pieces of equipment:
a. Atomic Force Microscope (AFM).
Given this list, would you have the capacity of conducting the test in-house on a regular
basis?
If not, what piece(s) of equipment would you require so that you could conduct the test?

4. The surface roughness test would require the following chemicals


a. ASTM Type 1 water, and
b. Potassium Chloride (KCl) optional,
Do you currently have or could readily order these chemicals?
If not, which chemical(s) would need to be specially ordered to conduct the test?

5. The surface roughness test would require the following materials


a. PTFE magnetic stirrer,
c. Powder-free gloves,
d. Scissors,
e. Double stick tape,

124
2012 Water Research Foundation. ALL RIGHTS RESERVED.

f. Knife cleaned with isopropanol, and


g. Borosilicate glass vessel
Do you currently have or could readily order these materials?
If not, which material(s) would need to be specially ordered to conduct the test?

6. Do you believe that your current laboratory and/or analytical staff would have the
capability of conducting the surface roughness experiment?

7. If the previous answer is a no, what aspect of the training would the staff require (such as
equipment training or specific procedure training)?

8. Any new test requires the staff to overcome a learning curve before consistent, accurate
results can be obtained. Given our description, how steep would you think this learning
curve would be for your staff to learn the surface roughness test?

9. We believe that the surface roughness test would take 3 hours including sample
preparation to image acquisition to complete once the process of understanding and
conducting the test has been established. Do you believe our time estimate is accurate?

10. Would the surface roughness test have a significant impact on other lab activities by
factors such as lead time or the complexity of the test?

11. Do you see this procedure as an accurate and precise method for measuring the surface
roughness?

13. Please provide any addition comments regarding the surface roughness test.

125
2012 Water Research Foundation. ALL RIGHTS RESERVED.

APPENDIX C
Instructions to Participating Laboratories for
Conducting AFM, Contact Angle, and Streaming Potential Measurements

126
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Dear Participating Laboratory,


Thank you for agreeing to participate in the project titled Evaluation of Membrane
Characterization Methods that is being sponsored by the Water Research Foundation (formerly
AWWARF). At this time, we have begun Task C - Development of Membrane Characterization
Standard Techniques and we are now requesting your assistance with this investigation. The goal
of Task C is to develop standard techniques for characterizing the following membrane
properties:

Contact angle with water

Streaming potential (used to calculate zeta potential)

Surface roughness using atomic force microscopy (AFM)

The three teams participating in the development of these techniques are the University
of Nevada, Reno (UNR; A. Childress, project PI), Duke University (DU; M. Wiesner) and
Colorado School of Mines (CSM; T. Cath). Developing the standard methods for measuring the
aforementioned characteristics will be carried out in three phases. During Phase 1, UNR is
distributing both control and membrane samples to the partner laboratories (DU and CSM). UNR
is also distributing general instructions for performing the characterization measurements; these
instructions are not comprehensive and are intended as guidance only. Each laboratory will
perform contact angle, streaming potential, and surface roughness measurements using the
instructions. UNR will collect the data and analyze the local (within each institution) and global
(among all institutions) precision of each of the methods. Additionally, each laboratory will
record in detail the methodology that they used when measuring contact angle, streaming
potential, and surface roughness. This information will then be submitted to UNR along with the
characterization results for subsequent analysis. In Phase 2, UNR will further develop and refine
the instructions for measuring contact angle, streaming potential, and surface roughness using
AFM in order to develop a standard technique for measuring each characterization parameter. In
Phase 3, UNR will again distribute the control surfaces, membrane samples, and draft standard
techniques to the partner laboratories. They will repeat the same measurements as in Phase 1;
however, they will carefully follow the standard technique. The data will again be collected by
each lab and submitted to UNR for analysis. If necessary, the standard technique may be iterated
on until a desired level of precision is achieved.
Choices of Control Materials for Method Development
Prior to performing characterization measurements on membrane samples, measurements
will be performed on standard or control surfaces. The control surfaces for both flat sheet and
hollow fiber configurations are given in Table 1. The ability to characterize the properties of the
hollow fiber configuration (either concave or convex depending on outside-in or inside out flow)
is an important aspect of this investigation because two of the seven membranes to be tested are
of the hollow fiber configuration. Thus, in addition to the instructions for the flat surfaces,
instructions are also given for the laboratories to perform contact angle measurements on one
convex surface (glass capillary coated with PMMA, prepared by UNR) and AFM imaging on
one concave surface (angle-cut Teflon tubing).

127
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Table 1.
Reference surfaces selected for method development.
A
B
Measurement\Sample
Flat
Concave
PMMA block
PMMA film cast on flexible
with small
support (polyethersulfone support
diameter hole
Streaming Potential
from commercial membrane)
(or holes)
drilled though
it.
Cleaned glass
Teflon flat sheet from Ram-Hart
capillary
Contact Angle
with contact angle range and
painted black
material origin certificate.
on the outside
HS-20MG nanostructure array (20
Teflon tubing
nm step height) for Z-axis
~1 mm outer
AFM Roughness
calibration from
diameter
budgetsensors.com, or similar grid.

C
Convex
PMMA
commercial
rods or tubes,
or glass rods
covered with
PMMA film
Glass capillary
covered with
PMMA on the
outside
Teflon tubing
~1 mm outer
diameter

AFM instructions for flat samples (membranes or control materials)


1)

Rinse membrane (or control material, except for metrological nanostructure for Z- axis
calibration) three times with saturated ultrapure water, store completely immersed in
saturated ultrapure water at 5C at least 48 hours prior to measurement. Ultrapure water
has conductivity of 0.05 S cm when not saturated with ambient CO2, and 0.8 S cm
when equilibrated with atmospheric CO2. Check pH of saturated ultrapure water (5.6).

2)

Prepare liquid cell and tip for tapping mode measurement in liquid.

3)

Calibrate instrument in air (especially Z-axis), if metrological nanostructure is available.

4)

Cut coupon suitable for imaging and install in liquid cell, fill with saturated ultrapure
water.

5)

Follow instrument-specific instructions, set 10m 10m scan area, 256 steps in X and Y
direction; acquire date in tapping mode.

6)

Record and describe any data processing procedures applied that may affect roughness,
such as automatic leveling.

7)

Save data in software specific binary format and export data into SDF, ASCII or Excel
format. Free program Gwyddion may be useful in data conversion.

8)

Calculate average roughness, RMS roughness, and skewness (if possible) with
instrument-specific software.

9)

Create surface image in jpeg, png or gif format.

128
2012 Water Research Foundation. ALL RIGHTS RESERVED.

10) If direct conversion to SDF format was not possible, edit ASCII/Excel data to conform to
SDF specifications as described in Internet-based Surface Metrology Algorithm Testing
System (http://syseng.nist.gov/VSC/jsp/index.jsp).
11) Upload data in SDF format to Internet-based Surface Metrology Algorithm Testing
System and perform average roughness, RMS roughness, and skewness calculations
again and calculate respective uncertainties (all described functions are available in
Internet-based Surface Metrology Algorithm Testing System).
12) Repeat measurements for three regions on three different coupons (for a total of nine
measurements) for each membrane or control material.
13) Collect data in a table. Report tip characteristic (tip curvature, material, etc.), instrument
make and type, operator name and date. Report sample name and origin, measurement
temperature and results (roughness ,skewness, uncertainties, instrument specific software
and Internet-based Surface Metrology Algorithm Testing System, include image) for
each nine imaged spots.
14) Archive raw data files, create one compressed archive with meaningful name (related to
sample name and laboratory name) per membrane/control material.
AFM instructions for hollow fiber membranes (inside-out) or control concave materials
1)

Rinse sample three times with saturated ultrapure water, use syringe to rinse inside the
fiber, store completely immersed in saturated ultrapure water at 5C at least 48 hours
prior to measurement.

2)

Prepare instrument as for flat-sheet membranes.

3)

Cut fiber with knife or blade at a sharp angle to expose inside surface.

4)

Affix angle-cut hollow fiber material with two-sided tape to measurement cell, with
exposed inside of the membrane upwards. Make sure angle-cut area does not interfere
with tip movement.

5)

Fill cell with ultrapure water.

6)

Proceed as for flat-sheet membranes.

7)

Report any baseline correction options different from flat sheet membranes (automatic
leveling), e.g., if software integrated with the instrument allows for curved baseline
correction.

8)

Proceed with data recording and analysis as for flat sheet membranes.

9)

Report the same data as for flat-sheet membranes.

For sample cutting technique see: M. Rafat, D. De, K.C. Khulbe, T. Nguyen, T. Matsuura,
Surface characterization of hollow fiber membranes used in artificial kidney, J. of Appl. Polym.
Sci. 101 (2006) 4386-4400.
Contact angle instructions for flat samples (membranes or control materials)

129
2012 Water Research Foundation. ALL RIGHTS RESERVED.

1)

Sample preparation: membranes or reference materials should be rinsed with saturated


ultrapure water and stored fully immersed in saturated ultrapure water at 5C for at least
48 hours prior to measurements.

2)

Prepare goniometer for captive bubble measurement in saturated ultrapure water.

3)

Cut membrane coupon using sharp scissors. Stretch coupon on steel support (depending
on instrument), ensuring that there are no gaps between the support and the sample, and
also ensuring that the edge of the membrane is straight and smooth.

4)

Follow goniometer instructions for baseline and light optimization.

5)

Use dynamic captive bubble technique with 10 L bubble size: bubble should be
detached from needle under membrane prior to touching it.

6)

Measure 3 captive bubbles on 2 coupons (6 bubbles total), perform 5 automatic


measurements per each bubble with 1 minute time delay, if instrument allows.

7)

Save 1 representative bubble image per sample.

8)

Collect data for all 6 bubbles in a single Excel 97 (xls) file.

9)

Report instrument make and type and operator name and date. Report sample name and
origin, measurement temperature, and results (contact angle averages for left and right
side of the bubbles).

10) Save and archive one Excel file per sample (with bubble image pasted) with meaningful
name (related to sample name and laboratory name). UNR will calculate statistics from
raw data (average, standard deviation, confidence interval at 95% level).
Contact angle instructions for hollow fiber membranes (outside-in) or control convex
materials
1)

Sample preparation: hollow fiber membranes should be rinsed with saturated ultrapure
water three times to remove preservatives; sample should then be dessicated over
phosphorus pentoxide for at least 24 hours prior to measurement. If control surface is
clean, water rinsing can be omitted or replaced with control-specific cleaning procedures.

2)

Prepare goniometer stage: sample holder, transparent cover with paper pads soaked in
water that do not obscure sample image (see Figure 1). If air around the sample is dry,
any drop will evaporate quickly, and contact angle measurements will not be stable.

3)

Mount hollow fiber over goniometer stage, so fiber axis is parallel to goniometer stage
plane, and perpendicular to camera axis (axis normal to the lens surface at its center, see
photograph). Inverted holder for captive bubble technique may work well. Stretch fiber
gently, so it has appearance of a cylinder.

4)

Follow the same baseline and light optimization procedures as for captive bubble
technique.

5)

Use 5-15 L hanging sessile drop on horizontal cylinder (larger clam-shell drop might
release from less hydrophilic membranes). Produce advancing sessile drop drop
pumped with syringe after it is created on the hollow fiber surface.

6)

Look on the fiber directly from above make sure drop is not leaning to the front or to
the back, adjust by turning fiber in the holder.
130
2012 Water Research Foundation. ALL RIGHTS RESERVED.

7)

Run measurement as for captive bubble technique, set baseline at bottom of the capillary
contour.

8)

Make the same number of measurements as for flat-sheet membrane, report the same data
and fiber (cylinder) outer diameter measured with a micrometer.

Figure 1. Capillary with hanging clam-shell drop photographed at UNR laboratory. Elements of
sessile drop holder were used to position capillary on the goniometer stage. Note that a small
reservoir with water and cover with moist paper pads were used to prevent drop evaporation.
Zeta potential instructions for flat samples (membranes or control materials)
1)

Perform calibration of streaming potential instrument conductivity and temperature


sensors.

2)

Prepare electrolyte solutions needed using saturated ultrapure water and ACS reagents
(see below).

1)

Sample preparation: rinse 3 times with saturated ultrapure water, store completely
immersed in 2.0 mM KCl solution (prepared from saturated ultrapure water and ACS
certified KCl) at 5 C for at least 48 hours prior to measurement,

2)

Cut coupon to size, install coupon in flow cell just prior to measurement.

3)

After cell is installed, start flow of electrolyte (2.00 mM KCl ACS certified in ultrapure
saturated water) at constant pressure and changing directions.

4)

Measure electrolyte conductivity and pH with external well-calibrated instruments and


make sure that conductivity is in agreement with values reported by internal sensor.

5)

When streaming potential is stabilized, collect data. When variable pH is desired, prepare
electrolytes using KOH or HCl (add KCl to keep total monovalent ion concentration at
4.0 mM).

6)

Follow the following pH cycle (5.7 corresponds to ambient air saturated ultrapure water)
keeping the same sample and changing electrolytes: 5.7
4.0 3.0 5.7 7.0
8.0.

9)

Run triplicate measurements (3 different coupons) for each sample.

131
2012 Water Research Foundation. ALL RIGHTS RESERVED.

10) Report instrument make and type and operator name and date. Report sample name and
origin, electrolyte composition and pH, and results: streaming potential vs pressure slope,
slope standard deviation, temperature, conductivity, calculated viscosity, calculated
permittivity, and zeta potential. Use the following formulas for viscosity () and
permittivity () , where t is temperature in degrees Celsius:

11) Make one Excel file per electrolyte and sample coupon. Create one compressed directory
(containing excel files) with meaningful name (related to sample name and laboratory
name) per membrane/control material. UNR team can calculate statistics from raw data
(report streaming potential vs pressure, temperature and conductivity).

132
2012 Water Research Foundation. ALL RIGHTS RESERVED.

APPENDIX D
Standard Protocols for Characterizing Membrane Surfaces:
Contact Angle, Zeta Potential, Surface Roughness

133
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Standard Protocols for Characterizing Membrane


Surfaces: Contact Angle, Zeta Potential, Surface Roughness

WaterRF Project #4102:


Evaluation of Membrane Characterization Methods

Funding Agency
Water Research Foundation
Project Number 4102
6666 W. Quincy Avenue
Denver, CO 80235-3098

Principal Investigator
Dr. Amy E. Childress
University of Nevada, Reno
Department of Civil and Environmental Engineering/258
1664 N. Virginia Street
Reno, NV 89557-0258

Co-Principal Investigator
Dr. Jonathan A. Brant
HDR Engineering, Inc.
500 108th Avenue NE, Suite 1200
Bellevue, WA 98004-5549

September 30, 2010

134
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Contents
CHAPTER 1: SURFACE WETTABILITY - CONTACT ANGLE MEASUREMENTS
General Discussion
Contact Angle Measurement Apparatus and Software
Materials and Reagents
Contact Angle Measurement Procedures
Procedure for Flat-Sheet Membrane Samples
Procedure for Hollow Fiber Membrane Samples (External Surface)
Contact Angle Data Analysis
Additional Recommendations
CHAPTER 2: IONIZATION OF SURFACE FUNCTIONAL GROUPS AND ADSORBED
IONS: SURFACE CHARGE - STREAMING POTENTIAL / ZETA POTENTIAL
General Discussion
Test Apparatus and Software
Materials and Reagents
Procedure
Procedure for Flat-Sheet Membranes
Procedure for Hollow Fiber Membranes
Streaming Potential Data Analysis
Additional Recommendations
CHAPTER 3: MEMBRANE SURFACE ROUGHNESS - AFM MEASUREMENTS
General Discussion
AFM Apparatus and Software
Materials and Reagents
Procedure
Procedure for Flat Membranes
Procedure for Hollow Fibers
AFM Roughness Data Analysis
Additional Recommendations

135

2012 Water Research Foundation. ALL RIGHTS RESERVED.

136
136
137
139
139
139
140
142
142
143
143
143
144
144
144
145
146
146
148
148
148
149
149
149
149
151
151

CHAPTER 1
SURFACE WETTABILITY - CONTACT ANGLE MEASUREMENTS
GENERAL DISCUSSION
The contact angle () that is formed at the three-phase interface between solid, liquid, and
gas / vapor phases may be used to elicit information regarding the surface energy properties of a
material, such as membranes. The contact angle is the angle that is formed between a surface and
the tangent line whose origin is at the three-phase junction and is measured through the liquid
phase (Fig. 1). The information that is gathered from contact angle analysis may be used to
calculate specific surface energy properties (van der Waals, Lewis acid-base) for detailed
interfacial analyses, as well as for qualitatively assessing the wettability, or hydrophobicity /
hydrophilicity, of a membrane surface. This latter application is perhaps the most common use
for contact angle by utilities and other membrane users.

Figure 1. Sessile drop configuration (left) for measuring contact angle in which a liquid droplet
is placed on a dry surface and captive bubble configuration (right) for measuring
contact angle in which a surface is immersed in a probe liquid and a air bubble is
placed on the wetted surface.
The two most commonly used methods for measuring contact angle on membrane surfaces
are the sessile drop and captive bubble techniques (Fig. 1). Of these techniques the captive
bubble method is being used more and more frequently as it allows the membrane to be
characterized in its native hydrated (wet) state. These two techniques are illustrative of the two
general types of contact angles that may be measured for membranes: the advancing and
receding contact angles. The advancing contact angle is that which is measured for a dry surface,
where the liquid is advancing over the previously dry surface. Conversely, the receding contact
angle is that which is measured for a previously wetted or hydrated surface. Contact angle
hysteresis, which is the difference between the advancing and receding contact angle values, is
one of the greatest challenges in contact angle analysis. Hysteresis is a challenge because it can
produce two different contact angle values (receding and advancing angle), as evidenced by a
drop on a tilted flat surface (Fig. 2). The challenge is to determine which value is the most
relevant for a particular application. For membrane analysis the receding contact angle would
seem to be more relevant as membranes are hydrated or wet in most practical applications.
Membrane surface roughness and porosity may also affect the value of the measured contact
angle values relative to those contact angles that are measured for a smooth non-porous surface

136
2012 Water Research Foundation. ALL RIGHTS RESERVED.

of the same material. Contact angles may be corrected for surface roughness using the Wenzel
Equation.

Figure 2. Contact angle formed between a water droplet on an inclined Teflon surface. This
image illustrates the difference between an advancing (angle formed on the lower left)
and receding (angle formed on the upper right) contact angle.
In the dynamic captive bubble technique proposed here, an air bubble is generated on the
tip of a needle that is U-shaped. The needle is immersed in a quartz cuvette, which contains the
probe liquid of interest (e.g., water). After detachment, the air bubble rises through the liquid and
comes into contact and subsequently attaches to the membrane surface. The type of contact angle
that is measured using this technique is referred to as the receding contact angle, because the
water has receded from the previously wetted surface. The dynamic captive bubble technique
as proposed here is recommended for measuring contact angles on membrane surfaces because it
characterizes the membrane in its most relevant condition (i.e., hydrated). Measurement of the
advancing contact angle using the captive bubble technique is not recommended because it
would require air to be withdrawn from the bubble (after it is formed on the sample surface) until
the contact angle reaches a stable value. This action would increases the risk of surface damage
(proximity of tip to the membrane), would complicate manual control of bubble volume, and
would distort the shape of the air bubble as a result of the needle penetrating into its volume. It
would also limit choices of image analysis method (asymmetric drop shape analysis, ADSA, was
modified to accommodate the needle only recently and may not be implemented in current
software).
CONTACT ANGLE MEASUREMENT APPARATUS AND SOFTWARE
Contact angle is most commonly measured using a goniometer (Fig. 3). Goniometers are
instruments that consist of a sample stage, digital camera, and a light source. Other add-ons are
also available that allow for digitally capturing and analyzing the bubble or droplet shape in
order to calculate contact angle. The primary equipment vendors for goniometers are Ram-Hart
and Krss Scientific, though others also exist. The principle components of a goniometer system
that is capable of operating in the sessile and/or captive bubble configuration, include:

Adjustable height sample stage,

Adjustable light source,

Computer interfaced digital camera,

137

2012 Water Research Foundation. ALL RIGHTS RESERVED.

Quartz cuvette (captive bubble),

Environmental chamber (temperature and humidity control)

Gas-tight micro syringe with U-shaped needle

Sample holders

Fixed drop standards images of idealized drops on a slide with certified contact angle
values (Figure 4, used for accuracy assessment)

Two rectangular plates cut out of microscope slide to sandwich a hollow fiber membrane
in sample holder (Figure 5

Figure 3. Example of a modern goniometer system (Ram-Hart, Netcong, NJ) that includes a
computer for data acquisition and analysis, environmental chamber, sample stage, light
source, automatic syringe, and digital camera for image acquisition.

Figure 4. Certified contact angle fixed-drop calibration reference tool (Ram-Hart, Netcong, NJ)
for determining the accuracy of contact angle values measured in a respective
laboratory.

138
2012 Water Research Foundation. ALL RIGHTS RESERVED.

MATERIALS AND REAGENTS


To facilitate the comparison of results between labs and to allow for the calculation of
surface energy parameters from contact angle values, it is highly recommended that all reagents
be of the highest available quality. For measuring the contact angle that water forms on a surface
it is recommended that ASTM Type I water be used (maximum conductivity = 0.055 S cm-1).
Following production and subsequent exposure to the surrounding atmosphere the ASTM
Type I water will be subject to carbon dioxide (CO2) dissolution until saturation occurs. This will
affect the conductivity and pH of the water. It is therefore recommended that the water to be used
for contact angle measurements be allowed to reach equilibrium with the atmosphere prior to
performing the contact angle measurements. This may be accomplished by stirring the water
sample (PTFE magnetic stirrer) in ambient air in properly cleaned borosilicate glass vessel
(Standard Method 1070 B, Standard Methods for Examination of Water and Wastewater, 1995),
until the conductivity reaches 0.7 S cm-1.
Note that measuring the pH of ASTM Type I water requires special electrodes. As an
alternative, we propose that ACS certified KCl can be dissolved in ultrapure water (0.50 g/L) and
pH tested periodically with standard pH meter (pH = 5.6 for ambient CO2 saturation). The water
sample that is being used to determine the pH of the ASTM Type I water should not be used for
contact angle measurements. As a precaution, the water sample for contact angle measurements
should be stored in borosilicate glass for no longer than one week.
CONTACT ANGLE MEASUREMENT PROCEDURES
Procedure for Flat-Sheet Membrane Samples
1) Contamination of the sample surface must be avoided. Powder-free gloves must be worn
at all times when handling the membrane samples and goniometer equipment that will
come into contact with the probe liquid and membrane. Sample preparation: membranes
should be rinsed with saturated ASTM Type I water and stored fully immersed in
saturated ASTM Type I water at 5C for at least 48 hrs prior to performing the contact
angle measurements. When selecting samples for analysis choose sample areas that are
free of visible defects, like scratches or stains.
2) Set-up and start the goniometer according to the vendor supplied instructions. Precaution
should be taken to ensure that the instrument is leveled and protected from external
vibrations. This is especially important when older instruments are used and can be
major source of error.
3) Test the instrument using a fixed drop image standard: measure all four contact angles
imaged on the standard and calculate deviations from certified values. Calculate the
relative value (%) of the greatest deviation (absolute value) as an estimate of systematic
error.
4) Prepare goniometer for captive bubble measurement in saturated ASTM Type I water.
When goniometric method can be chosen, specify polynomial fit to bubble edge, not
asymmetric drop/bubble shape analysis (ADSA). Although ADSA is accurate and uses
139

2012 Water Research Foundation. ALL RIGHTS RESERVED.

information of the whole drop/bubble shape to calculate contact angle, it is not applicable
to hollow fiber measurement as proposed below. Polynomial fitting is applicable to both
flat-sheet and hollow fiber samples and will assure better method uniformity.
5) Cut membrane coupon using sharp scissors. Stretch coupon on steel support (depending
on instrument), ensuring that there are no gaps between the support and the sample, and
also ensuring that the edge of the membrane is straight and smooth. Immerse sample in
wet cell.
6) Follow the instrument specific instructions for determining the sample baseline and for
optimizing the light conditions. Run measurements in the dark, to minimize stray light.
7) Test samples at ambient temperature 23 2 C.
8) Use the dynamic captive bubble technique and an air bubble volume of 10 L. The
volume of the air bubble may be specified if using a computer controlled syringe or
manually by the operator. The distance between the needle tip and the membrane sample
surface must be such so as to allow the bubble to detach from the needle under the
membrane prior to the two coming into contact. PTFE tubing can be put around the
syringe tip, if bubble breaks down before reaching desired volume. Tap U-shaped needle
at the bottom to detach bubble.
9) Measure at least 3 captive bubbles on 2 membrane coupons (6 bubbles total), perform 5
automatic measurements per each bubble with 1 minute time delay between
measurements. Treat average for left and right contact angle from 5 measurements as a
single data point. Report instrument make and type, report sample name and origin,
cleaning procedures, if any, measurement temperature if outside specified range, number
of bubbles and coupons, average, sample standard deviation, confidence interval at 95%
level).
Procedure for Hollow Fiber Membrane Samples (External Surface)
1) Prior to conducting the contact angle measurement it is necessary that the hollow fiber
membranes be rinsed with saturated ASTM Type I water three times to remove
preservatives and other contaminants that may be present on the membrane surface.
Some hollow fiber membranes are shipped with glycerol and bisulfite as preservatives.
Run DI water through plugged fiber, with pressure that allows water flow across the
pores. Check effluent for TOC. Pores may become flooded when pressure is applied,
even with hydrophobic material, this is usually irreversible process. Sample may appear
more hydrophilic (smaller contact angle) as compared to a sample with dry pores.
Describe cleaning procedure in the report.
2) After cleaning, the hollow fiber membranes should be stored fully immersed in saturated
ASTM Type I water at 5C for at least 48 hrs prior to measurements. When selecting
samples for analysis choose sample areas that are free of visible defects, like scratches or
stains. Use powder-free gloves to handle the membrane samples and associated
equipment at all times.

140
2012 Water Research Foundation. ALL RIGHTS RESERVED.

3) Set-up and start the goniometer according to the vendor supplied instructions. Precaution
should be taken to ensure that the instrument is leveled and protected from external
vibrations.
4) Test the instrument using a fixed drop image standard: measure all four contact angles on
the standard and calculate deviations from certified values. Calculate the relative value
(%) of the greatest deviation (absolute value) as an estimate of systematic error
5) Mount hollow fiber between two thoroughly cleaned (Saturated NaOH in isopropanol at
room temperature followed by 5-fold water rinses; water drop spreads uniformly and
quickly on clean surface) pieces of microscope slide (Fig. 5). Make sure that syringe
needle is thinner than the hollow fiber tested.

Figure 5. PTFE tubing sandwiched between two glass slides.


4) Mount hollow fiber over goniometer stage, so fiber axis is parallel to goniometer stage
plane, and perpendicular to camera axis (axis normal to the lens surface at its center).
6) Follow the instrument specific instructions for determining the sample baseline and for
optimizing the light conditions. Run measurements in the dark, to minimize stray light
7) Test samples at ambient temperature 23 2 C.
8) Use the dynamic captive bubble technique and an air bubble volume of 10 L. The
volume of the air bubble may be specified if using a computer controlled syringe or
manually by the operator. The distance between the needle tip and the membrane sample
surface must be such so as to allow the bubble to detach from the needle under the
membrane prior to the two coming into contact. PTFE tubing can be put around the
syringe tip, if bubble breaks down before reaching desired volume. Tap U-shaped needle
at the bottom to detach bubble.
5) Measure at least 3 captive bubbles on 2 membrane coupons (6 bubbles total), perform 5
automatic measurements per each bubble with 1 min time delay between measurements.
Treat average for left and right contact angle from 5 measurements as a single data point.
Report instrument make and type, report sample name and origin, cleaning procedures, if
any, measurement temperature if outside specified range, number of bubbles and
coupons, average, sample standard deviation, confidence interval at 95% level).

141

2012 Water Research Foundation. ALL RIGHTS RESERVED.

CONTACT ANGLE DATA ANALYSIS


For determining the average, standard deviation, and 95% confidence level at least three
measurements should be per sample per pH and electrolyte concentration. The noted statistics
should be reported for each test condition. The left and right contact angles should be used to
calculate a single average contact angle value per air bubble or water droplet. If the deviation
between the left and right angles exceeds 2, then the bubble/droplet should be discarded and the
measurement repeated.
ADDITIONAL RECOMMENDATIONS
The following recommendations are also suggested in addition to the standard method
proposed earlier for carrying out contact angle measurements on flat-sheet and hollow fiber
membranes:

Teflon standard surface flat surface can be used, but it will require re-certification
from provider at additional cost (re-measured for receding contact angle
measurement with captive bubble), as current certified value is advancing contact
angle measured with sessile drop technique,

A 3D fixed shape calibration standard (made from gauge ball and plate) should be
used to establish that the sample stage is properly leveled and to establish the
error that is associated with the goniometer being used,

Front-back inclination of the goniometer stage with respect to the camera axis is
the key geometrical parameter in goniometer setup affecting accuracy,

Organic contaminants can have profound effect on contact angle measurements.


TOC of ASTM water should be monitored and not to exceed 100 ppb. Use fresh
HPLC-grade water, if TOC monitoring is not possible,

Use zero air (from which traces of organics were removed) to equilibrate water,
and

Consider glove box filled with zero air, when VOC can be present in the
laboratory air.

142
2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 2
IONIZATION OF SURFACE FUNCTIONAL GROUPS AND ADSORBED
IONS: SURFACE CHARGE - STREAMING POTENTIAL / ZETA
POTENTIAL
GENERAL DISCUSSION
Zeta potential () is the potential difference between the bulk of solution and the sheer
(slipping) plane of the interfacial double layer (A. V. Delgado, F. Gonzles-Caballero, R.J.
Hunter, L. K. Koopal and J. Lyklema, Measurement and Interpretation of Electrokinetic
Phenomena (IUPAC technical Report) Pure and Applied Chemistry, 77 (2005) p. 1753-1805).
Zeta potential is a function of surface and solution chemistry (pH, ionic composition, and ionic
strength) at the solid-liquid interface.
Zeta potential is an important membrane characteristic when assessing membrane fouling
potential and in developing chemical cleaning protocols. Currently, zeta potential is used as a
substitute for surface charge, which so far is not accessible through direct measurement.
Membrane zeta potential is determined from streaming potential measurements. Streaming
potential is generated when electrolyte flows through a capillary channel, or a porous plug, and is
related to zeta potential of the capillary (plug)/electrolyte interface by the HelmholtzSmoluchowski equation:
E 0
=
Equation 1
p

where E is streaming potential due to electrolyte flow through a capillary channel, p is the
applied pressure driving the flow, is zeta potential, is electrolyte conductivity, is the
viscosity of the electrolyte solution, is the relative permittivity of the solution (dimensionless),
and 0 is the permittivity of a vacuum (fundamental constant). Values of E, p, and can be
determined using a specialized instrument, while and are calculated based on temperature
measurement (empirical fit functions for pure water data are used). Instruments with pH control
(titration) capabilities are available.
TEST APPARATUS AND SOFTWARE
Currently, there are two widely utilized vendors for streaming potential analyzers: CAD
Instruments, France and Brookhaven Instruments Corporation, Holtsville, NY. Streaming
potential analyzers consist of a sample cell whose design is tailored to the material being
analyzed (flat sheet, fiber, granular material), a pumping system (typically pressurized nitrogen)
to move an electrolyte solution through the sample cell, pressure/conductivity/pH/temperature
probes, and electrodes to measure the voltage across the sample cell. Other add-ons may also be
available to allow for titration of the test solution. The instrument is interfaced with a computer
to allow for data acquisition and analysis in real-time. Temperature is used to calculate the
viscosity and electric permittivity of the test solution, which are then used to calculate streaming
potential. While sample cells are commercially available in a variety of geometries, it may be
necessary to custom design some cells in order to meet system specific constraints as may be the

143

2012 Water Research Foundation. ALL RIGHTS RESERVED.

case for some hollow fiber membranes. Most commercially available instruments are supplied
with software packages for acquiring and analyzing the data.
MATERIALS AND REAGENTS
To facilitate the comparison of results between labs it is necessary that all reagents be of
the highest possible quality. Specifically, all electrolyte solutions should be made using ASTM
Type I water (maximum conductivity = 0.055 S cm-1, ASTM D1193-06) or its equivalent.
Following production and subsequent exposure to the surrounding atmosphere the ASTM Type I
water will be subject to carbon dioxide (CO2) dissolution until saturation occurs. This will affect
the conductivity and pH of the water. It is therefore recommended that the water to be used in the
streaming potential measurements be allowed to reach equilibrium with the atmosphere prior to
performing the contact angle measurements. This may be accomplished by stirring the water
sample (PTFE magnetic stirrer) in ambient air in properly cleaned borosilicate glass vessel
(Standard Method 1070 B, Standard Methods for Examination of Water and Wastewater, 1995),
until the conductivity reaches 0.7 S cm-1. ACS certified reagents must be used when preparing
all test solutions. The following reagents are required to prepare the test solutions as described in
this overview: potassium chloride (KCl), hydrochloric acid (HCl), potassium hydroxide (KOH).
Prepare 0.1 M solutions of HCl and KOH using ASTM Type I water. The KOH solution
should be stored in a polyethylene bottle, as it will leach silica from glass over time. Potassium
hydroxide that contains carbonates may be used, as atmospheric CO2 will be absorbed from the
air in absence of special precautions. A 2.0 mM KCl solution should be prepared as follows:
place 0.2983 g of ACS grade KCl in a 2 L volumetric flask and add ASTM Type I water. Adjust
the solution pH to the desired value using wither HCl or KOH. The volume of HCl or KOH
solution that must be added to achieve a pH between 3 to 8 will have a negligible effect on the
KCl concentration, but overall conductivity changes rapidly with acid addition as a result of the
high mobility of H+ ions. Acidified solutions (pH 3 or 4) are quite stable, in contrast to neutral
and basic solutions that absorb atmospheric CO2. Use basic solutions immediately after
preparation, create pH by 0.3 unit greater than the target, as pH measured in the bulk of the
solution after streaming potential experiment is usually decreasing.
PROCEDURE
The procedure used for preparing the membrane samples for streaming potential analysis
will depend on the membrane configuration being characterized. Two separate procedures are
outlined below for flat sheet membranes (i.e., samples from spiral wound elements) and for
hollow fiber membranes.
Procedure for Flat-Sheet Membranes
1) Sample preparation: rinse 3 times with saturated ASTM Type I water, store completely
immersed in 2.0 mM KCl solution (prepared from saturated ultrapure water and ACS
certified KCl) at 5 C for at least 48 hrs prior to measurement.
2) Set-up and start the streaming potential analyzer according to the vendor supplied
instructions.

144
2012 Water Research Foundation. ALL RIGHTS RESERVED.

3) Perform calibration of streaming potential instrument sensors: conductivity, temperature,


and pH (if equipped). Note the location of the temperature sensor: if it is removed from
the cell make sure that solution is thermally equilibrated with the ambient air. Avoid
conditions that may lead to temperature gradients (heaters, air conditioning air currents).
4) Test samples at ambient temperature (23 2 C)
5) Prepare the necessary electrolyte solutions (2.00 mM KCl with pH adjusted with KOH or
HCl) using saturated ASTM Type I water and ACS reagents (see above).
6) If an asymmetric cell is used (most frequently PMMA + sample) measure the zeta
potential of the fixed material under the exact conditions (electrolyte composition) to be
applied to the membrane sample.
7) Cut the membrane coupon to size, install coupon and spacers in flow cell just prior to
measurement.
8) After the cell is installed purge the cell and start flow of electrolyte (2.00 mM KCl ACS
certified in ultrapure saturated water) at constant pressure and changing directions. If
streaming potential shows changes over time, wait until it is stabilized. If streaming
potential is not stable, air bubbles in the cell may be present: purge and start again.
9) When streaming potential is stabilized, collect data. Run measurement three times to
estimate repeatability.
10) After experiment measure electrolyte conductivity and pH with external well-calibrated
instruments and make sure that conductivity is in agreement with values reported by the
internal sensor during experiment.
11) When variable pH is desired, prepare electrolytes using KOH or HCl as described above.
Follow the following pH cycle (5.7 corresponds to ambient air saturated ultrapure water)
keeping the same sample and changing electrolytes: 5.7 4.0 3.0 5.7 7.0 8.0
12) Run triplicate measurements (3 different coupons) for each sample.
13) Report instrument make and type, sample name and origin, electrolyte composition and
pH, and results: streaming potential versus pressure slope, temperature, conductivity,
calculated viscosity, calculated permittivity, and zeta potential.
Procedure for Hollow Fiber Membranes
1) Follow procedures for flat-sheet samples, except that a porous plug must be prepared
first.
2) Arrange fittings compatible with the instrument that will allow for installation of easily
available tubing like PVC (Figure 6).
3) Rinse hollow fibers as described in the contact angle procedure, then dry in dessicator.
145

2012 Water Research Foundation. ALL RIGHTS RESERVED.

4) Place 4-5 dry, clean hollow fibers centrally inside piece of tubing mounted vertically in a
holder, plug the bottom with Play-Doh or similar material.
5) Prepare marine grade epoxy mix. Start pouring epoxy inside tubing, when it has
consistency of honey so it is too viscous to soak into the hollow fiber membrane pores.
6) Wait overnight for the epoxy to cure, cut central section (about 5 cm long) for
measurement with tube cutter. Do not use saws, files or sandpaper otherwise dust goes
into the hollow fibers and is hard to remove. Flush the plug with running DI water, then
condition in 2.0 mM KCl solution for 48 hours.

Figure 6. Hollow fiber porous plug compatible with ZETA-CAD instrument.


STREAMING POTENTIAL DATA ANALYSIS
Check that the internally calculated values for solution viscosity, and permittivity,
agree to within 0.5% with the values that are manually calculated using Eq.2:

= 1.787 + 1.5615 107 t 4 - 2.221 10-5 t 3 + 1.4409 10-3t 2 - 6.044 10-2t


= 87.9144 - 1.32802 10-6t 3 + 9.58726 10-4t 2 - 0.404399 t

Equation 2

where t is the solution temperature (C). If agreement is satisfactory, then the zeta potential that
is automatically calculated by the streaming potential analyzer is acceptable. Otherwise, use the
Helmholtz-Smoluchowski equation (Eq. 3) and the values for and that were determined using
Eq. 2 to manually calculate zeta potential:

E (t )
p 0 (t )

Equation 3

For determining the average, standard deviation, and 95% confidence level at least three
measurements should be per sample per pH and electrolyte concentration. The noted statistics
should be reported for each test condition.
ADDITIONAL RECOMMENDATIONS

146
2012 Water Research Foundation. ALL RIGHTS RESERVED.

The following recommendations are also suggested in addition to the standard method
proposed earlier for carrying out streaming potential measurements on flat-sheet and hollow fiber
membranes:

Silver chloride electrodes are preferable over platinum electrodes,

Sample conditioning (soaking time, solution chemistry) is of the utmost importance


for obtaining reproducible results both within a single laboratory and between
laboratories. When comparing zeta potential results for membranes it is highly
recommended that the sample condition procedures used at each laboratory be published
and compared,

Design of a standardized streaming potential measuring cell that could be used with
all electrokinetic analyzers is desirable for further improving the reproducibility in
streaming potential results.

147

2012 Water Research Foundation. ALL RIGHTS RESERVED.

CHAPTER 3
MEMBRANE SURFACE ROUGHNESS - AFM MEASUREMENTS
GENERAL DISCUSSION
Surface roughness is an important property in determining membrane fouling propensity.
Surface roughness may be measured with atomic level resolution using an atomic force
microscope (AFM). AFM is widely used to characterize the physical and chemical properties of
membrane surfaces. The procedure that is outlined here specifically pertains to the
characterization of membrane surface roughness. An AFM operates using a light-lever
technique, where a laser is reflected off the back of a cantilever and its position determined using
a photo-diode. A fine tip exists on the end of the cantilever and is scanned over a surface using
any number of operating modes (e.g., tapping, contact). Tapping mode is most commonly used
to characterize membrane surfaces as contact mode may result in damage to the membrane
surface. Tapping mode uses rapidly oscillating cantilever in the vicinity of the surface, and
amplitude damping is used for imaging. Only short intermittent contact of AFM tip with sample
(tapping) occurs, which is especially suitable membrane surfaces. Measuring the position and
movement of the cantilever as it is scanned over a membrane surface allows for the direct
measurement of surface features.
The data that are collected by the AFM during imaging can be subsequently analyzed and
used to produce 3-D rendered images of the membrane surface and for calculating a variety of
surface roughness statistics, such as average roughness (Ra), root mean square (RMS) roughness
(Rq), and surface area difference. Of these various parameters, Rq and Ra are the most widely
used to describe the morphology of membrane surfaces. The RMS roughness, Rq, is a population
standard deviation for a set of z values (depth coordinate) collected for a given area, where zi are
heights (z-coordinates) from the AFM scan, N is number of data points (i = 1,2,,N), and i is
data index.

Rq =

(z

zavg ) 2
N

Equation 4

The average roughness uses averaged absolute values of deviations from average:

Ra =

| z

zavg |

Equation 5

AFM APPARATUS AND SOFTWARE


AFM systems and supplies for imaging membrane samples may be purchased from a variety
of vendors. A wet-cell accessory is recommended for imaging membrane samples and is required
for imaging wet membranes. AFMs are typically supplied with vendor specific image analysis
software, though there are numerous image analysis programs that may be purchased for
analyzing AFM generated data. Due to the nature of the instrument specific site constraints and
precautions must be utilized (e.g., vibration dampening, temperature control, etc.). The AFM
vendor should be contacted to determine instrument specific requirements.

148
2012 Water Research Foundation. ALL RIGHTS RESERVED.

MATERIALS AND REAGENTS


Water described here is the same as proposed for contact angle measurements. Ultrapure
water, like ASTM type I, degrades very quickly when exposed to ambient air, if conductivity is
considered as primary purity indicator. We propose that conductivity of water used here should
be 0.055 S cm-1 at the source, but it should absorb atmospheric CO2 in a controlled way.
Otherwise absorption will occur at the time of measurement anyway (wet cell is not protected
from atmospheric CO2), and characteristics of original water will not apply. ASTM Type I water
shall be stirred (PTFE magnetic stirrer) in ambient air in thoroughly cleaned borosilicate glass
vessel, until conductivity reaches 0.7 S cm-1.
Calibration standards are recommended for determining the accuracy and error that is
associated with surface roughness measurements in any given laboratory. These may be
purchased from the AFM vendor.
PROCEDURE
Procedure for Flat Membranes
1) Acquire image of at least one z-calibration standard in air in tapping mode. Verify that
measured depth is in agreement with certified value.
2) Rinse membrane three times with saturated ultrapure water, store completely immersed in
saturated ultrapure water at 5C at least 48 hours prior to measurement.
3) Prepare liquid cell and tip for tapping mode measurement in liquid.
4) Cut coupon suitable for imaging and install in liquid cell, fill with saturated ultrapure
water.
5) Follow instrument-specific instructions, set 10 m 10 m scan area, 256 steps in X and
Y direction; acquire data in tapping mode.
6) Record and describe any data processing procedures applied that may affect roughness,
such as automatic leveling.
7) Calculate RMS roughness with instrument-specific software. Save raw data and create
surface image.
8) Repeat measurements for three regions on three different coupons (for a total of nine
measurements).
9) Report tip characteristic (tip curvature, material, etc.), instrument make and type, sample
name and origin, measurement temperature and RMS roughness.
Procedure for Hollow Fibers

149

2012 Water Research Foundation. ALL RIGHTS RESERVED.

1) Rinse sample three times with saturated ASTM Type I water, use syringe to rinse inside
the fiber, store completely immersed in saturated ultrapure water at 5C at least 48 hrs
prior to measurement. Refer to the cleaning procedure for hollow fibers treated with
preservatives in Section 1.4.2.
2) Prepare instrument as for flat-sheet membranes.
3) For inner surface imaging cut fiber with knife or blade (previously cleaned with
isopropanol to prevent oil contamination often present on new steel blades) at a sharp
angle to expose inside surface (Fig. 8). One of the participating labs reported better
results with simple cut along the fiber axis.
4) Affix angle-cut hollow fiber material with two-sided tape to measurement cell, with
exposed inside of the membrane upwards. Make sure angle-cut area does not interfere
with tip movement. Fill cell with ASTM Type I water or its equivalent.
5) For imaging the outer surface of the membrane it is critical that the sample be properly
mounted. If fiber slides on two-sided tape, cut is in the middle along the axis and affix
with external surface facing upwards.
6) Proceed according to the procedure that was outlined for flat-sheet membranes (Section
3.4.1 Step 5).
a. Note: Make sure to report any baseline correction options different from flat sheet
membranes (automatic leveling), e.g., if software integrated with the instrument
allows for curved baseline correction.

150
2012 Water Research Foundation. ALL RIGHTS RESERVED.

2/3 part for outside


Cutting
2/3

1/3

AFM tip
1cm

1/3 part for inside image


Figure 8. Membrane angle cut for inner surface imaging.
AFM ROUGHNESS DATA ANALYSIS
Calculate the average, standard deviation, and 95% confidence level for Rq and Ra. A
minimum of six separate areas must be imaged on the membrane surface in order to generate
proper statistics.
ADDITIONAL RECOMMENDATIONS
The following recommendations are made in addition to the previously described
standard method for measuring the surface roughness of membrane surfaces using AFM:

The scan area for curved surfaces (i.e., hollow fiber membranes) should not be
larger than 5 m2,

A specially designed AFM tip (e.g., mounted at an angle) that is visible from
above cantilever could be considered for imaging curved surfaces, and

The AFM tip should have enough resolving power to define in detail at least a 2030 nm surface feature, without sticking to the membrane.

151

2012 Water Research Foundation. ALL RIGHTS RESERVED.

REFERENCES
(ASTM), A. I., Standard Practice for Conducting an Interlaboratory Study to Determine the
Precision of a Test Method. ASTM International: 2009; Vol. ASTM E691 - 09e1
Adamson, A. W. and A. P. Gast, Physical Chemistry of Surfaces. Wiley: New York, 1997.
Alklaibi, A. M. and N. Lior, Desalination 2005, 171 (2), 111-131.
Allcock, H. R., L. B. Steely and A. Singh, Polymer International 2006, 55 (6), 621-625.
Alsteens, D., E. Dague, P. G. Rouxhet, A. R. Baulard and Y. F. Dufrene, Langmuir 2007, 23
(24), 11977-11979.
Blom, D. A., J. R. Dunlap, T. A. Nolan and L. F. Allard, Journal of the Electrochemical Society
2003, 150 (4), A414-A418.
Boussu, K., J. D. Baerdemaeker, C. Dauwe, M. Weber, K. G. Lynn, D. Depla, S. Aldea, I. F. J.
Vankelecom, C. Vandecasteele and B. Van der Bruggen, ChemPhysChem 2007, 8 (3),
370 - 379.
Boussu, K., C. Vandecasteele and B. Van der Bruggen, Journal of Membrane Science 2008, 310
(1-2), 51-65.
Brant, J. A., K. M. Johnson and A. E. Childress, Colloids and Surfaces a-Physicochemical and
Engineering Aspects 2006, 280 (1-3), 45-57.
Chapman-Wilbert, M., S. Delagah and J. Pellegrino, Journal of Membrane Science 1999, 161 (12), 247-261.
Childress, A. E. and M. Elimelech, Journal of Membrane Science 1996, 119 (2), 253-268.
Childress, A. E. and M. Elimelech, Environmental Science & Technology 2000, 34 (17), 37103716.
Chung, T. S., J. J. Qin, A. Huan and K. C. Toh, Journal of Membrane Science 2002, 196 (2),
251-266.
Delgado, A. V., F. Gonzles-Caballero, R. J. Hunter, L. K. Koopal and J. Lyklema, Pure and
Applied Chemistry 2005, 77, 1753-1805.
Drelich, J., Journal of Adhesion 1997, 63 (1-3), 31-51.
Drelich, J., J. D. Miller and R. J. Good, The Effect of Drop (Bubble) Size on Advancing and
Receding Contact Angles for Heterogeneous and Rough Solid Surfaces as Observed with
Sessile-Drop and Captive-Bubble Techniques. Journal of Colloid And Interface Science
1996, 179 (1), 37-50.
EPA, U. S., Membrane Filtration Guidance Manual, EPA 815-R-06-009. 2005.
EPA, U. S., 40 CFR Parts 9, 141, and 142 National Primary Drinking Water Regulations: Long
Term 2 Enhanced Surface Water Treatment Rule; Final Rule. Federal Register: 2006;
Vol. 71, pp 653-786.
Ernst, M., A. Bismarck, J. Springer and M. Jekel, Journal of Membrane Science 2000, 165 (2),
251-259.
Gao, L. C. and T. J. McCarthy, Langmuir 2006, 22 (14), 6234-6237.
Garbassi, F., M. Morra and E. Occhiello, Polymer surfaces: from physics to technology. Wiley:
Chichester, New York, 1998.
Hagmeyer, G. and R. Gimbel, Separation and Purification Technology 1999, 15 (1), 19-30.
Jinnai, H., H. Hasegawa, Y. Nishikawa, G. J. A. Sevink, M. B. Braunfeld, D. A. Agard and R. J.
Spontak, Macromolecular Rapid Communications 2006, 27 (17), 1424-1429.
Khulbe, K. C., C. Y. Feng and T. Matsuura, Synthetic Polymeric Membranes. Characterization
by Atomic Force Microscopy. Springer: Berlin, 2008.

152
2012 Water Research Foundation. ALL RIGHTS RESERVED.

Kim, Y., J. T. Park, J. H. Koh, D. K. Roh and J. H. Kim, Journal of Membrane Science 2008,
325 (1), 319-325.
Li, Q. L. and M. Elimelech, Environmental Science & Technology 2004, 38 (17), 4683-4693.
Li, X., S. De Feyt and I. F. J. Vankelecom, Journal of Membrane Science 2008, 324, 67-75.
Mnttri, M., A. Pihlajamki and M. Nystrm, Journal of Membrane Science 2006, 280 (1-2),
311-320.
Mulder, M., Basic Principles of Membrane Technology. 2nd ed.; Kluver Academic Publishers:
1996.
Mller, R. H. and W. Mehnert, Particle and surface characterisation methods. Medpharm
Scientific Publishers: Stuttgart, 1997.
Nystrm, M., L. Kaipia and S. Luque, Journal of Membrane Science 1995, 98 (3), 249-262.
Peeters, J. M. M., M. H. V. Mulder and H. Strathmann, Colloids and Surfaces aPhysicochemical and Engineering Aspects 1999, 150 (1-3), 247-259.
Roudman, A. R. and F. A. DiGiano, Journal of Membrane Science 2000, 175 (1), 61-73.
Schfer, A. I., Natural organic removal using membranes, Ph. D. Dissertation. University of New
South Wales: Sydney, Australia 1999.
Schonhorn, H., Journal of Physical Chemistry 1966, 70 (12), 4086-4087.
Tadmor, R., Langmuir 2004, 20 (18), 7659-7664.
Tay, J. H., J. L. Liu and D. D. L. Sun, Water Research 2002, 36 (3), 585-598.
Wyart, Y., G. Georges, C. Demie, C. Amra and P. Moulin, Journal of Membrane Science 2008,
315 (1-2), 82-92.
Yamamura, H., K. Kimura, T. Okajima, H. Tokumoto and Y. Watanabe, Environmental Science
& Technology 2008, 42 (14), 5310-5315.
Yaroshchuk, A. E. and V. Ribitsch, Langmuir 2002, 18 (6), 2036-2038.
Zhu, X. and M. Elimelech. 1997. Colloidal Fouling of Reverse Osmosis Membranes:
Measurements and Fouling Mechanisms. Environ. Sci. Technol., 1997, 31 (12), pp
36543662

153
2012 Water Research Foundation. ALL RIGHTS RESERVED.

ABBREVIATIONS

AFM
ANOVA
AP

atomic force microscopy


analysis of variance
Anton Parr

CSM

Colorado School of Mines

DU

Duke University

EKA

Electro Kinetic Analyzer

FKKT

University of Maribor

MF
MLR

microfiltration
Multiple Linear Regression

NF

nanofiltration

OCWD

Orange County Water District

PA
PES
PP

polyamide
polyethersulfone
polypropylene

RO

reverse osmosis

SEM

scanning electron microscopy

TEM
TFC

transmission electron microscopy


thin film composite

UCR
UF
UNR

University of California, Riverside


ultrafiltration
University of Nevada, Reno

WaterRF

Water Research Foundation

154
2012 Water Research Foundation. ALL RIGHTS RESERVED.

You might also like