You are on page 1of 5

Desalination 261 (2010) 99–103

Contents lists available at ScienceDirect

Desalination
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d e s a l

The effect of feed pH on the performance of a reverse osmosis membrane


T. Hoang, G. Stevens, S. Kentish ⁎
Particulate Fluids Processing Centre, Department of Chemical and Biomolecular Engineering, University of Melbourne, Victoria 3010, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Membranes are predominantly characterized using single salts at one pH. In this paper, a more com-
Received 6 November 2009 prehensive characterization of a reverse osmosis membrane is provided. The zeta potential, flux and
Received in revised form 13 May 2010 rejection behaviour of a brackish water membrane is examined over a wide pH range. Performance in both
Accepted 14 May 2010
single salt solutions (sodium chloride, calcium chloride and sodium sulfate) and a binary solution of calcium
Available online 17 June 2010
and sodium chloride are considered. Results are consistent with the Donnan exclusion theory. That is, larger,
more highly charged ions are rejected more strongly than smaller or less strongly charged species. In
Keywords:
Membrane
particular, this behaviour can lead to hydronium ion rejections of up to −300% as this ion permeates
Reverse osmosis preferentially over other cations. Membrane flux is relatively constant with respect to pH although there is
Nanofiltration some evidence of an increase in flux in doubly charged salt solutions at high pH.
Donnan theory © 2010 Elsevier B.V. All rights reserved.

1. Introduction developed a general theory of multicomponent electrolyte separa-


tions based on the extended Nernst–Plank equation and Donnan
The permeate flux and the salt rejection are the two parameters potentials. Agreeing with this theory, their experimental data showed
usually used to evaluate the efficiency of membranes in desalination. constant values of sodium and chloride rejection for SS10 cellulose
Both these factors are influenced by pH as well as the composition of acetate membrane in sodium chloride solution as pH was lowered
the feed water [1–7]. In contrast to microfiltration and ultrafiltration, until pH 4–4.5, then chloride rejection decreased and sodium rejec-
the separation mechanism of nanofiltration (NF) and reverse osmosis tion increased [4]. Similar results were found for calcium and chloride
(RO) membranes is based not only upon molecular size but also upon rejection in a calcium chloride solution.
ionic charge. Changes to the feed pH may alter the nature of the The aim of this paper is to clarify the effects of feed composition
membrane surface charge, which can subsequently affect the and pH across a wide range on the performance of an RO membrane.
membrane performance [8–11]. The relationship between feed pH The change of permeate flux and both cation and anion rejection is
and the separation capability of NF membranes has been the focus for monitored and interpreted in terms of the membrane surface charge
many researchers over the past twenty years [5–18]. However, much and the Donnan theory of dielectric exclusion.
less work on RO membrane performance has been reported. Ozaki
et al. [19] concluded that the ES20 ultra low pressure RO membrane 2. Experimental
had a high rejection capacity of heavy metals (Ni, Cu and Cr),
which was influenced by pH and the presence of divalent cations such The RO membrane used for this study was a brackish water AK
as Ca2+ and Mg2+. Wagner et al. [1] reported a fairly linear relation- membrane from GE Osmonics. Salt solutions are prepared from AR
ship between sodium chloride salt rejection and feed pH for their RO grade crystalline NaCl, CaCl2∙2H2O and Na2SO4 as supplied by Chem
membranes due to the increasingly negative charge of the polyamide Supply (South Australia). The pH was adjusted with AR grade HCl,
membrane surface. The pH range investigated in this work was from NaOH, Ca(OH)2 and H2SO4 corresponding to the co-ions present in the
5 to 9 and the effect of pH under 5 on RO membrane performance is feed solution. All solutions were prepared to a total ionic strength of
less clear. 0.034 M, corresponding to 2 g/L of NaCl. Deionised water with
It was found by Lipp et al. [2] that the rejection of calcium by a resistivity of 18.2 MΩ cm was used exclusively in all experiments.
seawater membrane in single salt solutions was smaller than that of A schematic diagram of the flat membrane crossflow rig set up for
sodium under similar conditions. This contradicted the Donnan the laboratory test is shown in Fig. 1. The rig consists of a Hydra-Cell
exclusion theory, which argues that the rejection of multivalent G10 pump connected with a 60 L feed tank and three parallel
ions should be higher than that of monovalent species. Hall et al. [3] Sterlitech CF042 membrane cells. The feed solution is pumped from
the feed tank passing through a spiral tube placed in a water bath for
temperature control to the membrane cells. The retentate is circulated
⁎ Corresponding author. Tel.: +61 3 8344 6682; fax: +61 3 8344 4153. back to the feed tank through valves (which are used to adjust the
E-mail address: sandraek@unimelb.edu.au (S. Kentish). system pressure) and flowmeters. The permeate is collected by three

0011-9164/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.desal.2010.05.024
100 T. Hoang et al. / Desalination 261 (2010) 99–103

Fig. 1. Experimental apparatus. Pressure gauges denoted by P.

separate containers placed on an electronic balance (Pioneer Series, PTFE spacer creating a channel between them. An Ag/AgCl electrode
Ohaus, Australia), which transfers the measured weight data to a was mounted at each end of the channel. In a typical experiment,
computer for monitoring the weight change versus time. membrane samples were washed and soaked in deionised water
Membranes were washed and soaked in deionised water overnight before use. Each electrolyte solution was placed into the
overnight before use. They were then installed in the rig and com- feed container and the streaming potential measured as a function
pacted at 1600 kPa for 2 h. Water at pH 6.5 ± 0.2 was run for 2 h of pressure (dU/dp); initially at the unadjusted pH. Acid was then
through the unit at the experimental pressure, then replaced by salt gradually added to reduce the solution pH and the streaming potential
water at the experimental pH. The entire permeate from each was measured with each addition. The acid addition was carried
membrane unit and a small amount of retentate were collected every out very slowly around the isoelectric point (IEP) to ensure a good
30 min and stored for analysis. Six sets of permeate and retentate estimation of this point. After the IEP was passed, the solution was
were collected for every experiment. After completion of the replaced with a fresh supply of the same electrolyte, the membranes
experiment, fresh water was run to wash and clean the membranes. were washed thoroughly and repacked into the cell and the
Unless otherwise stated, the experimental pressure was 800 kPa and measurement was continued with increasing pH using hydroxide
the retentate flowrate was 12.5 L/min across each cell, giving a solution.
crossflow velocity of 2 m/s and a Reynolds Number of 12,000. The relationship between the measurable streaming potential and
Processing temperature was 30 ± 1 °C. the zeta potential is given by the Helmholtz–Smoluchowski equa-
A Varian 720-ES ICP-Optical Emission Spectrophotometer was tion [20]:
used for determination of sodium, calcium and sulfur contents.
Chloride concentration was determined by potentiometry using a dU η L
ζ = ð3Þ
Metrohm autotitrator Titrando Dosino with Ag–Pt electrodes and dp ε:εo Q :R
0.02 N silver nitrate solution. pH and conductivity were measured
using an Orion 720A+ pH-meter and an Orion 3 star conductivity where ζ is the zeta potential, dU/dp is the slope of a plot of streaming
meter. The apparent rejection (Rapp) was obtained from the following potential versus pressure, η is solution viscosity, εo is the permittivity,
equation: ε is the dielectric constant, L and Q are the length and the cross
  sectional area of the capillary system, and R is the AC resistance of
Rapp = 1 − Cp = Cr × 100 ð1Þ the cell.

where Cp is the average value of the element concentration of three 3. Results and discussion
permeate solutions collected and Cr is the retentate concentration at
the same period. 3.1. Validation of testing protocols
The normalized flux was calculated from:
Pre-compaction of the membranes was carried out at high
J = Jsalt = Jwater ð2Þ pressure before new membranes were used. The purpose of this
stage was to attain a stable permeate flux and improve reproducibility
where Jsalt is the mass of permeate collected in 1 min during salt water of the results [21–24]. The period of pre-compaction described in the
circulation and Jwater is the corresponding mass of permeate collected literature has varied from 20 [21] or 24 [22,23] to 48 h [24]. Other
during water circulation. researchers only condition the membranes under normal operating
The streaming potential was measured with a commercial pressure without high pressure pre-compaction [2,3,19,25,26]. In
electrokinetic analyser EKA from Anton Paar GmbH, Austria. Two our case, pre-compaction was performed for 2 h before a 2 h run of
membranes were placed within the rectangular measuring cell, with a water at the standard operating pressure of 800 kPa and then 3 h of
T. Hoang et al. / Desalination 261 (2010) 99–103 101

circulation of 0.034 M NaCl solution. Results showed that the most


stable permeate flux was achieved after pre-compaction at 1600 kPa.
The flux was a little higher after pre-compaction at 1200 kPa but
reduced to the same value over the period of the salt water run. Thus,
compaction at 1600 kPa was chosen for the remaining experiments.
To determine reproducibility, six pieces of the AK membrane
were tested in two runs using the 3-cell crossflow rig under the
standard processing conditions of 0.034 M NaCl, 2 m/s crossflow
velocity and 800 kPa transmembrane pressure drop. Results gave an
average salt water flux of 33.0 L m− 2 h− 1 with a standard deviation
of 0.6 L m− 2 h− 1 and an apparent rejection of 92.2% with a standard
deviation of 0.55%. These values are both lower than the manufac-
turer's specifications which indicate a permeate flux of 44 L m2 h and
a true rejection of 98.5%. However, these manufacturer's guidelines
are established at significantly lower salt concentrations (0.5 g/L Fig. 3. The zeta potential of an AK membrane in various salt solutions as a function of
NaCl) and this will contribute to the discrepancy. pH. The ionic strength of all solutions was 0.002 M.

3.2. Membrane performance


such pH values, the hydrogen rejection reaches a negative value of
3.2.1. Effect of pH on rejection −300% in an NaCl solution.
Fig. 2 illustrates the performance of the AK membrane at pH values
from 3 to 10. For a sodium chloride solution, there is considerable 3.2.2. Effect of ionic charge
variation in the apparent rejection of both sodium and chloride. At pH For a comparable solution of sodium sulfate (Fig. 5), the rejection
10 these rejection values are comparable at 95%. Rejection is also of sodium and sulfate are uniformly high at basic pH, reflecting
stable at this value at pH 8 and then falls steadily until pH 5. The the strong repulsion of the large, doubly charged sulfate ion by the
sodium rejection reaches a minimum value at this point and then rises negatively charged membrane. Due to electrical neutrality, sodium
again while the chloride rejection continues to fall. rejection is also high. As pH is decreased and the membrane charge
This behaviour can be explained by studying the charge properties weakens (Fig. 3), both ions permeate more readily. However, there is
of the membrane. The plot of zeta potential for the AK membrane in no inflexion point in rejection observed for this system at low pH. This
sodium chloride solution is shown in Fig. 3. The membrane is may reflect a lower IEP for this case. While Fig. 3 suggests an IEP of
negatively charged at high pH due to the deprotonation of carboxylic pH ∼ 3.9, other workers have shown that the IEP within the membrane
groups and adsorption of hydroxide ions on the membrane surface. pores can differ from that determined from zeta potential measure-
The charge changes only marginally when pH decreases throughout ments, which only measure surface charge [9,27,28]. In addition, at
the basic and neutral areas, but rises rapidly below pH 6 to reach the lower pH some of the sulfate ions will be protonated to form HSO− 4 .
isoelectric point (IEP) at pH 4.3. As pH decreases further, the zeta Fig. 6 shows the expected speciation within the solution as a function
potential becomes positive through the protonation of amine groups. of pH. This data has been simulated using the Electrolyte NRTL
The negative charge at high pH repulses chloride ions resulting in thermodynamic model within Aspen Plus© Version 2006.5. The more
a high chloride rejection. The sodium rejection is also high to main- weakly charged HSO− 4 species will permeate more readily leading to
tain the charge balance of the components in the solution. As the lower rejection.
isoelectric point is approached, rejection of both species falls, as these The pH dependence of the calcium rejection for the AK membrane
charge effects are diminished. At pH values below this point, the in a calcium chloride solution is similar to the sodium rejection in a
positive charge of the membrane repels the sodium ions, giving it a sodium chloride solution (Fig. 5). High values of calcium and chloride
high rejection. However, unlike the trend in the basic zone, chloride rejection are achieved at high pH. They reduce gradually until pH 5,
rejection continues to decrease. In this case, the charge balance is drop suddenly at pH 4.5 then separate, with the calcium rejection
maintained by the permeation of the smaller hydronium ions through rising steeply and the chloride rejection decreasing. However, all
the membrane. This is evident from Fig. 4 which shows that at rejection values are higher than in NaCl solution. These results are
quite different from a report by Lipp et al. [2], who found that the
rejections of sodium and chloride in a sodium chloride solution for a

Fig. 2. The effect of pH on apparent rejection of ionic species in single salt solutions with
total ionic strength of 0.034 M. Crossflow velocity is 2 m/s; transmembrane pressure is Fig. 4. The effect of pH on apparent rejection of H3O+ in solutions of total ionic strength
800 kPa. 0.034 M. Crossflow velocity is 2 m/s; transmembrane pressure is 800 kPa.
102 T. Hoang et al. / Desalination 261 (2010) 99–103

Fig. 5. The effect of ion charge on apparent rejection in single salt solutions. Ionic
strength of 0.034 M, crossflow velocity of 2 m/s and transmembrane pressure of
800 kPa. Fig. 7. A comparison of cation rejection in single and binary salt solutions. Crossflow
velocity is 2 m/s; transmembrane pressure is 800 kPa.

Dow FT-30 membrane were much higher than those of calcium and chloride ion through the membrane to retain neutrality. Conversely,
chloride in a calcium chloride solution. They suggested that calcium the sulfate ion is larger and doubly charged so permeation is less
ions better screened the negative surface charge of the membrane favourable. This is reflected in the hydrogen rejection which remains
than sodium ions and therefore caused a decrease in chloride positive at all pH values. The fall in rejection at low pH reflects the
rejection. Because of electrical neutrality, calcium would also be less increase in HSO− 4 speciation as discussed above.
rejected in a single salt solution. Our results, however, are consistent Fig. 9 displays the comparable rejection behaviour of OH− at pH
with the theory developed by Hall et al. [3], who reported a higher values above 7. In this case, negative rejection behaviour is observed
rejection for calcium chloride solution than that of sodium chloride in sodium sulfate solutions. The OH− ion will permeate in preference
[4]. They argued that the calcium concentration within the membrane to the sulfate ion due to its smaller size and lower charge. Conversely,
should be lower than that of sodium for the same feed concentration in chloride based systems, both anions are more comparable in size
since calcium is sterically larger and has a higher charge [4]. and charge and so positive rejection is observed for both. The
In a solution of mixed NaCl and CaCl2 at the same ionic strength the fall in rejection at high pH values suggests that charge effects are
sodium rejection is comparable to that of the single salt, except that more substantial for the chloride ion than for the OH− as this ion does
the minimum at the isoelectric point is not as prominent (Fig. 7). appear to preferentially permeate when the membrane surface is
Calcium on the other hand, achieves a higher rejection in the mixed strongly charged.
solution. As the smaller, less charged sodium ions will penetrate
through the membrane more readily, calcium ions are rejected more
strongly. The chloride rejection for the mixed salt falls between 3.2.3. Effect of pH on flux
the two single salt solutions, which is to be expected (Fig. 8). For any particular single salt solution, there is little change in
permeate flux throughout the applied pH range (Fig. 10). Childress
3.2.3. Hydrogen rejection and Elimelech also report a relatively constant flux of sodium chloride
In sodium and calcium chloride solutions as well as in their solution over the entire pH range for a nanofiltration membrane with
mixtures, hydrogen rejection has negative values throughout the the exception of a slight peak at the isoelectric point [9]. According to
whole pH range, indicated by the fact that the pH of the permeate is these authors, the peak in flux may be caused by an increase in pore
lower than that of the feed solution. The three curves in Fig. 4 have the size due to conformational changes of the cross-linked membrane
same shape, with rejection becoming less negative as pH increases. polymer structure at neutrality. Another reason may be the increased
These values reflect the chloride rejection results presented in Fig. 8. water permeability due to a decrease in electroviscous effects. When
The mobile and relatively weakly charged hydronium ion follows the

Fig. 6. Speciation of sulfate species in acidic solutions for a sodium sulfate solution of Fig. 8. A comparison of chloride rejection in single and binary salt solutions. Crossflow
0.017 M. velocity is 2 m/s; transmembrane pressure is 800 kPa.
T. Hoang et al. / Desalination 261 (2010) 99–103 103

Acknowledgments

Financial support for this project is provided by the CSIRO


Water for a Healthy Country Flagship and this support is gratefully
acknowledged. The authors would also like to thank GE Water for
providing the RO membrane used in this work and for useful feedback
on the draft manuscript.

References
[1] E.M. Van Wagner, A.C. Sagle, M.M. Sharma, B.D. Freeman, Effect of crossflow
testing conditions, including feed pH and continuous feed filtration, on
commercial reverse osmosis membrane performance, J. Membr. Sci. 345 (2009)
97–109.
[2] P. Lipp, R. Gimbel, F.H. Frimmel, Parameters influencing the rejection properties of
FT30 membranes, J. Membr. Sci. 95 (1994) 185–197.
Fig. 9. The effect of pH on apparent rejection of OH− in solutions of total ionic strength
[3] M.S. Hall, V.M. Starov, D.R. Lloyd, Reverse osmosis of multicomponent electrolyte
0.034 M. Crossflow velocity is 2 m/s; transmembrane pressure is 800 kPa. solutions. Part I. Theoretical development, J. Membr. Sci. 128 (1997) 23–37.
[4] M.S. Hall, D.R. Lloyd, V.M. Starov, Reverse osmosis of multicomponent electrolyte
solutions. Part II. Experimental verification, J. Membr. Sci. 128 (1997) 39–53.
[5] C.V. Chung, N.Q. Buu, N.H. Chau, Influence of surface charge and solution pH on the
an electrolyte solution is forced through a pore with a charged surface performance characteristics of a nanofiltration membrane, Sci. Technol. Adv.
Mater. 6 (2005) 246–250.
there is a back flow of counterions and water in the double layer
[6] Y. Xu, R.E. Lebrun, Investigation of the solute separation by charged nanofiltration
adjacent to the pore surface. This effect is minimized at the IEP where membrane: effect of pH, ionic strength and solute type, J. Membr. Sci. 158 (1999)
charge effects decrease. 93–104.
While all fluxes are comparable at low and moderate pH values, [7] X.-L. Wang, W.-N. Wang, D.-X. Wang, Experimental investigation on separation
performance of nanofiltration membranes for inorganic electrolyte solutions,
the permeate flux in doubly charged salt solutions increases relative Desalination 145 (2002) 115–122.
to NaCl at higher pH. A similar increase in water flux with increasing [8] G. Hagmeyer, R. Gimbel, Modelling the salt rejection of nanofiltration membranes
pH was also reported by Mäntärri et al. [28] for several NF mem- for ternary ion mixtures and for single salts at different pH values, Desalination
117 (1998) 247–256.
branes and justified by expansion of the membrane matrix due [9] A.E. Childress, M. Elimelech, Relating nanofiltration membrane performance to
to repulsion between like-charged dissociated functional groups membrane charge (electrokinetic) characteristics, Environ. Sci. Technol. 34
on the pore walls. It is possible that these effects are accentuated (2000) 3710–3716.
[10] J. Tanninen, M. Nystrom, Separation of ions in acidic conditions using NF,
with doubly charged ions such as calcium and sulfate due to their Desalination 147 (2002) 295–299.
higher charge and their tendency to adsorb onto the pore walls. [11] J.M.M. Peeters, J.P. Boom, M.H.V. Mulder, H. Strathmann, Retention measurements
However, this is in direct conflict with the above argument of of nanofiltration membranes with electrolyte solutions, J. Membr. Sci. 145 (1998)
199–209.
Childress and Elimelech [9], who suggested that pore expansion
[12] R. Rautenbach, A. Grosschl, Separation potential of nanofiltration membranes,
occurs at the IEP. Desalination 77 (1990) 73–84.
[13] M. Ernst, A. Bismarck, J. Springer, M. Jekel, Zeta-potential and rejection rates of a
polyethersulfone nanofiltration membrane in single salt solutions, J. Membr. Sci.
4. Conclusion 165 (2000) 251–259.
[14] S. Szoke, G. Patzay, L. Weiser, Characteristics of thin-film nanofiltration
For the membrane studied in this work, the rejection behaviour is membranes at various pH-values, Desalination 151 (2002) 123–129.
[15] J.-H. Tay, J. Liu, D.D. Sun, Effect of solution physico-chemistry on the charge
generally consistent with the Donnan exclusion theory. Larger, more
property of nanofiltration membranes, Water Res. 36 (2002) 585–598.
strongly charged ions such as sulfate and calcium are more strongly [16] J.-J. Qin, M.H. Oo, H. Lee, B. Coniglio, Effect of feed pH on permeate pH and ion
rejected than smaller ions such as sodium and hydronium. In rejection under acidic conditions in NF process, J. Membr. Sci. 232 (2004)
153–159.
particular, the smaller size and charge of the hydronium ion lead to
[17] S. Bandini, J. Drei, D. Vezzani, The role of pH and concentration on the ion rejection
rejections as low as −300%. Conversely, membrane flux is only in polyamide nanofiltration membranes, J. Membr. Sci. 264 (2005) 65–74.
marginally affected by pH. A slight increase in flux at high pH in [18] C. Mazzoni, L. Bruni, S. Bandini, Nanofiltration: role of the electrolyte and pH on
doubly charged salt solutions may reflect conformational changes to Desal DK performances, Ind. Eng. Chem. Res. 46 (2007) 2254–2262.
[19] H. Ozaki, K. Sharma, W. Saktaywin, Performance of an ultra-low-pressure reverse
the membrane pore structure in this highly charged environment. osmosis membrane (ULPROM) for separating heavy metal: effects of interference
parameters, Desalination 144 (2002) 287–294.
[20] M. Smoluchowski, Handbook of Electricity and Magnetism, vol. 2, Barth, Leipzig,
1921.
[21] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Fouling of reverse osmosis and nanofiltration
membranes by humic acid — effects of solution composition and hydrodynamic
conditions, J. Membr. Sci. 290 (2007) 86–94.
[22] W.S. Ang, M. Elimelech, Fatty acid fouling of reverse osmosis membranes:
implications for wastewater reclamation, Water Res. 42 (2008) 4393–4403.
[23] W.S. Ang, M. Elimelech, Protein (BSA) fouling of reverse osmosis membranes:
implications for wastewater reclamation, J. Membr. Sci. 296 (2007) 83–92.
[24] R.D. Cohen, R.F. Probstein, Colloidal fouling of reverse osmosis membranes,
J. Colloid Interface Sci. 114 (1986) 194–207.
[25] M.A. Zazouli, S. Nasseri, A.H. Mahvi, M. Gholami, A.R. Mesdaghinia, M. Younecian,
Studies on rejection and fouling of polyamide reverse osmosis membrane in the
treatment of water solutions containing humic acids, World Appl. Sci. J. 3 (2008)
434–440.
[26] J.S. Louie, I. Pinnau, I. Ciobanu, K.P. Ishida, A. Ng, M. Reinhard, Effects of polyether–
polyamide block copolymer coating on performance and fouling of reverse
osmosis membranes, J. Membr. Sci. 280 (2006) 762–770.
[27] G. Hagmeyer, R. Gimbel, Modeling the rejection of nanofiltration membranes
using zeta potential measurements, Sep. Purif. Technol. 15 (1) (1999) 19–30.
[28] M. Mänttäri, A. Pihlajamäki, M. Nyström, Effect of pH on hydrophilicity and charge
and their effect on the filtration efficiency of NF membranes at different pH,
J. Membr. Sci. 280 (2006) 311–320.
Fig. 10. Variation in normalized flux as a function of pH in single salt solutions.
Crossflow velocity is 2 m/s; transmembrane pressure is 800 kPa.

You might also like