w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/watres

Accurate and self-consistent procedure for determining pH in seawater desalination brines and its manifestation in reverse osmosis modeling Oded Nir a,*, Esra Marvin b, Ori Lahav a a b

Faculty of Civil and Environmental Engineering, Technion e IIT, Haifa 32000, Israel Faculty of Environmental Engineering, Helsinki Metropolia University of Applied Sciences, Finland

article info


Article history:

Measuring and modeling pH in concentrated aqueous solutions in an accurate and

Received 24 March 2014

consistent manner is of paramount importance to many R&D and industrial applications,

Received in revised form

including RO desalination. Nevertheless, unified definitions and standard procedures have

24 June 2014

yet to be developed for solutions with ionic strength higher than ~0.7 M, while imple-

Accepted 2 July 2014

mentation of conventional pH determination approaches may lead to significant errors. In

Available online 11 July 2014

this work a systematic yet simple methodology for measuring pH in concentrated solutions (dominated by Naþ/Cl) was developed and evaluated, with the aim of achieving consis-


tency with the Pitzer ion-interaction approach. Results indicate that the addition of 0.75 M


of NaCl to NIST buffers, followed by assigning a new standard pH (calculated based on the


Pitzer approach), enabled reducing measured errors to below 0.03 pH units in seawater RO


brines (ionic strength up to 2 M). To facilitate its use, the method was developed to be both


conceptually and practically analogous to the conventional pH measurement procedure.


The method was used to measure the pH of seawater RO retentates obtained at varying recovery ratios. The results matched better the pH values predicted by an accurate RO transport model. Calibrating the model by the measured pH values enabled better boron transport prediction. A Donnan-induced phenomenon, affecting pH in both retentate and permeate streams, was identified and quantified. © 2014 Elsevier Ltd. All rights reserved.



The pH value is a parameter of major significance in reverse osmosis (RO) applications. Most significantly, it affects the permeation rate of potentially toxic weak-acid elements, e.g. boron (Tu et al., 2010), NH3 (Hurtado and Cancino-Madariaga, 2014) and arsenic (Teychene et al., 2013) and at the same time controls the development of chemical scaling of minerals e.g.

calcite and brucite (Nir et al., 2012). Additionally, the pH affects the membrane's lifespan (Donose et al., 2013) and may induce a change in its physical properties and thereby in its performance (Wang et al., 2014). Despite this, an accurate measurement procedure and a reliable predictive model for pH in desalination brines are currently lacking in the literature. Regarding the measurement procedure, the knowledge gap is associated with theoretical and practical difficulties arising from standardization of pH in high ionic strength (I)

* Corresponding author. E-mail addresses: [email protected], [email protected] (O. Nir). http://dx.doi.org/10.1016/j.watres.2014.07.006 0043-1354/© 2014 Elsevier Ltd. All rights reserved.


w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5

solutions (Buck et al., 2002). Regarding modeling, the gap is attributed to the high complexity of transport and reaction processes, affecting the evolution of pH in seawater-brines in full-scale RO operations (Nir and Lahav, 2013). To date, a widely accepted definition and a measurement procedure are available for both dilute solutions (I < 0.1 mol kg1) and seawater, but not for seawater desalination brines (approximately twice SW concentration). In practice, pH is regularly measured in desalination feeds and brines by a combined glass electrode, calibrated by standard NIST buffers (NIST stands for U.S. National Institute of Standards, which develops and maintains pH standards; the value measured by this procedure is termed pHNIST). However, this concept, which was developed for dilute solutions (see brief discussion in the supporting material file), encompasses two assumptions which are theoretically invalid for concentrated solutions: (1) In the process of assigning primary pH standards, the activity coefficient of chloride ions can be estimated by the Bates-Guggenheim convention; (2) The liquid junction potential term, appearing in the process of measuring pH using the glass electrode arrangement, is practically cancelled out as a result of the calibration procedure (Buck et al., 2002). As a result of relying on these erroneous assumptions, measuring pHNIST in concentrated solutions invariably leads to considerable errors. Extensive work has been dedicated to measurement, interpretation and standardization of pH in seawater (see brief discussion in the supporting material file). Although high precision pH measurements have been shown feasible by potentiometric and spectrometric methods, interpretation and standardization are still under debate (Marion et al., 2011). Overall, the seawater pH scale approach is based on the relatively constant composition of seawater and therefore cannot be readily extended to desalination brines of varying compositions, nor to seawater brines at salinity >45‰ or pH values significantly different from 8.1 (Millero et al., 2009). The Pitzer approach, has been long recognized as a potential framework within which the definition and traceability of pH values could be soundly extended to higher ionic strengths (Covington, 1997; Ferra et al., 2009; Millero, 2009). Based on statistical mechanics approach, the Pitzer equations for ion activity coefficients take into account both longerange interactions, represented by the Debye-Huckel term, and short-range specific interactions between dissolved species (Pitzer, 1973). Since the establishment of a new pH reference system requires extensive and precise analytical work (which, ultimately, will have to be conducted in institutions responsible for standards) and extensive uncertainty analysis (Spitzer et al., 2011), the progress in this direction is slow. Meanwhile, industries such as desalination or oil&gas are in urgent need for a practical solution, which will allow for improved modeling and better process monitoring and design. Different approaches introduced for pH measurement in NaCl brines include the use of a liquid-junction free cell coupled with ion selective electrode (Knauss et al., 1990), calibration techniques involving acid/base titrations (Mesmer, 1991) and spectroscopic methods (Millero et al., 2009). These procedures were, by and large, not adopted by desalination professionals, probably due to the relatively large disparity between them

and the conventional NIST concept in terms of both measurement complexity and the pH scale employed. The problem of consistency between the measured pH and the applied thermodynamic model was already recognized by Harvie et al. (1984) whose Pitzer-based model formed the basis for many studies on thermodynamic properties of desalination brines. In their work, this issue was addressed by adopting the extended Macinnes convention (i.e. the chloride ion activity coefficient is equal to the mean activity coefficient of a KCl solution of the same ionic strength) for the representation of activity coefficients. The Macinnes scale was set as a default in the implementation of the Pitzer model in the geochemical program PHREEQC (Plummer et al., 1988). Although not considered superior from the thermodynamic standpoint, the extended Macinnes convention is acknowledged the most appropriate due to its higher compatibility with the NIST scale at high ionic strengths, as compared with other conventions (e.g. Bates-Guggenheim) (Harvie et al., 1984). However, model-measurement discrepancies resulting from liquid junction potential are still apparent in determining pH of concentrated solutions (Plummer et al., 1988). The Pitzer model has been increasingly applied to desalination brines mainly for the assessment of scaling tendency (Azaroual et al., 2004; Huff, 2004; Schausberger et al., 2009; Radu et al., 2014; Sousa et al., 2014) and energy requirements (Mistry et al., 2013). In a previous work the writers developed a computer simulation code for predicting the transport and equilibrium state of weak acid-base species within SWRO streams (Nir and Lahav, 2013, 2014). The simulation was based on a reactive-transport approach, i.e. membrane transport equations (solution-diffusion-film model) coupled with elaborated thermodynamic and chemical-equilibrium calculations, facilitated by the use of the Pitzer concept, as implemented in PHREEQC. By performing a full species distribution analysis at each numerical brine recovery step the simulation code enabled the prediction of the pH evolution in the rejected solution as it flowed through a full-scale membrane train. This approach was shown to improve the modeling predictions of boron permeation (Nir and Lahav, 2013), compared to the approach applied in most of the works published thus far on boron membrane transport, in which the hypothesis is that the pH of the reject remains constant from the raw seawater to the brine produced at the outlet of the SWRO step (e.g. Taniguchi et al., 2001; Sagiv and Semiat, 2004; Mane et al., 2009; Park et al., 2012). In fact, considerable differences in pH values were measured at the feed and brine in many SWRO applications (Waly et al., 2011; Andrews et al., 2008). Being a key parameter in acid-base equilibria, the exact knowledge of the pH value throughout the retentate path within the membrane train is imperative in any predictive model, both as an input parameter and throughout the membrane train path, as means of calibrating and assessing model predictions. However, little attention has been thus far given to errors associated with the measured pH in the context of desalination brines. In the current work a heuristic measurement procedure for desalination brines dominated by Naþ/Cl was developed and evaluated. Similarly to Nordstrom et al. (1999), a computer program implementing the Pitzer approach was used to assign pH values to non-standard buffer

w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5

solutions according to the Macinnes convention. Subsequently, the significance of the new measurement procedure to SWRO modeling was demonstrated by comparing simulated and experimental pH values.


Material and methods



pH measurements were made using the Metrohm Aquatrode Plus (6.0257.600) combined glass electrode with integrated Pt 1000 temperature sensor and a Metrohm780 pH meter. Temperature was kept constant at 25 ± 0.6  C with a MRC BL-30 circulating bath. Sample measurements and calibrations were carried out in mixed 25 ml beakers. Certified secondary standard buffers phthalate (pH ¼ 4.01), equimolal phosphate (pH ¼ 6.86), and carbonate (pH ¼ 10.01) from Merck were used for calibration when the pH was measured in the NIST scale (see results in Fig. 1 and part of the results in Fig. 3). For the NaCl-buffers: the abovementioned NIST buffers were prepared in the lab using analytical grade chemicals. Prior to NaCl addition, the mV of the prepared buffers were compared with the certified value (margin of error DmV < 0.2). Chemicals used in the preparation of samples and standards were oven-dried at 60e100  C for a minimum of one hour and stored in a desiccator before weighing. Sodium carbonate was dried at 250  C for a minimum of two hours and stored in a desiccator over CaCl2 salt. Sample and calibration buffers were prepared on a volume (molar) basis and modeled accordingly. Accurate compositions and pH values of the three buffers used are available in the SM files. In the first experimental set (Fig. 1), synthetic SW salts were weighed and added cumulatively to sample solutions, while in the other experimental sets samples and calibration standards were prepared from stock solutions. Synthetic seawater was prepared according to a typical seawater composition (35 g/kg). CaCl2 was not added to high pH carbonate-containing samples to avoid CaCO3(s)/ CaSO4(s) precipitation. Model calculations always followed the


exact composition of the prepared samples. The test solutions (Fig. 4) consisted of 0.0025 M carbonate, 0.001 M borax, synthetic sea-water ranging from 0 to 2.5xSW and HCl ranging from 0 M to 0.01 M. HCl was added to samples using Metrohm 775 Dosimat automatic pipette. All samples were prepared separately in a systematic order: synthetic-SW was added first, then Borax, HCl, dilution with decarbonized water to about 80 ml and then CaCl2 (for low pH samples). Carbonate addition and filling to 100 ml with decarbonized water was carried out immediately before measurement of each sample to minimize atmospheric CO2(g) exchange. RO experiments were performed using a pilot-scale system, supporting a 4” diameter 40” long spiral wound module (Dow SW304040-HRLE). 100 l of real Mediterranean seawater was used as feed. To avoid CaCO3 scaling at high pH and recovery, the seawater pH was first adjusted to pH4 using HCl. Air was bubbled subsequently for ~1 h for reducing the inorganic carbon concentration to ~0.05 M (estimated from base dose required to raise pH back to 7.8). The pH was then raised to pH 9.08 using NaOH. For each recovery ratio, the system operated in a full recirculation mode for ~20 min before sampling. Operating pressure, temperature and cross-flow velocity were maintained at 65 ± 1 bar, 25 ± 1  C and 0.17 ± 0.1 respectively. pHNIST was measured in the brine and permeate solutions using Orion-Ross 8102BN combination electrode with Eutech Instruments pH1500 pH meter. When calibrated by the same buffers, deviations between the two electrodes (Orion-Ross & Metrohm) were 0.999). Concentrations of boron and seawater major ions were measured using ICP-AES.


Software and modeling

Theoretical pH calculations, based on charge balance, were performed using PHREEQC (Parkhurst and Appelo, 1999), a software package developed by the USGS (United States Geological Survey) for modeling hydrogeochemical systems. For speciation of aqueous species PHREEQC offers various

Fig. 1 e Average measured pHX,NIST values of the borax and carbonate standard buffers (n ¼ 3),compared to pH values calculated in PHREEQC using different databases/species activity models, as a function of ionic strength resulting from the addition of seawater salts.


w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5

Fig. 2 e DpH resulting from theoretical liquid junction potential considerations as function of ionic strength upon calibration with I ¼ 0.1 M buffers at T ¼ 25  C. DpH for seawater and NaCl ( 0.7 M) solutions was Calculated according to Eq. 1. Theoretical calculations were carried out using Pitzer's ion activity model with/without Macinnes convention application.

thermodynamic data sets and theoretical approaches. In this work the database pitzer.dat was mainly used. This database implements the Pitzer approach and retrieves thermodynamic data mostly from Harvie et al. (1984). The results of implementing pitzer.dat were compared in this work (Fig. 1) with results emanating from the use of two other database files: (1) minteqV4.dat e implementing an extended DebyeHuckel approach and thermodynamic data compiled by the U.S environmental protection agency; and sit.dat e which implements the Bronsted-Gugenheim approach and uses thermodynamic data compiled by the European nuclear energy agency. The pH of the NaCl buffers used for calibration was also calculated by PHREEQC, using the Pitzer.dat database. Missing Pitzer interaction coefficients (i.e. phthalic acid (Chan et al., 1995; Ferra et al., 2009) and phosphoric acid

(Covington and Ferra, 1994)) were added manually. Ion transference numbers used in Eq. (1) for generating Fig. 2 were adopted from the literature (Della Monica et al., 1979; Poisson et al., 1979; Panopoulos et al., 1986). Our SWRO computer model coupled the solution-diffusion-film model (Python code) and the Pitzer approach, implemented within PHREEQC_COM object (Charlton and Parkhurst, 2011). For detailed model description see Nir and Lahav (2013) and Nir and Lahav (2014). Membrane permeabilities for water (4.18 * 107 m/s/bar), salt (1.66 * 108 m/s) and boric-acid (6.3 * 107 m/s) were determined from independent RO experiments using distilled water and seawater at pH7. The correlation for Sherwood's number (Sh ¼ ARebSc0.25) was used for calculating the mass transfer coefficient used within the film concentration-polarization model (Schock and Miquel, 1987). A and b were estimated from independent experiments to be 0.079 and 0.73, respectively. The measured composition of the seawater feed, which was used as input to the simulation, is given in Table 1.


Results and discussion

3.1. Measuring and modeling pH in highly concentrated solutions The pH measurement approach applied in this work was to minimize liquid junction potential errors by adding NaCl to standard NIST buffers and assigning new pH values to the newly-generated saline buffers using the Pitzer approach. The combined glass electrode was then calibrated with the assigned pH values. This procedure was tested systematically in light of the following criteria: (1) The new measurement technique should be almost identical to the conventional one (i.e. three-point calibration), with the aim of facilitating adoptability by both industry and academia; (2) wide range of salinities and pH values should be covered; (3) The pH calculated by the thermodynamic model used for assigning pH values to the calibration buffers (pHcalc) should be highly consistent with the pH assigned by NIST to standard reference buffer solutions

Fig. 3 e The difference between the calculated (PitzereMacinnes) and the average (n ¼ 3) measured pH using four different sets of calibration buffers, as a function of ionic strength.

w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5


Fig. 4 e Measured (markers) and calculated pH for HClþ0.0025 M Na2CO3 þ0.0025 M NaHCO3þ 0.001 M Na2B4O7:10H2O in synthetic seawater at: SSWx1 (a), SSWx1.5(b), SSWx2(c) and SSWx2.5(d).

(pHS,NIST) in order to maintain continuum pH scale from low to high ionic strengths; (4) The differences between pH values measured on the NIST scale (pHX,NIST) and pHcalc should be as small as possible even at high ionic strengths in order to retain meaningfulness of the pHNIST (which will most likely remain the most widely use method) and allow practitioners to maintain their intuition for pH, thereby encouraging them to practice the new method; (5) The pH measured by the new method, pHX,PM, (PM: PitzereMacinnes) should be consistent with pHcalc. The three chosen standard buffer solutions (see Experimental section) resembled the pH4, pH7 and pH10 three-point calibration system, which is both widely-used and also covers the pH range required in desalination applications (in compliance with the first two criteria). The choice of the buffer weak-acid systems was also induced by the reliability and availability of Pitzer's coefficients characterizing the interactions of the weak acid species with NaCl. Pitzer coefficients absent from the PHREEQC database were adopted from the literature and incorporated manually into the

Table 1 e Seawater feed major ions and weak acid species concentrations (mg/l). Cl





SO2 4


BT (mg/l as B)

CT (mg/l as CO2)







database. As a prerequisite step, the equilibrium model was tested for consistency with standard NIST pH values and compared against other databases embedded in the software, ~ es et al. analogously to the theoretical analysis made by Camo (1997). The pH values reported by NIST were generally the closest to those calculated by the Pitzer model for almost all the temperature range considered (see Fig. 1A on supporting material file), implying high consistency of the Pitzer model and database with the dilute NIST buffers, in compliance with criterion (3). Criterion (4) was tested by adding seawater salts to borate and carbonate NIST buffers, measuring pHX,NIST and comparing the results to the theoretical pH values obtained by different thermodynamic models. As shown in Fig. 1, the differences between calculated and measured pH values significantly increased in all the models except for the PitzereMacinnes model, for which the difference remained relatively constant. A one-tail, unequal variance T-test was performed to compare the average absolute difference, (DpH ¼ jpHX,NIST  pHcalcj) for this model with each of the other three models. The PitzereMacinnes model was found to be significantly closer to the average pH-NIST value. The average error was 0.063 pH units as compared to 0.22 (p ¼ 0.011) pH units for MinteqV.4.dat, 0.18 pH units (p ¼ 0.00038) for SIT.dat and 0.12 pH units (p ¼ 0.0011) for unscaled Pitzer. The lower average DpH obtained with Pitzer.dat compared with SIT.dat and MinteqV4.dat was attributed to both the inherent inaccuracy of the latter models at high ionic strength and liquid junction potential. However, the


w a t e r r e s e a r c h 6 4 ( 2 0 1 4 ) 1 8 7 e1 9 5

lower average DpH obtained for the MacinnesePitzer model, as compared with the unscaled Pitzer model was a direct result of the Macinnes activity convention, which was assumed to decrease the absolute value of the liquid junction potential term appearing in Eq. (1). This relation was examined by calculating the theoretical DpH resulting from an ideal liquid junction potential for both Macinnes and unscaled activity conventions using Eq. (1) (Harvie et al., 1984) KClð3MÞ Z

DpH ¼

DLJP ¼ RT lnð10Þ=F


X tk k



dln ack  S

X tk k


dln ack

lnð10Þ (1)

where: tk is the ion transference number of ion k (i.e. the fraction of electrical current carried by each ion when the solution is subject to an electromagnetic field); aCk is the conventional activity of ion k. The first integral in Eq. (1) refers to the potential of the liquid junction between the measured solution X and the 3M KCl reference filling solution, while the second integral refers to the liquid junction potential forming when the electrode is dipped in a standard solution S (represented here by a 0.1M NaCl solution). The results of this calculation for seawater and NaCl solutions of different salinities and 25  C are presented in Fig. 2. Although, as expected, Eq. (1) did not provide accurate predictions of the experimental DpH (liquid junction potential), the trends appearing in Fig. 2 followed the empirical results, i.e. using the unscaled Pitzer model resulted in increased DpH as a function of ionic strength, while using the Macinnes convention maintained it relatively constant. Another observation obtained from Fig. 2 is the high similarity between DpH for NaCl and DpH for seawater when the Macinnes convention was used. These theoretical considerations suggest that a buffer system containing single NaCl concentration is sufficient for a wide ionic strength range and varying ion compositions when the Macinnes scale is used, thereby corroborating the approach used in this work. The amount of NaCl needed to be added to the NIST buffers was determined according to criterion (5). The glass electrode was calibrated using three NIST þ NaCl buffers at varying NaCl concentration and then used to measure the pH of borax and carbonate buffers in seawater-based solutions of varying concentrations (i.e. SW,1.5, SW,2, etc.). The measurements were compared to the calculated pH in the ionic strength range 0.5e2.0 M, adequate for most desalination applications. Since DpH~0.03 was estimated as the uncertainty associated with the NIST procedure (Buck et al., 2002) it was selected as the accuracy limit of the current measurement approach. As shown in Fig. 3, the difference between measured and modelled pH (DpH) amounted to 0.09 when the standard NIST buffers were used, while calibration with 0.75 M NaCl buffer resulted in DpH < 0.03, with almost all standard deviations inside this range, represented by the dotted lines in Fig. 3. Calibrations with 0.5 M and 1M added NaCl (not shown) produced very similar, but slightly less consistent results. Standard phtalate, phosphate and carbonate buffers comprising additional 0.75 M NaCl were chosen as the threepoint calibration buffer system for measuring pH in

desalination brine streams. The new procedure was tested by measuring the pH of a borate/carbonate mixture in synthetic seawater media (SSW) at four different concentrations (SSW,1, SSW,1.5, SSW,2 and SSW,2.5). DpH < 0.02 was obtained for all SSW concentrations used, indicating high consistency between the model and the chosen measurement procedure. Subsequently, the pH of borate-carbonate-HCl in SSW was measured and compared to the model to evaluate the consistency over a wide pH range. As shown in Fig. 4, the titration curve was reasonably well predicted by the model at all SSW concentrations. Most importantly, at the high pH range (pH8-9), which is the most critical for both boron rejection by RO and CaCO3 solubility, DpH < 0.03 was obtained for all SSW concentrations examined. At the pH range of 6e7 DpH was in the range of 0.03e0.09 while in the low range of 3e5 0.07

Accurate and self-consistent procedure for determining pH in seawater desalination brines and its manifestation in reverse osmosis modeling.

Measuring and modeling pH in concentrated aqueous solutions in an accurate and consistent manner is of paramount importance to many R&D and industrial...
1MB Sizes 0 Downloads 3 Views