[membrane science and technology] ion exchange membranes - fundamentals and applications volume 12...

Electro-deionization

Chapter 4

4.1. OVERVIEW OF TECHNOLOGY

Electro-deionization (EDI) was applied to concentrate radioactive aqueouswastes at first by Walters et al. (1955) in 1955. Kollsman, (1957) performed someexperimental work for de-ionizing water in 1957. However, the EDI technologywas not commercialized until the late 1980s. It is now applied to the supplies ofpurified water (PW) in power generation, the electronics, pharmaceutical, foodand beverage industries. EDI is essentially a mixed-bed (MB) deionization processbuilt into an electrodialysis (ED) system that continuously regenerates the ionexchange resin using electrical power. The process produces high purity water inthe range of 8–17MO from feed water, which contains 1–100mg l1 of totaldissolved solids (TDS). The electrically generated hydrogen and hydroxyl ionsrecombine to form water in the concentrate and produce no extra salts. EDI hasthe advantage of being a continuous process with constant stable product quality,which is able to produce high purity water without the need for acid or causticregeneration. In a process train in which reverse osmosis (RO) is followed by EDI,water can be produced that is comparable to MB ion exchange treated water.

Ganzi et al. (1987) show a schematic of the EDI process as in Fig. 4.1. Inthis figure, an applied DC potential is denoted by (+) and (). Here, ionexchange membrane sheets are represented by the vertical lines labeled in termsof their ionic permeability. These membranes, used as barriers to bulk waterflow, define district compartments, through which liquid streams containing ions(Na+ and Cl ions) can flow tangentially. A DC electrical potential, applied byan external power supply, causes the transfer of ions to occur. In the dilutingcompartment, the space between membranes is filled with cation and anionexchange resins (or resin fibers).

The transfer of ions is represented schematically by arrows. Ions enteringthe diluting compartment react with the ion exchange resin. Ions then transferthrough the resin in the direction of the potential gradient. Ions simultaneouslytransfer across the membranes, maintaining neutrality in all compartments.Because of the permselective properties of the membranes and the directionalityof the electrical potential gradient, ions in the solution become depleted in thediluting compartment and become concentrated in adjacent (concentrating)compartment.

The use of ion exchange resins in the diluting compartment is the key tothe process. One reason for this is that without the ionic conductivity affordedby the resin, ion transfer does not occur at a practical rate for most fresh-water

DOI: 10.1016/S0927-5193(07)12018-0

dx.doi.org/10.1016/S0927-5193(07)12018-0.3d

Enhancedtransferregime

Electroregenerationregime

Anionpermeablemembrane

Cationpermeablemembrane

Na+

Cationexchanges

Na+

Cl−

H+

H+

H+

OH−

FeedAnion

permeablemembrane

Cationpermeablemembrane

Cl−

Cl−

Cl−

OH−H+

OH−

Concentratingcompartment

ConcentratingcompartmentDiluting

compartment

ConcentrateProduct

Concentrate

Anionexchanges

Na+

Na+

Figure 4.1 Schematic of IONPURE process (Ganzi et al., 1987).

Ion Exchange Membranes: Fundamentals and Applications438

source. Without the ion exchange resins, the solution in the diluting compartmentis deionized and its electrical resistance increases to such an extent that usefulelectrical current transfer is outweighed by membrane inefficiencies, resulting inback-diffusion of ions from the concentrating to diluting compartments. By fillingthe ion exchange resins in the diluting compartment, the conductivity of thecompartment is increased and enhanced ionic transport rates are achieved in theenhanced transfer regime depicted in Fig. 4.1. In this regime, the resins remain inthe salt forms and the rate limiting step is most often film-diffusion of the ions inthe bulk solution to the ion exchange resin surfaces. The film-diffusion leads tolimiting current density under which the electric current is limited by the diffusionof counter-ions to a resin surface and co-ions away from a resin surface.

A second reason for the need for resin in the diluting compartment relatesto water dissociation reaction generated in the electro-regeneration regime inFig. 4.1. In this regime, high potential gradients are created at the interfacebetween the cation exchange resins and the anion exchange resins or at theinterface between the cation (anion) exchange resins and the anion (cation)exchange membranes. Significant amounts of H+ and OH ions are produceddue to the water dissociation (water splitting) and the resins become electro-chemically regenerated to the H+ and OH forms. The H+ ions and OH ions

Electro-deionization 439

produced in this regime are transported toward the electrodes under the appliedpotential field. They recombine in the desalting or the concentrating compart-ments and regenerate water. Under these conditions, the resin acts as a contin-uously regenerated MB ion exchange column, exchanging H+ ions and OH ionsin stoichiometric amounts with salts in the solution. EDI devices operating inthe electro-regeneration regime can remove weakly-ionized compounds. Thesephenomena account for the ability of the EDI process to efficiently de-ionizewater to the 10 to 0.1microsiemen cm1 region of electrical conductivity (Ganziet al., 1997).

Allison (1996) suggests that the feed water to the EDI process must havevery low levels of fine suspending matter because the ion exchange resin essen-tially acts as a media filter and there is no developed method to backwash themedia. This means extensive pretreatment such as ultra-filtration or RO isneeded to assure continuous low levels of the suspending matter. Stack repairsare also more difficult. Stacks are filled with resin after assembly, but there is noeffective means to unload the resin prior to disassembly. Each compartmentmust be cleaned to remove the resin beads (or resin fibers) prior to re-assembly.There are no real advantages and there are these disadvantages over ED whennon-water splitting operation is considered. When EDI is operated in the watersplitting mode, the efficiency of utilizing power for desalting is low. Typicallyonly 10–20% of the applied DC current transports ionic salts. The rest of thecurrent splits water. With the low power utilization efficiency, the process isreally practical only on low TDS waters such as RO permeate with a TDS in therange of 100mg l1 or less.

4.2. MASS TRANSFER IN THE EDI SYSTEM

In order to understand the mass transport phenomena arising in the EDIsystem, Glueckauf (1959) discussed the two-stage ion transport process in thedesalting compartment. The first concerns the diffusion transfer from the flowingsolution into the ion exchange resin particles. The second concerns the transfer ofthe ions along the ion exchange particle chains from the interior of the compart-ment to the ion exchange membranes. Matejka (1971) reported the experimentaland theoretical work that advanced the theory of ionic transport in the EDIsystem.

On the other hand, Verbeek et al. (1998) proposed the following conceptexpressing the mass transport in the EDI system as presented in Fig. 4.2. In thissystem, the anode compartment and cathode compartment are filled respectivelywith cation exchange resins and anion exchange resins. Accordingly, the watersplitting in this system occurs only at the interface between the electrodesand ion exchange resins. It is therefore possible to describe the physicochemicalprocesses in the cell without terms such as water dissociation described inSection 4.1. Further it is possible to present the fluxes for each ions in the liquid

CM AM

Concentratecompartment

Anionexchange

compartment

Cationexchange

compartment

OH−

OH−

H2

−X−

Deionized water: 55 nS/cm

+O2

H+

H+

M+

RO: 10 µS/cm

Concentrate: 200 µS/cm

Figure 4.2 Mass transport in the EDI system (Verbeek et al., 1998).


(solution) phase and the solid (ion exchange resin) phase in a differential balanceelement as illustrated in Fig. 4.3, in which both phases are connected throughthe opening marked with a circle.

In Fig. 4.3, ions i (counter-ion denoted by the subscript (i) are assumed tobe supplied upward with the feeding solution from the inlet of the element. Theyflow in the liquid phase and flow out to the outside of the system. The ionicconcentration change in the element is expressed by the following partial differ-ential equation expressed on the z-axis of the space coordinate in the direction offluid flow.

@Ci

@t¼ v

@Ci

@z aS

1

Ji (4.1)

Here, Ci is the counter-ion concentration (mol l1) in the liquid phase, t time,aS (m2 per m3) the specific surface of the ion exchange resin, e the porosity ofion exchanger bed, and Ji (molm2s1) the flux of ion i passing through theopening.

We define the counter-ionic flux in the solid phase J i from the left handside in the element, passing through the solid phase, taking in Ji from theopening, and flowing out at the right hand side of the system. These phenomenaare expressed by the following partial differential equation expressed on the

z

xCi

Ji +

Solidphasedx

JiJi

dz

liquidphase

Ci + dZ

dx∂x∂Ji

∂Ci

∂Z

Figure 4.3 Material balance in a liquid and a solid phase (Verbeek et al., 1998).


x-axis of the space coordinate in the direction of external electrical field, takingCi as the counter-ionic concentration in the solid phase.

@Ci

@t¼ aSJi

@J i

@x(4.2)

The counter-ions i are assumed to be supplied from the liquid phase to theopening along the z-axis perpendicular to the surface of the solid phase (ionexchange resins filled in the compartment). So, we have the following Nernst–Planck equation established on the z-axis drawn in the liquid film formed on thesurface of the ion exchange resins for counter-ions i and co-ions j as follows:

Ji=j ¼ Di=j

@Ci=j

@z zi=jCi=jui=j

@c@z

(4.3)

in which, Di/j, zi/j and ui/j are respectively the diffusion constant, charge numberand mobility of the ions i or j in the solution film. For neutral species k, the fluxis expressed by the following diffusion equation.

Jk ¼ Dk

@Ck

@z(4.4)

Next, the following ions are considered to be important in ultrapure water(UPW) production: H+, Na+, K+, NH4

+, Ca2+, Mg2+, OH, Cl, NO3,


SO42, HCO3, CO3

2, Si(OH)O, and in addition, dissolved neutral CO2,Si(OH)4. Some of these components are coupled by the following chemicalreactions with the corresponding reaction constants:
Water dissociation
H2O2Hþ þOH KW ¼ 1014 (4.5)
Carbonate system
CO2 þH2O2HCO3 þHþ KC;1 ¼ 106:43 (4.6)

HCO3 2CO2

3 þHþ KC;2 ¼ 1010:33 (4.7)
Silicate system
SiðOHÞ42Hþ þ SiOðOHÞ3 KS ¼ 109:8 (4.8)

The co-ions j are not absorbed by the ion exchanger and they do not participatein a chemical reaction, so their net fluxes Jj become zero:

Jj ¼ 0 (4.9)

There is no net current in the liquid film formed on the surface of ion exchangeresins:

Xmi¼1

ðziJiÞ þXnj¼1

ðzjJjÞ ¼ 0 (4.10)

In addition, the rule of electro-neutrality applies at every position of thesolution:

Xmi¼1

ðziCiÞ þXnj¼1

ðzjCjÞ ¼ 0 (4.11)

Finally, the main driving force for ion transport in the solid phase is the externalelectric field. Diffusion can be neglected as the resulting fluxes are much smallerthan the flux caused by migration. In this situation, the counter-ion flux in thesolid phase J i is expressed by

J i ¼ ziuiCi

@c@x

(4.12)

The total electric current I and external field gradient @c/@x are given by Ohm’slaw as follows:

@c@x

¼I

F ð1 Þb=d R l

z¼0

R d

x¼0ðSni¼1z

2i uiCiÞdxdz

(4.13)


Here, b, d and l are respectively the width, the thickness and the length of thecompartment.

The EDI process expressed by the equations described above was calcu-lated using the computer program. The results are exemplified as follows:

Fig. 4.4 shows the dynamic Na+ ion concentration changes in the courseof time in the cation exchange compartment filled with totally regenerated cationexchange resins (H+ type). Part (a) is calculated without continuous electro-chemical regeneration, indicating H+ ions to be replaced by Na+ ions. Part (b)is calculated with continuous electrochemical regeneration due to water splitting,indicating the decreasing of Na+ ion concentrations.

10

8

6

4

2

00 0.1 0.2 0.3

z (m)

(a) without electrochemical regeneration

CN

a (1

0−5M

)C

Na

(10−5

M)

10

8

6

4

2

00 0.1 0.2 0.3

z (m)

(b) with electrochemical regeneration

Figure 4.4 Na+ ion concentration changes in a cation exchange compartment (Verbeeket al., 1998).

3.0

2.5

2.0

1.5

1.0

0.5

00 0.05 0.10 0.15 0.20

z (m)

(b) in a solid phase

IIIIII

Ci (

M)

0 0.05 0.10 0.15 0.20z (m)

I II III

(a) in a liquid phase

0

1

2

3

4

5

Ci (

10−5

M)

= Na+ = Ca2+ = H+

Figure 4.5 Ionic concentration profiles in a cation exchange compartment (Verbeeket al., 1998).


Fig. 4.5 shows the ionic concentration profiles in a cation exchange com-partment in a steady state in case of abnormally increased calcium concentrationin a feeding solution (e.g. due to failure in the softening stage). In the first part ofthe compartment, part I, the cation exchange resins are selectively loaded withCa2+ ions because of high selectivity for Ca2+ ions of the cation exchanger asshown in (b). Accordingly Ca2+ ion concentration in the solution is graduallydecreased, however, Na+ ion concentration is unchanged (a). In the next part,part II, Na+ ions in the solution are gradually loaded to the cation exchangeresins, so Na+ ion concentration is decreased in the solution (a) and increased inthe resins (b). However, Na+ ion concentration in the resins begins to decreasevia the peak due to the Na+ ion concentration decrease in the solution (b). Inthe final part, part III, the concentrations of both Na+ ions and Ca2+ ions are


decreased and instead of them the concentration of H+ ions is increased (due towater splitting) in the solution and in the resins (a, b).

The reasonability of the changes in Fig. 4.5 is confirmed experimentally(Neumeister et al., 2000).

4.3. STRUCTURE OF THE EDI UNIT AND ENERGY CONSUMPTION

The structure of the EDI unit is illustrated in Fig. 4.6 (Deguchi andKaribe, 1998). Based on the principle of ED, cation exchange membranes andanion exchange membranes are arranged alternately. Desalting cells and con-centrating cells incorporated with spacers are arranged between the membranes.Cation exchange resins (or resin fibers) and/or anion exchange resins (or resinfibers) are filled in the desalting cells. A cathode chamber and an anode chamberare placed at both outsides of the cell arrangement and a direct current is passed.The electric current is expressed as:

I ¼QF ðN t NpÞ

nZ(4.14)

where I is the direct current (A), Q the feeding solution flow rate (ml s1), Nt theion concentration in the feeding solution (eq l1), Np the ion concentration in theproduct solution (eq l1), n the constant, F Faraday constant and Z the currentefficiency.

Energy consumption is

P ¼ I2 Rm (4.15)

where P is the direct current electric power (W), Rm the electric resistance of thesystem (O).

Ebara Corporation developed an EDI (GDI) apparatus with ionexchange nonwoven fabric (IEN) and ion conducting spacer (ICS) (Akahori

Feeding solution

Desalted solution

Concentrated solution

Cathode

−+

Anode

A C A C A C A

A = Anion exchange membraneD = Desalting cell

c = Cation exchange membraneC = Concentrating cell

C A C A C A C A C A C A C A C A C

D C D C D C D C D C D C D C D C D C D C D C D

Figure 4.6 EDI cell arrangement (Deguchi and Karibe, 1998).

20

18

16

14

12

10

8

6

4

2

0

Electric power consumption (Wh/m3)0 100 200 300 400 500 600

100

90

80

70

60

50

40

30

20

10

0

SiO

2 re

mai

ning

rat

io (

%)

Spe

cific

res

ista

nce

(MΩ

cm

)

Figure 4.7 Energy consumption vs. electric resistance and silica removal in the GDIprocess (Akahori and Konishi, 1999).


and Konishi, 1999). IEN and ICS are prepared by radiation graft polymer-ization and placed between ion exchange membranes. They reduce cell pairresistance and promote water dissociation. Fig. 4.7 gives electric resistanceand silica removal plotted against energy consumption in the GDI process,showing that UPW (over 17.5MO cm) is produced at energy consumption of250Wh m3.

4.4. WATER DISSOCIATION IN AN EDI PROCESS

Fig. 4.8 is a schematic, sectional view through an EDI apparatus, illus-trated ion flow direction through an ion-depleting and ion-concentrating com-partment. DiMascio and Ganzi (1999) suggest that the surfaces of ion exchangeresins and ion exchange membranes are in contact with each other, and form theresin/membrane interfaces and resin/resin interfaces at which water splittingreactions arise at over limiting current density. They discussed the auto-catalyticinfluence of functional groups (quaternary ammonium groups and tertiary alkylamine groups) in an anion exchange membrane and anion exchange resin to theintensities of the water dissociation reaction. The feature of the water dissoci-ation generated in the EDI system considerably resembles to that in the bipolar

AM CR AR CR CM AM

Anode Ion depleting compartment Ion concentrating Cathodecompartment

CM: Cation exchange membrane AM: Anion exchange membraneCR: Cation exchange resin AR: Anion exchange resinJH: H+ ion flux JOH: OH− ion flux H2O: H2O regeneration ratea: Water dissociation generation at a resin/membrane interfaceb: Water dissociation generation at a resin/resin interfacec: H2O regeneration

JOH

σ σ

σ

JH JH JH JOH

ca b c

JOH+ −

H2OH2O

Figure 4.8 Water dissociation reaction in an EDI system.


membrane ED system described in Chapter 3. Here, we discuss the mechanismof the water dissociation reaction in the EDI system.

Water dissociation reaction is presented by the following equation and it isgenerated in the water dissociation layer formed at the resin/membrane interface‘‘a’’ and the resin/resin interface ‘‘b’’ in Fig. 4.8 as suggested by DiMascio andGanzi (1999).

H2O3ka

kbHþ þOH (4.16)

ka and kb are respectively forward and reverse equilibrium reaction rateconstant.

The water dissociation reaction produces the H+ ion flux JH and OH

flux JOH in the water dissociation layer formed at the resin/resin interfaceand membrane/resin interface. The H+ ions and OH ions are recombined at‘‘c’’ at the outside of the water dissociation layer to regenerate H2O. The H2Oregeneration rate sH2O is equivalent to JH and JOH.

JH ¼ JOH ¼ sH2O (4.17)

Applying the discussions on the water dissociation reaction in the bipolarmembrane ED, the mechanism of the water dissociation in the EDI system isunderstood with the illustration in Fig. 4.9 (cf. Fig. 3.10 in Section 3.3.2),showing the intermediate layer to be formed between the resin and membrane.

Waterdissociation

layer

k A

H+ H+OH- OH-

lK lAL

Cathode Cationexchange membrane

Intermediatelayer

Anionexchange

resin

Anode

l : Thickness of the water dissociation layerL: Thickness of the intermediate layer

Figure 4.9 Water dissociation layer formed in an EDI system.


The similar illustration is conceivable between the resin and resin. The thicknessof the intermediate layer L is estimated to be L ^ 107m according to theatomic force microscope (AFM) images observed by Sata et al. (1995) (cf.Section 3.3.1, Fig. 3.7).

Water dissociation reaction is generated in the water dissociation layerformed in the intermediate layer. Electric current efficiency Z of the waterdissociation reaction is expressed as follows (cf. Section 8.10, Eq. (8.53) inFundamentals) (Tanaka, 2007).

i

F

Z ¼ ðkaCH2O kbC

0HC

0OHÞl (4.18)

Electric current efficiency in the electro-deionization process ZE is expressedby Eq. (4.19), which is introduced in the bipolar ED process (cf. Section 3.3.2,Eq. (3.13)).

ZE ¼ ZK þ ZA ZKZA (4.19)

4.5. REMOVAL OF WEAKLY-IONIZED SPECIES IN AN EDI PROCESS

Weakly-ionized species such as silica, carbon dioxide, boron and ammoniaare difficult to remove through the membrane process such as RO, electrodialysisreversal (EDR) etc. However, EDI removes these species effectively. Hernon et al.(1999) discussed how the EDI process successfully removes various weakly-ionized species as follows.

- - - -Cathode (-)

Ca++ Mg++ Na+ H+

High-Purityproduct water

OH−

++++ Anode (+)

Anion-Exchange membrane

HSiO3−HCO3

−Cl−SO4= CO3

=

CO2

SiO2Ca++

Cl−Na+

SO4=

HCO3−

Cation-Exchange membrane

Figure 4.10 Basic flow scheme of the EDI process (Hernon et al., 1999).


4.5.1 Effect of Water Dissociation in an EDI Process

Fig. 4.10 illustrates the basic flow scheme of EDI process. In EDI, the ionexchange resins facilitate mass transfer of weakly-ionized species mainly due towater splitting. Namely, in the diluting cell, the DC electrical field splits water atthe surface of the ion exchange resin beads, producing hydrogen and hydroxylions which act as continuous regenerants of the ion exchange resin. This allows aportion of the resins in the EDI to always be in the fully generated state. In turn,the fully generated resins are able to ionize weakly-ionized species. Once ionized,these species are quickly removed under the influence of the DC electrical field.Fig. 4.10 roughly depicts how different ions are removed as water travelsthrough the EDI diluting cell, strongly-ionized ions being removed first in theflow-path and weakly-ionized species removed as the water moves down theflow-path. EDI’s performance in removing these weakly-ionized species will thusassuredly leads to greater use of EDI in the production of UPW.

4.5.2 Silica Removal in an EDI Process

To further explore the performance limits of EDI on weak ion removal,Hernon et al. (1999) conducted the following laboratory study, focusing on theeffect of flow rate and amperage on silica removal in a full-scale, 50 gpm unit.EDI feed water for this test contained 20 ms cm1 of NaCl and 0.5 ppm of CO2

when the feed water was treated with RO. As shown in Fig. 4.11, silica removaldecreases with increasing flow rate for constant amperage. Increasing flow rateresults in lower residence time, more utilization of available current for strongion transfer, and less water splitting available for weak ion removal. However,even at velocities that exceed nominal velocities by 50%, silica removal remainedabove 95%. Fig. 4.12 shows the effect of amperage on silica removal for aconstant flow. Initially, silica removal increases with increasing amperage due tohigher extent of resin/membrane regeneration; however, it levels off at a certain

Feed conductivity: 20micros/cm, CO2=6 ppm

100

98

96

94

92

90

Sili

ca r

emov

al (

%)

Flow rate (gpm)40 50 60 70 80

Figure 4.11 EDI performance at varying flow rate (Hernon et al., 1999).

Sili

ca r

emov

al (

%)

90

92

94

96

98

100

1.00.80.6 1.2I /I0

Feed conductivity=20micros/cm, CO2=6 ppm

Figure 4.12 EDI performance at varying current density (Hernon et al., 1999).



current I0 which is a function of flow rate and feed composition. Operation atcurrent densities higher than I0 does not impact silica removal any further as thelimiting factor at this point for silica removal is the residence time.

The above studies confirm that the degree of weak ion removal in anEDI unit is tied to the degree of water splitting occurring within it. The abovestudies also provide a basis for calculating design conditions for the removal ofweakly-ionized species.

4.5.3 Carbon Dioxide Removal in an EDI Process

Carbon dioxide removal by EDI is an important facet of its performance.When carbon dioxide is present in an ion exchange feed stream, the carbondioxide competes with silica for ion exchange sites on an anion resin. Carbondioxide often represents the largest anion load on an ion exchange system,especially when the ion exchange unit is preceded by an RO unit. The presenceof excess amounts of carbon dioxide both limits the capacity of ion exchangeresin to remove silica and limits the efficiency of silica removal by the ionexchange bed. As will discuss in Section 4.6.1 silica removal is critical in bothpower generation and in semiconductor production. The presence of an excessamount of carbon dioxide in effect exposes a plant to potentially serious silicaproblem. Carbon dioxide cannot normally be removed via other membranedemineralization process such as RO and EDR unless chemical adjustmentsare made to change alkalinity levels. EDI on the other hand, is a membranedemineralization process routinely reduces CO2 levels by over 99% in mostapplications (cf. Table 4.4).

4.5.4 Boron Removal in an EDI Process

Boron is often present in water and is problematic as it can cause defectsin semiconductor chip manufacturing. Boron is not well removed by the ionexchange resin process due to its poor ionization and low selectivity (Yagi et al.,1994). Table 4.1 shows removal of boron by both RO and EDI at four plants.The RO units typically remove only a small portion of the boron from the feedwater. However, the EDI units consistently remove over 96% of the boron in thefeed water.

4.5.5 Ammonia Removal in an EDI Process

The ammonia study was part of a largest project addressing the develop-ment of total water-recycle systems for Space Station. In a self-contained lifesupport system, such as the future-planned Space Station, the ability to efficientlyrecycle water is key to long-term mission success. It is projected that a four-personteam will use a supply of 225 lbs of water for general use during a space mission toMars and back. All water used will need to be recycled through an intricaterecovery system and treated through a sequence of processes to produce drinkingwater. Ammonium ions are generated as a decomposition product of urea and

Table 4.1 Boron levels and removals at various sites

PowerPlant #2

SemiconductorPlant #2



RO feed boron (ppb) 22 280 110 85RO product boron (ppb) 14 170 83.5 64.5EDI feed boron (ppb) 14 71 83.5 64.4EDI product boron (ppb) 0.45 2.75 2.8 0.74RO boron removal (%) 36.3 39.3 24.1 24.1EDI boron removal (%) 96.8 96.1 96.6 98.9RO feed pH 6.5 7.7 7.8 7.9EDI feed pH 5.7 6.5 6.4 6.4

Source: Hernon et al. (1999).


over time will build up in concentration, thus rendering the water unsuitable forhuman consumption. EDI, being one of the purification steps, is required toreduce the level of ammonia from 200ppm down to the range of 0.5–1ppm.

Fig. 4.13 shows the percent ammonia removal by EDI. EDI operationwas optimized in terms of flow rate and current. The system was operated in asemi-continuous mode with a continuous 200 ppm ammonia feed at 160ml min1.The concentrate stream in turn was continuously recycled from a separate wastetank. The waste tank volume was kept constant by periodically removing a smallamount of water that would make up for any water transfer from the dilute to theconcentrate through the membranes. The amount of periodic waste blowdowncorresponds to a water recovery value of 99.9%. Due to such continuous recycle,the waste concentration increased over time as pictured in Fig. 4.13. Such highconcentration resulted in turn to a slight decrease in ammonia removal due toback-diffusion.

4.6. PRACTICE

4.6.1 Ultrapure Water Production in Electric Power Generation and

Semiconductor Manufacturing Processes

In the nuclear power plants, UPW is used for makeup to high-pressurestream boiler. In the semiconductor plants, the UPW is used for rinsing semi-conductor wafers after various processing steps. Ionic Inc. applied the EDIprocess operating in the following locations (Hernon et al., 1995).

(a)
Grand Gulf Nuclear Station (GGNS) (b) Arkansas Nuclear ONE (ANO) (c) Seabrook Nuclear Station (SNS) (d) A New England semiconductor manufacturing plant (Semi 1) (e) A Midwest semiconductor manufacturing plant (Semi 2) (f) An Arizona semiconductor manufacturing plant (Semi 3)

35000

30000

25000

20000

15000

10000

5000

0

ppm

NH

4+ in

con

cent

rate

NH

4+ r

emov

al (

%)

100

99

98

97

96

95

94

93

92

91

90168 336 504 672 840 1008 1176 1344

Operating hours

pH = 10.3, 99.9% water recovery

NH4+ removal (%)

ppm NH4+ in

concentrate

Figure 4.13 NH4+ removal by EDI (Hernon et al., 1999).


In all these installations, EDI is one building block of a multi-step treat-ment process. The use of various treatment steps in the treatment processdepends largely on differences in feed water characteristics such as TDS, pH,organic load, dissolved gases, temperature and the presence of various unde-sirable constituents, such as trihalomethanes. Table 4.2 summarizes some of thefeed water variables to each of these units as well as the unit processes that areused in each installation.

Fig. 4.14 shows conductivity reductions which is directly related to TDSremoved across the EDI unit at GGNS. In the figure, 100% reduction corre-sponds to the minimum conductivity achievable due to water dissociation. Duringall of the operating terms, the overall conductivity reduction averages over 99.5%.The values in the other stations and plants have been steady in the range of 99%to 99.5+%.

EDI feed water and product water samples were analyzed for sodium,calcium, magnesium, chloride and sulfate. Rates of removal range from one tothree orders of magnitude depending on the specific ion and its feeding waterconcentration. EDI product water quality approaches MB ion exchange quality.Table 4.3 summarizes the results of these analyses.

In power plants high silica levels in the UPW supply to steam generators canlead to silica deposition on the electrical generator turbine blades. CO2 competeswith silica ions for exchange sites, so the amount of CO2 in a feed stream affectsboth bed life and silica removal efficiency in the ion exchange bed. Table 4.4

Table 4.2 EDI installations, raw feed water sources and pretreatment steps

Location FeedSource

Feed TDS(ppm)

Temp.(1C)

Flowrate(gpm)

Stream Flowsheet

GGNS Surface 400 17–25 50 MMF–ACF–UF–EDR–RO–EDIANO Surface 100 10–30 200 MMF–ACF–UF–EDR–RO–EDISNS Well 475 25 150 UF–RO–EDISemi 1 Surface 230 20 65 CF–UF–RO–EDISemi 2 Well 900 15–25 100 MMF–DEGAS–UF–RO–EDISemi 3 Surface 700 20–25 200 UF–RO–EDI


100

99

98

97

96

953000 4000 5000 6000

Operating time (h)

Con

duct

ivity

red

uctio

n (%

)

Figure 4.14 EDI conductivity reduction, GGNS (Hernon et al., 1995).


presents summary of silica and carbon dioxide feed levels and the correspondingpercent removals achieved with EDI at the six sites mentioned above.

4.6.2 Ultrapure Water Production in Pharmaceuticals

Various grades of water are used by the pharmaceutical industry for anumber of applications. Definitions of each water type and quality are providedin the United States Pharmacopoeia (USP). Of this water, water for injection(WFI) and PW are of primary interest to water treatment engineers and equip-ment suppliers. Table 4.5 provides a summary of the numerical interpretation ofthe Pharmacopoeia chemistry limits for this water.

A leading manufacturer of ophthalmic products (eye drops, contact lenscleaner etc.) planned to increase its capacity by establishing a new productionfacility in the continental United States. As water is the main ingredient inophthalmic solutions, the choice of a design for the new plant’s central water

Table 4.3 EDI feed and product ion levels

Ion Na+

(ppb)Ca2+

(ppb)Mg2+

(ppb)Cl

(ppb)SO4

2

(ppb)

GGNS Feed 1560 10 4 58 70Prod 2 o0.5 o0.5 o2 o4

ANO Feed 871 42 7 594 88Prod 3 0.7 o0.5 o2 o4

SNS Feed 1210 165 nd 528 157Prod 2.4 2.5 nd o2 o2

Semi 1 Feed 289 158 nd 140 245Prod o2 o0.5 nd o2 o4

Semi 2 Feed 1710 320 160 2200 1110Prod o2 o0.5 o5 o2 o4

Semi 3 Feed 437 35 25 78 31Prod 4 o3 o5 o2 o4


Table 4.4 Sillica and CO2 levels and removal

Silica CO2

Feed SiO2

(ppb)Product

SiO2 (ppb)Rejection

(%)Feed TotalCO2 (ppb)

ProductTotal CO2

(ppb)

Rejection(%)

GGNS 640 3 >99.5 2920 17 99.4ANO 3830 138 96.4 4780 8 99.8SNS 170 8 95.2 5910 o10 >99.8Semi 1 208 o2 >99.0 2800 o10 >99.6Semi 2 86 o2 >97.7 3260 o10 >99.7Semi 3 165 o2 >98.8 5340 o10 >99.8



purification system was a primary concern. The operational and performancerequirements for the water purification system were as follows (Parise et al., 1990):

(a)
Product quality equal to or better than that specified for USP PW. (b) Product quantity of 16 gpm, to be available on a continuous basis, 24 h
per day, 6 days per week.
(c) Minimize the use of hazardous chemicals and chemical waste discharge. (d) Minimize the opportunity for bacterial growth in the system. (e) Achieve consistent product quality and minimize equipment downtime.
After evaluation of the available water purification technologies, thefollowing RO/Continuous Deionization (CDI) system developed by Ionpure

Table 4.5 Water quality standard (Numerical interpretation of USPC standard)

Constituent Purified Water Water for Injection

pH 5.0–7.0 5.0–7.0Chloride (mg l1) %0.5 %0.5Sulfate (mg l1) %1.0 %1.0Ammonia (mg l1) %0.1 %0.1Calcium (mg l1) %1.0 %1.0Carbon dioxide (mg l1) %5.0 %5.0Heavy metal %0.1mg l1 as Cua %0.1mg l1 as Cua

Oxidizable substancesb Passes USPpermanganate test

Passes USPpermanganate test

Total solids (mg l1) %10 %10Total bacterial count %50CFUml1 %10CFU/100mlPyrogen Non specified %0.25EUml1

Source: Parise et al. (1990).aLimits for other heavy metals may be determined.bLimits for specific oxidizable substances may be determined.

Table 4.6 Purified water system feed quality (Raw water analysis)

Cations ppm Anions ppm

Ca 119.1 OH 0.0Mg 20.4 CO3 0.0Na 31.1 HCO3 92.0K 4.9 SO4 35.1Fe 0.1 Cl 37.1Cu 0.1 NO3 2.2Ba 0.0 F 2.4Sr 0.3Al 0.4

Note: Other dissolved SiO2 ¼ 4.5 ppm; pH ¼ 8.2.Source: Parise et al. (1990).


Technologies Corporation was selected (Parise et al., 1990). The CDI moduleis the EDI system developed by Millipore Corporation. The analysis of themunicipal feed water in Table 4.6 indicated that a single pass RO system, fol-lowed by a single pass CDI unit would meet the client’s final product qualityspecification.

The system design is shown in block format as Fig. 4.15. Raw municipalwater entered the system at 60 psig. And was pumped through dual on-linemultimedia filters to remove medium- and large-sized particles down to a 10 mmlevel. The next process in the system was filtration by activated carbon toremove organic contaminants and chlorine. Following the carbon beds weredual on-line cation exchange softeners to reduce the incoming total hardness to

Storage tanksterile vent filter

Boosterpump

254 nmUV unit

0.2 Micronpost filter

20 gpmto pointsof use

Recirculationpump

Product

CDIRO

system

5 Micronprefilters

Dualsofteners

Dualcarbonfilters

Dualmultimedia

filters

Rawmunicipal

Feedwater

Figure 4.15 EDI system to produce purified water meeting the United States Pharma-copoeia (Parise et al., 1990).


less than 1 ppm. The last step in the pretreatment train was filtration to anominal 1 mm level with polyethylene cartridge filters. The pretreatment trainhad one more design feature to minimize bacterial growth. Whenever the ROsystem was idle, a pump was activated to recycle water around the pretreatmenttrain, from just before the RO inlet, to the carbon filter. This eliminated thestagnant condition that would otherwise occur when the RO was inoperative.

The RO system currently utilizes 1500 square feet of Hi-Flux CP poly-sulfone membrane to remove suspended and dissolved contaminants, and isoperated at 400 psig to produce 20 to 32 gpm of product water. The RO iscurrently operating at 50% recovery, but is capable of operating at up to 70%recovery.

The 20 gpm RO product is next fed to a CDI installation consisting of two-cell-pair CDI modules. The CDI modules remove dissolved ionic contaminants topolish the RO product water up to the required final production resistivity0.50MO cm. The CDI system is producing 16 gpm of final product water. The60-cell-pair stacks are running at 220V and 1.1A each. The final product qualityfrom the CDI system varies with the RO product conductivity. With an RO prod-uct TDS of 17.7ppm, the CDI product resistivity is 1MO cm. With an RO prod-uct TDS of 10.6 ppm, the CDI product resistivity stays between 3 and 4MO cm.The CDI modules operate at a recovery of 80%.

The CDI product water is fed to a storage tank, from which water isre-circulated at 20 gpm through a 254 nm ultraviolet (UV) unit and final 0.2 mmfinal filters, before being sent to the distribution loop.


The multimedia filters are back-washed twice per week. The carbon filtersare steam sterilized and back-washed every two weeks. The softeners areregenerated automatically based on total flow. The RO system is sanitizedmonthly with 100 ppm of sodium hypo-chlorite, and cleaned as needed tomaintain flux, the frequency of which depends on the SDI of the incoming rawwater. No periodic maintenance is performed on the CDI modules.

4.6.3 Economic Comparison Between EDI and Mixed-Bed Ion Exchange

EDI is an effective technology alternative to MB ion exchange polishingfollowing RO. Edmonds and Salem (1998) provided the following economiccomparison of EDI and MB for new plant installation. In this study, a com-parison of three product flow cases (50, 200 and 600 gpm, see Table 4.7)and three feed water TDS cases (low, medium and high) were considered. Thedeionized water quality required was normally defined as greater than or equalto 17MO cm and silica reduction to less than 20 ppb.

Capital costs for the EDI and the MB equipment were obtained and givenin Table 4.8. Installation costs were estimated to be percentage of the totalcapital cost. A factor of 0.2 was used for the EDI systems and a factor of 0.4was used for the MB systems. The major operating costs considered for eachprocess were as follows: Both systems had labor and water use/wastewater costs.Electricity and replacement EDI stacks were other operation costs for the EDI.MB system costs also included chemicals and replacement resins. The followingcost factors were used: operating labor, $40/h; electricity, $0.07/kWh; water,$1/kgal; wastewater, $2/kgal; sulfuric acid, $0.05/lb of 100%; sodium hydroxide,$0.15/lb of 100%; cation resin, $55/ft3; anion resin, $150/ft3 and stack replacementcost, $6300/12.5 gpm product.

A summary of the annualized costs for each case is given also in Table 4.8.The total system annualized costs were developed from the capital equipmentcosts, the operating cost items and installation factors as described above. As theRO product water TDS increases, EDI becomes increasingly cost effective. Atthe high flow, low TDS case (600 gpm, 4.16 ppm), EDI is slightly (10%) more

Table 4.7 EDI/MB feed water quality (ppm)

Constituent Low TDS Medium TDS High TDS

Na+ 1.2 5.27 9.25HCO3

2.68 3.15 3.78Cl 0.28 6.32 11.9CO2 0.9 1.0 1.2SiO 0.5 0.5 0.5TDS 4.16 14.74 24.9

Note: pH ¼ 6.7.Source: Edmonds and Salem (1998).

Table 4.8 Capital equipment cost and system annualized cost

Flow gpm Capital Equipment Cost(1000 $)

System Annualized Cost(1000 $)

50 200 600 50 200 600

High TDS EDI 60 150 385 30.8 83.6 177.8MB 152 216 397 73.8 120.9 240.2EDI/MB 0.39 0.69 0.97 0.42 0.69 0.74

Medium TDS EDI 60 150 385 28.7 75.6 153.5MB 152 216 354 60.2 94.2 176.7EDI/MB 0.39 0.69 1.09 0.48 0.80 0.87

Low TDS EDI 60 150 385 27.6 71.1 140.3MB 151 216 352 49.6 72.6 126.2EDI/MB 0.40 0.69 1.09 0.56 0.98 1.11

Source: Edmonds and Salem (1998).


expensive than MB system. This is primarily the result of the replacementfrequency of the ED stacks. EDI demonstrates a clear capital cost advantage inthe 50 to 200 gpm range over MB polishing for new plants. MB polishers requireconsiderable ancillary equipment (chemical storage and delivery systems, wasteneutralization system) that negatively impacts their cost effectiveness in low-to-moderate flow range. The 600 gpm case shows that the capital cost of EDIand MB systems are approximately equal.

It was found that EDI, in low-to-moderate flow rate system cases, issignificantly less expensive, on an annualized cost basis, than MB. In addition,EDI offers other cost benefits over MB that was not quantified in this study.These include minimization of wastewater monitoring and discharge activity,significantly reduced space requirements, reduction in spill protection plandesigns and tasks, and less chemical handling for plant-operators. EDI also hasthe ability to respond to feed water TDS changes without compromising de-ionized water production and quality, with minimal impact on operation costs.

REFERENCES

Akahori, M., Konishi, S., 1999, Development of electrodeionization apparatus contain-ing novel ion-exchange materials, J. Ion Exch, 10, 60–69.

Allison, R. P., 1996, The continuous electro-deionization process, American DesaltingAssociation 1996 Biennial Conference & Exposition, Monterey, CA, USA August 4–8.

Deguchi, T., Karibe, T., 1998, Electro deionization, In: Water Treatment Handbook,Maruzen Publishing Co., Tokyo, pp. 220–225.

DiMascio, F., Ganzi, G. C., 1999, Electrodeionization apparatus and method, U. S.Patent 5,858,191.

Edmonds, C., Salem, E., 1998, An economic comparison between EDI and mixed-bed ionexchange, Ultrapure Water, November 1998.


Ganzi, G. C., Egozy, Y., Giuffrida, A. J., Jha, A. D., 1987, High purity water byelectrodeionization performance of the ionpure continuous deionization system,Ultrapure Water, 4(3), 43–50.

Ganzi, G. C., Jha, A. D., DiMascio, F., Wood, J. H., 1997, Electrodeionization, theoryand practice of continuous electrodeionization, Ultrapure Water, 14(6), 64–71.

Glueckauf, E., 1959, Electro-deionization through a packed bed, Brit. Chem. Eng., 4,646–651.

Hernon, B., Zanapalidou, H., Prato, T., Zhang, L., 1999, Removal of weakly ionizedspecies by EDI, Ultrapure Water, 16(10), 45–49.

Hernon, B. P., Zhang, L., Siwak, L. R., Schoepke, E. J., 1995, Process report,Application of electrodeionization in ultrapure water production, 1995, 56th AnnualMeeting International Water Conference, Pittsburg, PA, October 29–November 2.

Kollsman, P., 1957, Method and application for treating ionic fluids by dialysis, USPatent No. 2,815,320.

Matejka, Z., 1971, Continuous production of high-purity water by electrodeionization,J. Appl. Chem. Biotech., 21, 117–120.

Neumeister, H., Flucht, R., Furst, L., Nguyen, V. D., Verbeek, H. M., 2000, Theoryand experiments involving an electrodeionization process for high-purity waterproduction, Ultrapure Water, 17(4), 22–30.

Parise, P. L., Parekh, B. S., Waddington, G., 1990, The use of ionpure continuousdeionization for the production of pharmaceutical and semiconductor grades of water,Ultrapure Water, 7(8), 14–27.

Sata, T., Yamaguchi, T., Matsusaki, K., 1995, Effect of hydrophobicity of ion exchangegroups of anion exchange membranes on permeability between two anions, J. Phys.Chem., 99, 12875–12882.

Tanaka, Y., 2007, Water dissociation reaction generated in a water dissociation layerformed on an ion exchange membrane, J. Membr. Sci., the article in submitting.

Verbeek, H. M., Furst, L., Neumeister, H., 1998, Digital simulation of an electrodeion-ization process, Computers Chem. Eng., 22(Suppl.), S913–S916.

Walters, W. R., Weiser, D. W., Marek, L. J., 1955, Concentration of radioactive aqueouswastes, Ind. Eng. Chem., 47(1), 61–67.

Yagi, Y., Hayashi, F., Uchitomi, Y., 1994, Evaluation of boron behavior in ultrapurewater manufacturing system, Proceedings of the Semiconductor Pure Water andChemical Conference, San Jose, CA, pp. 54–62, March 8–10.

[membrane science and technology] ion exchange membranes - fundamentals and applications volume 12...

Documents