Paper
TP1087EN.doc Mar-10
transfer membranes, alternating flow streams lent of salt. This law is used as the basis for
dilute streams and concentrate streamsare created. calculating the amount of electric current needed
in an ED system to transfer a specific quantity of
Cell pairs consist of a cation transfer membrane, a
salts.3 For ED calculations, Faradays Law is:
dilute flow spacer, an anion transfer membrane
and a concentrate flow spacer. Up to 600 cell pairs (Nfeed-Nproduct) = I * #cp * ce / (F * Q)
are stacked on one another to make up a mem-
Nfeed = feed normality
brane stack as shown in Figure 1.
nproduct =product normality
I = current
#cp = number of cell pairs
F = Faradays constant
ce = current efficiency for ion transport
Q = dilute flowrate
Current efficiency for ion transport is a function of
the ion-exchange membranes and the stream
concentrations and is in the range of 70-90%.
Faradays law says that, for a membrane stack
with a fixed number of cell pairs and flowrate, the
greater the DC current applied to the stack, the
greater the salt removal. It would appear that the
Figure 1: Electrodialysis Stack best way to operate an EDR stack would be to
apply as high a current as possible to make a very
System Configurations high cut and minimize capital equipment cost.
Each membrane stack can desalt a nominal There is, however, a mass transfer limitation on the
amount of water depending on the number of cell amount of current that can be applied to the
pairs. A typical 600-cell pair membrane stack can stack. At a certain limiting current density, polariza-
desalt about 200,000 gpd (757 m3/day). The tion occurs.
amount of salt removal, or cut, is a function of sev-
eral factors that are discussed in this paper. A Polarization
nominal cut might be 50%. For applications requir- Consider the transfer of chloride ions from the
ing a greater amount of desalting, stacks are dilute stream through an anion-exchange mem-
placed in series, or stages. Each additional stage brane. Figure 2 illustrates the mechanisms for
removes 50% of the remaining salt. For larger mass transfer from the dilute stream through the
volume applications, stacks are placed in membrane. If the transfer number of chlorides in
parallel lines. the membrane is tm and the transfer number of
chloride ions in solution is ts, then the rate of trans-
Salt Removal Rate fer from the membrane surface through an ion-
The amount of salt that can be removed in a stack exchange membrane due to a direct current of
depends on the feedwater quality, the water tem- density CD is given by tm CD/F, and the rate of
perature, the properties of the ion-exchange transfer to the membrane surface through the
membrane, and the spacer properties. This paper solution due to a direct current is given by ts CD/F.
focuses on the spacer properties, but discusses
other factors where relevant.
Faradays Law
Faradays Law, as related to the ED process, states
that the passage of 96,500 amperes of electric cur-
rent for one second will transfer one gram equiva-
Page 2 TP1087EN
This rise in pH in the concentrate stream increases
the likelihood of precipitation of scaling salts in the
anion transfer membrane. This will damage the
membranes over time, so it is desirable to operate
at a current density below polarizing conditions.
The current density at which polarization occurs is
defined as the limiting current density and is
mathematically described by the condition that Nm
= 0. Hence, (CD/N)lim = D F/((tm- ts) ).
The factor, (CD/N)lim, is related to the mass transfer
coefficient. The limiting current density is always
normalized by the dilute stream normality because
Figure 2: Mass Transfer Through an Ion-Exchange the more concentrated the dilute stream, the
Membrane greater the current required to reach polarization.
tm is greater than ts, so the concentration of ions at To minimize the membrane area required for an
the membrane drops, forming a concentration application, (CD / N)lim must be maximized by
boundary layer between the membrane surface changing the boundary layer thickness, the other
and the bulk dilute stream. The concentration gra- parameters being physical properties of the system.
dient causes ions to diffuse from the bulk dilute
stream towards the membrane surface. The ion Flow Spacers
flux is determined by the diffusion coefficient for The boundary layer is a function of velocity and
ion transport through the solution, the concentra- flow spacer geometry. The boundary layer thick-
tion difference between the bulk stream and ness may be minimized by promoting turbulent
the membrane surface, and the boundary flow, and thus increasing the limiting current den-
layer thickness. sity.5,6,7 The amount of turbulence promotion is a
The mass balance at the membrane surface is function of the velocity in the flow spacerthe
given by:4 higher the velocity, the greater the turbulence. The
operating velocity in an ED stack is limited by the
(tm ts) CD/F = D(Nb Nm) /
pressure drop along the flow spacer and through
where the stack. There is a maximum inlet pressure at
which to operate a membrane stack to prevent
D = diffusion coefficient
external leakage. If the pressure drop is too great,
Nb = normality in bulk solution then the number of stages that can be utilized in
Nm = normality in solution at membrane series is limited. Hence, it is desirable to find a tur-
surface bulence promoter that optimizes the performance
by maximizing limiting current density and mini-
= boundary layer thickness mizing pressure drop.
As the current density is increased, the concentra- The type of flow spacer that has had the most
tion at the membrane surface, Nm, decreases until commercial success in terms of the total installed
there are no ions at the membrane surface to capacity of ED/EDR plants is the tortuous path
transfer. At this point, water splitting, or polariza- spacer.8 This spacer is manufactured by using two
tion occurs. Water dissociates into hydrogen (H+) sheets of low-density polyethylene with die-cut
and hydroxyl (OH-) ions. Experimentally, polariza- flow channels. The two sheets of polyethylene are
tion is found to occur at the anion transfer mem- glued together to form an under/over flow path
brane first. Since the anion transfer membrane that creates turbulence. Since the under/over
transfers negatively charged ions, the hydroxyl straps that create the turbulence are fairly infre-
ions travel through the membranes. The pH of the quent, the amount of turbulence promotion is lim-
dilute stream decreases, and the pH of the concen- ited. At higher velocities, however, the turbulence is
trate stream increases. sufficient to obtain reasonable limiting current
densities. The pressure drop per unit length for this
TP1087EN Page 3
design is low, so a long flowpath can be used with- retentate spacer in an RO element needs to pro-
out having excessive pressure drops. By mote turbulence to minimize the thickness of the
utilizing a long tortuous flowpath and operating at concentration polarization boundary layer at the
velocities in the 20 - 40 cm/s range, this design membrane surface. These retentate spacers are
optimizes turbulence promotion and stack pres- manufactured from non-woven polypropylene net-
sure drop. The disadvantage of this approach is ting. The RO industrys requirements for high vol-
that the tortuous design has significant membrane ume production, as well as high tolerances to
area wasted as sealing area. In a typical commer- maximize the amount of membrane area in an RO
cial design, 36% of the membrane area is shad- element, have led to improved manufacture of
owed by sealed area, so only 64% of the non-woven netting. It has become useful to take a
membrane area is usable for desalting. Figure 3 new look at screen spacers for EDR systems.
illustrates a tortuous path flow spacer, designated
Mark III by the manufacturer. Comparison of Spacer Properties
Bench scale spacers were manufactured, and their
performance, in terms of limiting current density
and pressure drop, was measured in the laboratory
for a sodium chloride solution at 70F. The limiting
current density was defined as the current density
at which the current efficiency for hydroxyl ion
transfer was 2 x 10-5 in an all-anion stack.
Experience has shown that this amount of increase
Figure 3: Mark III Flow Spacer in pH in the concentrate stream would be needed
to significantly increase the risk of precipitation.
Several manufacturers of electrodialysis equip-
ment have employed screen spacers to promote Figure 4 shows a graph of limiting current density
turbulence. In these spacers, woven or non-woven versus velocity for conventional tortuous path
netting is used in the flow path to create turbu- spacers and a screen spacer manufactured with
lence. In the 1950s the authors company looked non-woven netting. This netting is similar, but not
into the development of screen spacers and found identical, to the netting used in many commercial
that promoting turbulence with a screen spacer RO elements. For a given velocity, the limiting cur-
was better than with a tortuous path spacer. The rent density for the screen spacer is about three
use of screen spacers was not practical at the times that of the tortuous path spacer.
time, because netting was relatively expensive, and The limiting current density defines the maximum
methods for producing screen spacers in large amount of salt that can be removed per unit
volumes were limited. For practical long-term length. For example, for a 0.01 N solution of sodium
operation of large electrodialysis systems, the chloride at 70F operating at 70% of the limiting
thickness tolerances of flow spacers are critical. current density, the salt removal per unit length is
Since a thousand or more spacers are piled on top shown in Figure 5. The screen spacer removes
of one another in an electrodialysis stack, even more salt per unit length than the tortuous path
slight variations in component thickness can lead spacer, so a shorter flowpath can be used to ob-
to large variations in the stack heights. The thick- tain a desired level of salt removal. For both spac-
ness of the sealing area around the flowpath must ers, the lower the velocity, the greater the salt
be very close to the thickness of the flow spacer. If removal. The lower the velocity, however, the lower
the sealing area is thicker than the turbulence the flowrate per unit width, and so a wider flow-
promoters, the turbulence promoters do not touch path is needed to desalt a given volume of water.
the membrane surface. This can cause laminar
flow at the membrane surface, which in turn The pressure drop per unit length is much greater
reduces limiting current density significantly. for the screen spacer than the tortuous path
spacer, as shown in Figure 6. Since the frequency
In the 1990s, with the development of the spiral- of turbulence promoters is much higher for the
wound RO element, the technology for manufac- screen spacer than for the tortuous path spacer,
turing non-woven netting has improved dramati- these results are not surprising. To minimize the
cally. As with flow spacers in an ED system, the pressure drop, the optimum velocity for the screen
Page 4 TP1087EN
spacer is in the 6 -12 cm/s range, and a short flow-
path length is selected.
Power Consumption
Figure 6: Pressure Drop per Unit Length versus
Velocity In electrodialysis plant design, there is a trade off
between power consumption and capital cost. For
High Performance Screen Spacer any particular spacer, the greater the number of
cell pairs, the lower the DC power consumption
The outside dimensions of 18x 40 (46 x 102 cm) and vice versa (Figure 8). In developing the new
for the high performance screen spacer were spacer, it was important to make sure that capital
based on the dimensions of pre-existing ion- cost savings from using fewer stacks and cell pairs
exchange membrane production lines and the were not accompanied by increases in power con-
requirement for membranes and spacers to be sumption. The use of the thinner spacer means
easy to handle by one person for stack mainte- that the electrical resistance of each cell pair is
nance. A U-shaped flowpath was developed to fit reduced, so that the overall DC power
the optimum flowpath length and width into the consumption for a given number of cell pairs
18 x 40 configuration. This spacer design, known decreases. Therefore, at constant DC power, fewer
as the Mark IV, has about 74% usable area in cell pairs can be used with the screen spacer, pro-
contrast to 64% for the Mark III spacers (see Figure 7). viding savings in capital cost without increases in
operating costs.
TP1087EN Page 5
the average performance of the product from two
stages of Mark III stacks. This test demonstrated
the ability of the Mark IV stack to consistently
achieve the same salt removal as two stages of
conventional stacks. Figure 10 shows that the DC
power consumption was equal to or lower than the
Mark IV stack, even though there was half the
number of cell pairs.
Page 6 TP1087EN
compensate. A comparison of four stages of tortu-
ous path to three stages of screen spacer would
have placed both spacers at the point
of equal DC power consumption on their
operating curves.
Suffolk,Virginia
The pressure drop of both stacks increased as the
temperature decreased, and the Mark IV prototype The Robert G. House Water Treatment Plant in Suf-
stack had almost half of the pressure drop of the folk, Virginia, installed an EDR plant in 1990. This
Mark III stack. Assuming these stacks would be system treats 3.8 mgd and consists of three EDR
used on a new EDR system that uses VFDs to con- units, each with 8 lines and 3 stages of conven-
serve energy, the reduced pressure drop represents tional Mark III stacks.10
a 27% saving in pumping power, as shown in Table 1. The first stage of one line was recently replaced
Figure 12 shows the DC power consumption versus with a Mark IV stack. The Mark IV stack is the same
percent salt removal for both stacks during the pe- size as the Mark III stack, although it includes 600
riod from 500 1500 operating hours when the cell pairs rather than 500 cell pairs in the Mark III
average temperature was 6C and the feed con- stacks. The Mark IV stack is operating at the same
ductivity was constant at 2,100 S/cm. conditions as the other Mark III stacks in parallel
Although the pressure drop was lower, and a 46% lines. Each line has the same flowrate of approxi-
savings in cell pairs was achieved, the DC power mately 110 gpm (0.4 m3/h), and the first stage
for the prototype Mark IV stack was greater than applied voltage was 300 V in each case.
the DC power for the Mark III stack. The reason for Table 2 compares the salt removal for the Mark IV
this is that the systems are not operating at points stack and a Mark III stack in parallel. This data
of equal DC power consumption on their operating shows the increase in salt removal that is made
curves, as illustrated in Figure 8. The reduction possible by the Mark IV stack. The Mark IV stack
from four stages of tortuous path spacers to two achieves a cut of 50% while the Mark III stack
stages of screen spacers was such a large achieves a cut of 30% under the same
decrease that the power consumption increased to operating conditions.
TP1087EN Page 7
Table 2: Suffolk Stack Performance
Summary 7. Zhong, K.W.; Zhang, W.R.; Hu, Z.Y. and Li, H.C.,
Effect of Characterizations of Spacer in Elec-
The new high performance spacers make it possi- trodialysis Cells on Mass Transfer, Desalina-
ble to reduce the overall life cycle costs of an EDR tion, 46, 1983, pp. 243-252.
plant. At Lake Granbury, the comparison demon-
8. Wangnick, K., 1996 IDA Worldwide Desalting
strated minimizing cell pairs to reduce capital cost
Plants Inventory Report, No. 14, pp. 176-185.
at constant operating costs. At Foss Reservoir, the
comparison demonstrated reduced capital costs 9. von Gottberg, A.J.M. and Siwak, L.R.,
with only a small increase in power cost. At Suffolk, Re-Engineering of the Electrodialysis Reversal
the comparison demonstrated better product qual- Process, Intl Desalination & Water Reuse
ity from a single membrane stack. Quarterly, Vol. 7/4, Feb./Mar. 1998.
10. Thompson, M.A. and Robinson, Jr., M.P., Suffolk
References Introduces EDR to Virginia, Proceedings,
American Water Works Assoc. Membrane
1. Mason, E.A. and Kirkham, T.A., Design of Elec-
Conf., Orlando, FL, 1991.
trodialysis Equipment, Chem. Eng. Prog. Sym-
posium Ser. No. 24, Vol 55, 1959.
2. Siwak, L.R., Heres How Electrodialysis
Reverses... and Why EDR Works, Intl Desalina-
tion & Water Reuse Quarterly, Vol.
2/4, 1993.
3. Meller, F.H., Ed., Electrodialysis (ED) & Elec-
trodialysis Reversal (EDR) Technology, Ionics,
Incorporated, 1984.
4. Rosenberg, N.W. and Tirrell, C.E., Limiting Cur-
rents in Membrane Cells, Ind. Eng. Chem., 49,
1957, p. 780.
5. Belfort, G. and Guter, G.A., An Experimental
Study of Electrodialysis Hydrodynamics,
Desalination, 10, 1972, pp. 221-262.
6. Chiapello, J.M. and Bernard, M., Improved
Spacer Design and Cost Reduction in an Elec-
trodialysis System, Journal of Membrane Sci-
ence, 80, 1993, pp. 251-256.
Page 8 TP1087EN