LinkedIn +1.949.240.2188 |


Auto-Translation by Google
Auto-Translation by Google:

Mechanism of Silica Removal in Electrodeionization (2006)

Mechanism of Silica Removal in Electrodeionization (3-Zone EDI Model)

Michael J Snow, Ph.D.

May 2006

Keywords: Electrodeionization, EDl, CEDI, Water, Pure, Ultrapure, Silica, SiO2, Silicic acid, H4SiO4, Reactive; Non-reactive; Colloidal; Reverse Osmosis; lon-Exchange; Resin; Anion; Cation; Fouling; Rejection; Process design; Membrane; Zones; Electropure; SnowPure


Electrodeionization (EDI) is a continuous separation process that uses a DC voltage force to remove charged species from a liquid stream – typically water.  Removing silica from industrial waters is very important for power generation (boiler-water makeup) and electronics (semiconductors).  However, since EDI is based on ionic charge and silica (SiO2) has no fixed charge under normal conditions in pure water, it is not obvious how EDI removes this important species.  There are several zones within an operating EDI module.  The Working Zone (Zone 1) primarily transfers charged ions, The Intermediate Zone (Zone 2) primarily transfers bicarbonate ions, and the Polishing Zone (Zone 3) is where silica is ionized.  This paper intends to shed more light on the mechanism of silica removal in EDI.  It then proposes some strategies for improving the efficacy of silica removal which is important to industrial applications where very-low silica levels are desired.  It also offers suggestions on how to avoid silica problems in geographic regions, like Mexico and Japan, where high-silica issues are prevalent.


Silica is present in natural waters in the range of 5 to 110 mg/l (ppm).  In waters with low levels of silica, normal water treatment (softening, filtration, reverse osmosis, and EDI) works well.  At high levels, above 30 mg/ml, extreme and optimum measures must be taken to both avoid scaling problems and to maximize the silica removal in the product water.

High pressure boilers typically require ultrapure water below 10 µg/l (10 ppb).  Even lower levels 0.3µg/l (300ppt) are necessary for semiconductor applications, (Balasz, 2007).  This means that the water purification system must remove over 99.99% of the silica in the feedwater.


The chemistry of silica in water is complex (Bates 2001, and Meyers 1999).  Silica in water to be treated for industrial use is composed of “reactive (monomeric) silica and “unreactive” (colloidal, polymerized) silica.  At neutral pH, monomer silica is in the form of silicon dioxide, or the hydrated version which, in simplistic terms, is silicic acid, H4SiO4.

When pH rises, the hydrated silica ionizes, H4SiO4 <—>H+ + (H3SiO4) to give an ionic silica species.  However, the pKa of H4SiO4 is about 9.7 so, under normal pure water conditions of pH 7, there is very little anionic silica.

Nearly 100% of colloidal silica (unreactive polymerized silica, (SiO2)n) and colloidal mixed silicates can be removed physically by filtration and reverse osmosis (RO) via a physical barrier size-exclusion mechanism.  Monomeric uncharged reactive silica may be passed by the RO filter.  This silica must then be removed by downstream processes such as ion-exchange (anion resin) or by EDI.  Since silica is uncharged at normal pH, this is a difficult task for EDI and the mechanism is not widely understood.


RO rejects silica by two mechanisms.  The physical size of colloidal silica means that essentially 100% of colloidal silica will be rejected.  Monomeric dissolved silica may not be removed by polyamide thin film membranes which remove individual ions based on anionic charge and molecular size.  Practically, high-quality commercial RO membranes claim to remove over 99.5% of feed silica (the feed silica’s form is seldom defined).


EDI is a technology whereby charged species can be removed continuously from a water stream.  Typically, water with impurities is introduced into the EDI purifying chambers – where they encounter ion-exchange resins trapped between two ion-exchange membranes – under a DC voltage driving force.  The ions are taken from the aqueous phase onto the solid phase of the resins, and the purer water passes further into the module.  The ions now diffuse under the DC potential through the ion exchange membranes and into the concentrate chamber where they are carried away.

In the simplest model, the ions in the feedwater are pulled equally from the aqueous phase onto the solid resin phase where they then diffuse within the DC voltage gradient, enter the pass through the ion-exchange membranes and are concentrated and carried away by a small water stream.  Anions diffuse toward the anode and cations diffuse toward the cathode (see Figure).  The purified water, with ions removed, becomes the product stream.

3-Zone Electropure EDI Model: In this model, the EDI internals are figuratively broken into a “working bed” first zone, wherein anions and cations are removed by the simple mechanism described above, a “CO2 removal bed” second zone, and a “polishing bed” third zone.  The physical chemistry in each zone is different.


The first zone is where most of the feed conductivity is removed.  From a normal RO, most of the conductivity can be considered Na+, Cl, or their equivalents.  Therefore the anion resin surfaces in this working ion-transfer bed are modeled to be primarily in the chloride form.  There are very few open (+) sites with only a few in the OH form.  The chloride anions diffuse along the surfaces of the beads from (+) site to (+) site, through the membranes and into the adjacent concentrate streams.  As a chloride frees a (+) site on the surface, another anion chloride from the aqueous phase is pulled onto the anion surface phase where it joins the other chloride ions diffusing toward the anode.  Similarly, the cations, modeled primarily in the sodium form, are adsorbed onto the cation resin.  These adsorb onto the (-) sites as they become free to jump from (-) site to (-) as they diffuse toward the cathode.

In the first zone, the removal of conductivity contributes to the DC current (amps) at a high efficiency.

The water also contains carbon dioxide, CO2, and bicarbonate ion, HCO3.  Generally, these are not well adsorbed on the resins in the working bed for two reasons.  First, the bicarbonate anions have only ¼ of the selectivity of chloride anions.  Second, the bicarbonate is in equilibrium with carbon dioxide, (or H2CO3) which is uncharged.  Therefore, in this model CO2 and HCO3 pass through the working zone.

Molecular silica in neutral pH feedwater is uncharged.  It also passes through the working bed since it has nearly zero selectivity on Cl-form anion resins.


The second zone primarily removes the CO2 species in this model.

The resin surfaces become depleted of Na+ and Cl since the bulk of the primary ions are removed coming out of Zone 1.  The zone downstream of the working bed is now capable of removing the carbon dioxide.  The anion resins are depleted of chloride, leaving behind more hydroxide, and the cation resins primarily take the hydrogen form.  Now the bicarbonate ions can become well adsorbed on the hydroxyl-form anion resin bead surface being up to 24x more selective for this surface (Meyers, 1999).  This removal from the aqueous stream shifts the carbon dioxide equilibrium toward the bicarbonate ion which is adsorbed and diffused away under the DC potential.

The model also shows silica to be neutral which allows it passage through Zone 2.  This is because silica does not adsorb well to anion resins with high levels of bicarbonate adsorption.  As such, they tend to pass through the bicarbonate removal zone into the final, polishing zone,


The DC voltage potential causes ions to diffuse in the Zone 1 (working bed) and Zone 2 (bicarbonate removal bed).  The movement of ions in the first zones contributes to the DC current.  Once the purified water enters the polishing zone, the DC voltage potential engenders a different thermodynamic mechanism.  This is where “water splitting” occurs.  Water splitting is the quintessence of efficient silica removal and the prevention of silica fouling.


Water splitting is the process where H2O is lysed, or split, into its ionic constituents.  It is a very important process to EDI.

There is the normal equilibrium equation, H2O = H+ + OH.  Under static conditions, there is a continuous splitting and reforming of the water.  However, under a voltage gradient, the splitting rate dominates the reforming rate at the junction between anion and cation adsorbing surfaces.  The byproducts, H+ and OH, are pulled away from the junction.  The DC voltage gradient is strong in an EDI device at up to 5-8 V/cell (if a cell is 5mm wide then the gradient can be up to 1-2V/mm).

This physical-chemical process provides a never-ending supply of hydrogen and hydroxide ions in the polishing zone.  The splitting of water and the separation of the hydrogen and hydroxide ions is the “electrical work” done in this zone which adds to the current drawn by the EDI module.

The high, local concentrations of hydrogen and hydroxide provide at least two benefits.  First, the process creates regions with very high and low pH (approximately 13 and 1 respectively) which creates a hostile environment for bacteria.  This causes EDI modules to be, at a minimum, bacteriostatic.  Second, and important for this discussion, the supply of hydrogen and hydroxide ions keep the surfaces of the resin beads fresh and fully in their best forms for absorption.  Anion and cation resins are highly “regenerated” in their hydroxide and hydrogen forms.  This is the key to obtaining high efficiency of silica removal and avoiding silica scaling within the module.


Since silica is uncharged at neutral pH (ultrapure water is pH 7), it is weakly attracted to the anion resin.  Furthermore, the anion resin’s surface is shielded from the uncharged silica by fixed (+) charges by chloride, bicarbonate or hydroxide ions.

One way to write the reactive silica equation is:

H4SiO4 = H+ + (H3SiO4).

However, at the surface of the hydroxide-form anion resin, the better description is:

H4SiO4 (aq) + OH (surface) = (H3SiO4)(surface) + H2O (aq)

This describes how the hydroxyl ion on the surface of the polishing resin pulls a proton, H+, from the silicic acid.  A charged anionic silicic ion can now adsorb onto the free (+) site on the resin surface.

The adsorbed silica anion, (H3SiO4), can now migrate across the surface of the anion resin from (+) site to (+) site.  It diffuses toward the anode (+ DC voltage) and away from the cathode (- DC Voltage) until it reaches the positively charged membrane.  It is then pulled into and carried away by the concentrate stream.

The key to the mechanism for efficient removal of silica in EDI is (1) water splitting, (2) the abundant supply of hydroxide ions in the polishing zone and the subsequent formation of a very high pH anion resin surface fully populated by aforementioned ions.

There are other considerations.  The anionic silica, (H3SiO4), must be sufficiently adsorbed to the polishing resins in order to pull it out of the aqueous phase giving it residence time within the EDI module.  And, it must not be too strongly adsorbed or else it will not diffuse rapidly toward the anode.


Silica is energetically favored to form polysilica (SiO2)n at neutral pH.  This reaction, favored in high concentrations, is the source of colloidal silica and it is very difficult to remove from surfaces.  In an EDI module, if the concentration of SiO2 is relatively high in the feedwater, there is a risk of forming polysilica on the anion resin surfaces.

The level of water splitting is high when the DC voltage potential is set correctly.  This maintains the high surface pH of the anion resins in the polishing zone.  The silica is kept in the anion form where it is less likely to precipitate.  The correct DC voltage is able to diffuse the anionic silica steadily across the resins toward the concentrate streams.  If the voltage is kept low then the silica will not be removed as rapidly, it will rise in concentration at the anionic surfaces, and it has a tendency to form polysilica.  This is the EDI “silica fouling” that is discussed in Japan, Mexico, and other areas prone to high-silica feedwater.

The solution to preventing silica fouling is to raise the voltage high enough to ensure that water splitting is abundant; thereby ensuring that anionic silica can be removed rapidly.


The Electropure™ 3-Zone Model for EDI says that simple ions (most of the conductivity in the RO permeate-EDI feed) are removed in Zone 1, that bicarbonate is removed in Zone 2, and that silica (and boron) can only be removed thereafter.  Silica which is neutral in its normal aqueous phase becomes anionic silica when an OH-form resin in the polishing zone extracts a proton.  This is enhanced by the physical-chemical phenomenon of water splitting.  The anionic silica can now be as efficiently removed as other impurity ions.  Water splitting may be seen by some as “current inefficiency,” but it is a vitally important part of EDI performance.

Silica removal is improved by minimizing the size of the first zone, which is accomplished by minimizing the RO permeate conductivity. The size of the second zone is minimized by lowering the CO2 + HCO3in the EDI feed, and by ensuring the proper DC voltage is applied.  These strategies maximize the size and effectiveness of Zone 3, the polishing zone.

Silica fouling is avoided by lowering the level of silica in the feedwater and by ensuring the proper DC potential is used in the EDI system.

It is worth noting, especially so for semiconductor production, that Boron chemistry parallels Silicon chemistry, meaning the same strategies apply.


  1. Bates, Wayne (2001) “RO Water Chemistry”, Hydranautics Membrane Corp. technical paper posted on website,
  2. Meyers, Peter (1999) “Behavior of Silica in ion Exchange and Other Systems” ResinTech, Inc., IWC-99-64
  3. Balazs Analytical (2007), “Ultrapure Water Monitoring Guidelines” Revision 2.0
  4. Ning, Robert Y. (2002), “Discussion of silica speciation, fouling, control, and maximum reduction”, Elsevier Science B.V.