Ionic species are not all removed by the EDI process with equal efficiency. This fact impacts the quality and purity of the product water.
Easy ions are removed first.
The ions with the strongest charge, the smallest mass, and the highest adsorption to the resins are removed with the highest efficiency. These typically include: H+, OH-, Na+, Cl-, Ca+2, and SO4-2 (and similar ions).
In the first section of the EDI module, these ions are removed preferentially to other ions. The relative quantity of these ions affect the removal of the other ions. The pH approaches 7.0 in this section since the H+ and 0H- ions become balanced.
The first section of the EDI module is known as the "working bed".
- Moderately ionized and polarizable ions are removed next (e.g., CO2).
Carbon Dioxide Removal
This graph shows the effect of CO2 in relatively pure deionized water. CO2 is the next most common EDI feedwater constituent. CO2 has complex chemistry depending on the local concentration of protons, and is considered moderately ionized:
CO2 + H2O => H2CO3 => H+ + HCO3- => 2H+ + CO3-2
Since the pH is forced to be near 7.0 in this section, most of the CO2 is forced into the bicarbonate (HCO3-) form. Bicarbonate is weakly adsorbed by the anion resin, so cannot compete with "easy" ions such as Cl- and SO4-2.
In the second section ("Polishing Section") of the EDI module, CO2 (in all of its forms) is removed preferentially to weaker ions. The amount of CO2 and HCO3- in the EDI feed strongly effects the final resistivity of the product water and the efficiency of silica and boron removal.
It is found that as long as CO2(in all forms) is less than 5 mg/l, high quality ultrapure water can be achieved. If the CO2 concentration is greater than 10 mg/l, it can interfere with the total removal of ions and strongly impacts the EDI product quality and the silica removal.
- Weakly ionized species are removed last (e.g., dissolved silica and boron).
Since species such as molecular silica are very weakly ionized, and difficult to adsorb on ion-exchange resin, they are the most difficult to remove using any deionization process.
If all of the "easy" ions are removed, and all of the CO2 is removed, the EDI module can focus its force on removing these weakly ionized species. The residence time available in this third section of the module is important. The longer the residence time available in the module, the higher the removal efficiency. A long third-section residence time can be achieved by minimizing the conductivity of the RO product (the quantity of "easy" ions to be removed) and minimizing the quantity of CO2 in the RO product.
The second and third sections of the EDI module are known as the "polishing bed".
Silica is one of the more important minerals to remove from water for power generation and semiconductor applications. It also one of the most difficult to remove.
Silica chemistry is complex. On the most basic level, silica comes in "colloidal" and "reactive" forms. silica level in feedwater depends upon the geology of the region and whether the source is surface water or well water. Silica in raw water will range from less than 2 ppm to over 100 ppm.
Physical processes such as reverse osmosis (RO) will remove colloidal silica. EDI will only remove reactive silica.
Removal of reactive silica depends upon its charge. silica has little, if any, charge at neutral pH near 7 since the pK1 of silicic acid is 9.8 This makes it difficult to exchange with ion exchange resin, or to remove with RO or EDI. Raising the pH to above 9.8 helps with the driving force.
Silica scaling is also an issue. The solubility of silica at pH 6-8 is only 120 ppm at 25oC. this means that 30 ppm of silica in an RO feedstream with 75% recovery will begin to scale. There are two prevention techniques for silica scaling. One is the use of an antiscalant in the RO process, which will delay the precipitation of solid silica. The other is raising the pH, which increases the solubility limit of silica. At pH 10 silica is soluble up to 310 ppm. Of course, high pH will cause hardness scaling if the feed is not adequately softened.
It is important to maintain the silica content of the EDI feed stream to under 0.5 ppm as silica in order to:
- Avoid scaling in the EDI concentrate stream
- Minimize silica levels in the product water
Typical commercial RO modules will reject silica at only twice the passage of chloride ion. Most spiral RO module manufacturers claim 99.7% rejection for individual high quality elements. 99.0% - 99.5% is a reasonable silica rejection for a well designed RO system.
With a 20 ppm silica feed and 75% recovery, a 99.0% rejection element will maintain the RO product at 0.5 ppm silica.
For higher levels of silica in the feed, the RO system should be designed with higher quality RO elements and/or lower recovery. Using 99.7% rejection elements and 65% recovery, the RO feed can approach 90 ppm and still maintain 0.5 ppm silica in the effluent.
Water splitting and Module Resistance
As described above, water splitting is critical for the removal of weakly ionized species like silica, carbon dioxide, and boron. It is also critical for the removal of strong ions in thick cell, separate bed EDI. The amount of water splitting can be quantified using Faraday's law to compare the theoretical amount of current needed to transfer a given amount of ionized species out of one electrochemical cell to the actual applied current through that cell. Faraday's law states that
I = theoretical current, amps
Eq = number of equivalents transferred per cell
t = time, seconds
F = the Faraday constant = 96,500 coulombs/equivalent
If we define the current efficiency as the theoretical current given by Faraday's law, divided by the actual applied current, we obtain the following equation:
h = current efficiency, %
Ia = applied current, amps
To make this equation easier to use, we can substitute flow rate and feed concentration for the equivalents removed per unit time per cell so that:
TDS = total dissolved solids, mg/l as CaCO3
Q/n = product flow rate per cell, l/min/cell
3.22 = conversion factor
In Eq. 11 we show TDS as the total feed load to the cell but the total feed load also includes weakly ionized species, such as carbon dioxide and silica, which are also removed. To take these into account, Ionpure uses a term called Feed Conductivity Equivalent (FCE) and E-Cell uses a term called Total Exchangeable Anions (TEA). To calculate true current efficiency, TEA can be substituted directly for TDS in Eq. 11. However, to calculate current efficiency using FCE the conversion factor changes as shown:
We can make a couple of important conclusions from Eq. 10 and 11. First, thick cell EDI, which operates at higher flow per cell than thin cell EDI, will require higher current to maintain the same current efficiency. Second, current efficiency can be lowered, and hence water splitting increased, by increasing the applied current at a constant flow rate and feed concentration. Current efficiency is especially important with regard to weak ion removal. It is not uncommon to operate below 10% current efficiency for improved removal of silica, boron, etc.
E-Cell™ (General Electric corp.) uses a term, which is a variation of current efficiency. They actually refer to it as the E-factor, which is the inverse of the fractional current efficiency. Therefore, increasing the E-factor is analogous to lowering the current efficiency.
In any case, increasing the current requires either increasing the voltage or lowering the module electrical resistance. Increasing the voltage imparts several drawbacks including greater power consumption and increased safety risk. So reducing the module resistance becomes paramount in EDI module design.
Because all EDI devices use ion exchange resin in the diluting cells, the concentrating stream is really the limiting resistance in a module. For modules that don't have resin in the concentrate, the only way to increase the conductivity is to increase the conductivity of the water. This is done by increasing the recovery or by direct injection of a salt, such as sodium chloride, to the concentrate. In addition, many manufacturers use recirculation of the concentrate stream along with high recovery or salt injection to provide a less variable concentration along the length of the module.
There are several drawbacks to concentrate recirculation. Concentrate recirculation requires the use of a pump and additional ancillary equipment for control such as motor starters and throttling valves. This adds complexity to the system design and increases overall cost. In systems with fluctuating operating conditions, operation or adjustment of the pump can make the process significantly more labor intensive. Also, the power required to operate the pump can be a large portion of the total power consumption of the system. A typical industrial EDI system with concentrate recirculation would consume about 1 to 2 kilowatt-hours per thousand gallons of product water (kWh/kgal), where about 0.5 kWh/kgal is for the recirculation pump alone.
Salt injection also has several drawbacks. Increasing the salt concentration in the EDI concentrate stream:
- Limits the ability to recover that stream
- Increases the concentration gradient between the diluting and concentrating cells facilitating co-ion back diffusion if the membranes are not ideally permselective
- Increases the possibility of salt bridging and electrical shorting, and
- Leads to the formation of chlorine at the anode when fed with recirculated concentrate water.