Membrane and System Design Considerations in Producing High Purity Water
By: Nancy Mulhern
The use of membrane technology to produce water of greater and greater purity is rapidly evolving under the pressure of new product quality specifications, including those of the pharmaceutical and beverage industries. Membrane technology is well suited to achieving multi-component water specifications, given the fundamental nature of the separation process. Meeting a resistivity or conductivity specification as the sole gauge of water quality, however, can be more challenging. Membrane technology is capable of producing water with resistivity greater than 1 megohm when applied in a 2-pass RO system that is properly designed and operated.
A thorough study of polyamide (PA) membrane performance variations and possible interactions with feedstream components provides insight into the parameters which are most important in reducing permeate conductivity. This insight can also be useful in troubleshooting systems in the field, when unexplained performance fluctuations cannot be resolved by examining only the more common performance variables.
A series of controlled experiments has shown that membrane rejection will fluctuate in response to feed TDS and pH values, crossflow rates and element recovery levels. The performance of elements in the second-pass of a reverse osmosis (RO) system can be most dramatically affected. These variations, while not significant in the majority of applications, become crucial to the success of high-purity water processing. In addition, the effect of minor feedwater constituents, such as alkalinity and ammonia, are seen to play a dominant role in achieving high-purity permeate.
Polyamide (PA) thin-film composite membranes have a surface charge that plays a role in their separation ability, and the nature of this charge can be altered by the pH of the feedwater pH. The majority of PA (RO) membranes are negatively charged when operated on the pH levels most commonly encountered in water applications.
When the pH drops below a membrane's isoelectric point (generally between pH 4 and 5), these membranes become positively charged. The isoelectric point is that pH point at which the membrane has no net charge. This substantially decreases their performance when the permeate quality is being measured by conductivity. Acid transport through the membrane accounts for much of this apparent fall-off in performance. The effect is completely reversible when the pH is returned to near-neutral levels. The acid transport is facilitated by the presence of unreacted "end" groups (amines) in the polyamide barrier layer. Depending on the amount of unreacted groups present in a particular membrane, different responses to pH changes may be seen.
High pH levels can also reduce the rejection of PA membranes as measured by conductivity. As with the low pH phenomenon, the threshold value at which this decline occurs is unique to each membrane type. In general, pH values above 8.5 can be problematic (Chart 1). Acid addition to lower the pH will correct this condition. Table 1 contains field data that illustrates this effect. The reason(s) for this membrane performance change at high pH is not as well understood as that for pH values below the isoelectric point but some hypotheses can be made.
Many studies have demonstrated that PA membrane undergoes structural changes (possibly "swelling") in response to pH. This physical change may result in higher ion passage. Another possibility is linked to the fact that the carboxylic acid content and subsequent strength of membrane charge vary with the membrane chemistry. Certain membranes will a greater carboxylic acid content and a strong charge, which in turn may provide better rejection of the HCO3 and CO; present at higher pH values. Field data does suggest an associative interaction between alkalinity content and pH levels with respect to PA separation performance.
PA membrane performance is also a function of the relative conductance of the feedwater. Below a certain level of total dissolved solids (TDS), the membrane rejection will decline with the TDS of the feed solution (Chart 2). Given this proportional relationship, it is possible to have a first-pass permeate that is "too pure" for the optimum overall performance of a two-pass RO system.
Recovery and crossflow rates also play an important role in optimum separation performance. Charts 3 and 4 show the relationships observed in laboratory and field-testing. The optimum rates are dependent upon the feed TDS level, which is why the second RO machine may have different design points than the first RO machine.
Also basic to membrane separation is the effect of feedwater chemistry. Chemistry takes center stage when the desired product is high-purity water and the benchmark is conductivity. Dissolved gases such as carbon dioxide (CO2) can dramatically affect permeate conductivity (Table 2), and these constituents cannot be effectively dealt with in their original state by membrane technology alone. In the case of CO2, however, it is possible to force a conversion to bicarbonate (HCO3) and carbonate (CO3-2) ions by raising the feedwater pH. These ions are well rejected by PA (RO) membranes whereas CO2(and carbonic acid) are not rejected at all.
By adjusting the pH of the feed solution, a portion of the CO2present is shifted to HCO3- and/or CO3 -2 depending on the pH level reached. See Figure 1 for the equations that govern the chemistry of CO2in solution. Up to 98% percent of bicarbonate and carbonate can be removed in the first pass of a two-pass system. This method of control is generally most effective when used prior to the first-pass RO. Caustic injection prior to the second-pass RO can also work. It is much more difficult to control interstage caustic injection as the low TDS of the feed to R02 makes pH measurement at this point unreliable.
If elevating the pH of the feedwater is not practical, CO2and a portion of the feedwater alkalinity can be removed through the use of a degasifier. Acid injection ahead of such a unit would make this technique most efficient as this will convert the majority of the alkalinity present to CO2. The degasifier can be located either ahead of the first RO machine or between the two passes. Any alkalinity passing into the R02 permeate will re-equilibrate, forming H2C03, HCO3 and/or CO3 -2in proportion to the pH.
Another water chemistry variable that can play a large role in successfully achieving high-purity water is the presence of ammonia. The ammonia can be present due to chloramination or organic contamination of the feedwater. The use of chloramine treatment by municipalities is becoming more common, particularly for surface water sources. There may be an ammonia residual present in the water from the initial chloramine generation and/or the subsequent liberation of ammonia during its treatment by activated carbon or ion exchange.
Ammonia is ionized at neutral and acidic pH values; the addition of a strong alkali will produce molecular ammonia. Figure 2 illustrates the ammonia - ammonium relationship relative to pH. When a system feedwater contains ammonia, the need to add caustic for CO2removal must be carefully balanced with the need to eliminate ammonia in the permeate.
Ammonia (NH3) can pass through the membrane system in either the molecular or ionic (NH4+) form. Ammonium hydroxide (NH4OH) would be the most likely ionic form to pass through PA (RO) membrane, particularly if caustic is being used to raise the pH in the system. Ammonium hydroxide is less conductive than ammonium carbonate [(NH4)2CO3] so it is not uncommon to find off-line samples or storage tank water with conductivity higher than that of on-line readings. The pH values, however, will be lower. This shift is due to absorption of CO2from the air. Without the presence of ammonia, this type of contamination of high-purity water with CO2would generate higher conductivity as well as the reduced pH.
CO2+ H2O
Typical 2-Pass RO System Rejection Profile
FIRST PASS / SECOND PASSBased on Coductivity 99.5% / 81.6%
Based on Chloride Rejection99.8% / 98.9%
Typical Results for 2-Pass HR(PA) Membrane Systems
Conventional Design / Optimized DesignConductivity ( S/cm) / 2.6 / 0.8
TOC (ppb) / 300-500 / <100
2-Pass RO System: Effect of Acid Adjustment
Parameter* / Softened Feed / Permeate RO1 / Permeate R02pH / 8.7 / 9.3 / 9.2
Conductivity ( S) / 778 / 20 / 10
Chloride (ppm as ion) / 89 / 2 / <0.01
Sodium (ppm as ion) / 196 / 2.5 / 1.1
Alkalinity (ppm as CaCO3) / 60 / 4 / 4
pH / 7.5 / 6.3 / 5.8
Conductivity ( S) / 662 / 11 / 2
Chloride (ppm as ion) / 27.3 / 2.0 / 0.03
Sodium (ppm as ion) / 156.9 / 1.5 / 0.1
Alkalinity (ppm as CaCO3) / 32 / 4 / <2
*Off-line sampling/analysis
Conductivity vs. CO2Concentration at 250C
Conductivity s/cm / ppm0.05
0.09 / 0
0.01
0.12
0.16
0.19
0.21
0.24
0.26
0.28
0.3
0.32 / 0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0.48
0.61
0.71
0.81
0.89
0.97
1.04
1.11
1.17 / 0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.69
2.09
2.42
2.72 / 2.0
3.0
4.0
5.0