The United States Pharmacopeial Convention Inc. (USP) has proposed new specifications
for Water for Injection (WFI) and Purified Water (PW). These specifications will include
quantitative conductivity, and total organic carbon (TOC) measurements will replace
100-year-old qualitative wet chemistry tests. The impetus behind these changes is to
create a means of analysis that is less labor-intensive, yet more reflective of the water
quality. A comparison of the current and proposed specifications is given in Table A.
For the study summarized in this article, extensive testing was conducted to determine
the optimum design and operation of a two-pass reverse osmosis (RO) machine to produce
water meeting the requirements for United States Pharmacopeia 23 (USP 23) PW.
The study included the development of high-performance, sanitary-design membrane
elements; a comprehensive study of two-pass ROs currently in operation; and an intensive
production-scale evaluation of a two-pass RO performance, examining various operational variables.
CARBON DIOXIDE AND CONDUCTIVITY
The proposed USP 23 PW monograph will call for on-line (or immediate off-line) conductivity at or
below 1.3 microSiemens per centimeter (S/cm) (when the temperature is at or above 25C [77F])
as Stage 1 testing. The second-stage testing calls for off-line analysis showing a conductivity of
2.4 S/cm (at 25C C) This off-line conductivity requirement is higher than the on-line requirement,
allowing for the increase in conductivity caused by the contribution of dissolved carbon dioxide
(CO2) gas present in the water. Thus, a key to producing water meeting the on-line requirement
is the removal of CO2 from the water.
When CO2 gas is dissolved in water, a portion reacts with the water (H2O) molecules to form
carbonic acid. Being a dissolved gas, the CO2 passes completely through an RO membrane,
and once the CO2 reassociates with water molecules to form bicarbonate (HCO3) in the RO
product water, it contributes to the conductivity of the permeate water.
There are three reactions that govern the chemistry of CO2 in water (Equations 1-3):
CO2 + H2O <---> H2CO3
(carbonic acid) Eq. 1
H2CO3 <---> H+ + HCO3-
(bicarbonate ion) Eq. 2
where PKa = 6.38
HCO3- <---> H + CO32-
(carbonate ion) Eq. 3
where PKa = 10.37
When gaseous CO2 is dissolved in water, a portion is hydrated to form carbonic acid (Equation 1).
This carbonic acid dissociates into bicarbonate and hydrogen ions. At a pH of 4.3, very little
of the carbonic acid is dissociated. At a pH of 6.38, the molar concentration of carbonic acid
equals that of the bicarbonate and hydrogen ions. At a pH of 8.3, there is no longer an appreciable
amount of CO2 or H2CO3 present in the water. Above this pH, the bicarbonate ion is converted to
carbonate H+, as shown in Equation 3.
As the pH increases, all three equations are driven to the right and there is less CO2 available
in the gaseous form. Since RO membranes are unable to reject gaseous CO2, the permeate
conductivity is lowest when the feed pH is near or above 8.3. When the pH is above 8.3, the
CO2 is found in the form of the carbonate and bicarbonate ions, which are easily rejected by
As a part of the study to determine the optimum performance of two-pass RO, four different
types of spiral-wound elements manufactured with polyamide (PA) membrane were compared
to identify the optimum element type for the production of USP 23 PW. Several membrane
elements of each type were evaluated in first- and second-pass simulation tests to
determine their flow and rejection performance on softened, dechlorinated feedwater with
an average conductivity of 350 S/cm, a pH of 8 to 8.5, and an alkalinity of 250 ppm as CaCO3.
Pretesting verified that all elements showed rejection of 99% + 1% at a high conductivity
(2,000 S/cm) made up of monovalent salts (sodium chloride [NaCl]). First- and Second-pass
performances were evaluated based on conductivity measurements. Conductivity and pH
were measured off-line using calibrated instruments, and flowrates were measured using
a timed volumetric method.
During first-pass tests, salt passage ranged from 0.27% to 0.43%. Permeate from the
first-pass test was then collected and used as feed for the same membranes, creating a
second-pass simulation. Two tests were conducted; the first at the natural pH of the
permeate, and the second at a pH of 8.0 to 8.5. The latter pH was achieved by
addition of sodium hydroxide (NaOH caustic). This pH was selected to minimize
the effects of CO2 on the conductivity of the second-pass permeate. The percentage
passing averaged 6.5% prior to NaOH addition and 3.5% after NaOH addition. Even
though the addition of NaOH increased the conductivity of the feedwater, the apparent
rejection of all elements increased. This is attributable to the removal of dissolved CO2.
DESCRIPTION OF TEST MACHINE
A 7.5-gallon-per-minute two-pass RO machine was used to conduct testing. The machine was configured with a pump for each pass, with a total of sixteen 4-inch elements. The second-pass permeate piping and all instrumentation were sanitary in design. The machine was configured so that the recovery, membrane crossflow, and other operating parameters were adjustable. Two chemical injection ports were installed at the feed to each pass. Feedwater was typically softened city water with an average conductivity of 450 S/cm, pH of 8.0 and alkalinity of 270 mg/L.
TWO-PASS RO PERFORMANCE
There are several variables to consider when designing an RO machine. Those having the greatest effect on performance include membrane type, recovery, crossflow, and feedwater. Of course, these variables also need to be controlled within limits to allow economical production of Purified Water.
During this experiment. baseline readings were taken before and after each experiment. Baseline recovery was considered to be between 70% and 80% for the first pass and between 50% and 80% for the second pass. Overall, 75 baseline data points were collected.
The average first-pass rejection was 99.5%; second-pass rejection, without chemical addition, averaged 81.4%. The apparent high baseline second-pass passage (as measured by conductivity) is likely reflective of the effects of carbon dioxide. Working with chloride (CI) figures only, first-pass rejection rates were 99.8% (permeate of 270 parts per billion [ppb] from an average of 152 ppm (CI) and second-pass rejection rates were 98.9% (permeate of 29 ppb from an average of 2.55 ppm CI). Therefore, when the CO2 is excluded from the equation, the second-pass conductivity (as based on salt concentration only) would be well below the on-line limit proposed by UPS 23 of 1.3 S/cm.
Comparison of Purified Water Monographs
||USP 23 (Proposed)|
||5 to 7
||5 to 7|
||< 1.3 (on-or off line; T > 25C[77F])|
||< 2.4 (off line at T=24 to 26C [75 to 79F])|
||2.4 to 5.8 (off line at T=24 to 26C [75 to 79F])|
|n/a: not applicable
C02 REMOVAL BY pH ADJUSTMENT
As shown previously, increasing the pH converts CO2 and carbonic acid to bicarbonate (HCO3-) and carbonate (CO3=). Unlike CO2, carbonate and bicarbonate ions are rejected by RO membranes. The most common method of increasing pH is the injection of caustic prior to the first pass or between the two passes. This study evaluated the effects of caustic injection at both points and determined the optimum pH for each.
When no caustic was added, the average second-pass permeate conductivity was 2.4 S/cm. When caustic was injected between the two passes, the average final permeate conductivity was 1.2 S/cm, but it fluctuated from 1.0 to 1.8 S/cm. Caustic injection prior to the first pass produced second-pass water with consistent conductivity of 1.0 S/cm. Note: A water softener is required for this scenario since operating at a high pH on unsoftened water would promote hardness scaling on the membrane surface.
The optimum pH varies for each water source, mostly dependent on alkalinity levels. The optimum feed pH on the test water supply was found to be 8.0 to 9.0, resulting in an interstage pH of 7.5 to 8.5.
When interstage caustic injection was used alone, the optimum interstage pH was 8.0, but it was difficult to control because of the problems associated with measuring pH on low-conductivity water. The advantage of first-stage caustic injection is obtaining permeate with consistently low conductivity by maintaining the benefit of high pH through both passes. While there has been a tendency to design the second pass at high recovery (e.g., 90%+), this study showed that there is merit in designing the second pass at a lower recovery (e.g., 70% to 80%).
EFFECT OF RECOVERY
In an effort to avoid fouling, a first-pass RO machine operating on an average water supply is typically designed at a recovery of 70% to 75%. An important point to note is that the concentrate from the second pass is then typically recycled back to the feed of the first pass because it has a lower conductivity than the feedwater.
Increasing the recovery of the second pass does show a pronounced negative impact on the final product water, while it does not decrease the volume of water sent to drain. Higher recoveries really allow only a slight reduction in the capacity of the first pass.
Second-pass recovery tends to affect permeate quality somewhat more than does first-pass recovery. Overall, it appears that the optimum recovery for the production of high-quality permeate with the least waste was 70% to 80% in the first pass and 70% to 75% in the second pass. This arrangement showed consistently positive results.
TOTAL ORGANIC CARBON
In the new specifications, the "oxidizable substances" test will be replaced by a TOC test with the limit for both PW and WFI proposed to be 500 ppb. Reverse osmosis membranes will typically reject 99.9% of all organics with a molecular weight of 150 daltons or greater. The test machine and field data show that a two-pass RO easily reduces TOC levels from 2 to 3 ppm to below 100 ppb.
RO MICROBIOLOGICAL PERFORMANCE
As in USP XXII the proposed USP 23 monograph does not address limits for microbiological concentrations. Both list a recommended "action limit of 100 colony-forming units per milliliter
(cfu/ mL) for Purified Water in the Information Chapter. The USP 23 is expected to also list a recommended action limit in the information chapter for WFI of 10 cfu/100 mL, while USP XXII does not include an action limit for the microbial limits of WFI, other than the endotoxin specification, which will not be changed.
A test was also conducted to determine the ability of a two-pass RO to remove bacteria. An average concentration of 80,000 cfu/mL was injected into the two-pass test machine for two hours. The average bacteria concentration in the first-pass permeate was 19 cfu/mL, and an average of only 3.0 cfu/ mL was detected in the second-pass permeate. This is more than a 4-log reduction across both passes. However, it is important to underscore that while an RO system may appear to have the filtration capabilities to physically remove bacteria cells, it is not reasonable to expect the device to produce sterile or even WFI-quality water on a consistent basis, since the product water can become recontaminated, unless immediately heated or treated with a chemical sanitizing agent such as ozone.
USP 23 PURIFIED WATER SYSTEM
Figure 1 shows a schematic of a complete "USP 23PW" system. As specified in the USP, the feedwater must meet U.S. Environmental Protection Agency National Primary Drinking Water Regulations (EPA NPDWR). The water should be first processed through multimedia filtration (manganese greensand, anthracite, and gravel) to remove particles. Next, the water should be softened to below 5-mg/L hardness. This is necessary to avoid scaling of the RO membrane at the higher pH levels. Some residual chlorine should remain in the feedwater through these processing steps, if possible. However, the chlorine should then be removed by sulfite injection to protect the PA membranes from oxidation. Activated carbon may be an option, but it does facilitate microbial growth and makes microbial and TOC control much more difficult. Caustic should be injected prior to the RO until the pH reaches 8.0 to 8.5.
Figure 1: USP 23 Purified Water system
The RO machine should have a maximum recovery of 80% for the first pass and 75% for the second pass. The second-pass concentrate should be recycled to the first-pass feed. The system should use USP 23 design elements that have a maximum NaCl passage of 1%. These membranes should be of a sanitary design with stainless steel permeate tubes to ensure effective cleaning and sanitation. The second-pass permeate piping should also be sanitary design.
A final treatment step to be considered for USP 23 PW production might be ozonation. An ozonated storage loop protects the water from microbial recontamination and further reduces TOC levels. Average TOC levels in an ozonated loop are typically well below 10 ppb. An ultraviolet light may be used to revert the O3 back to O2.
A two-pass RO machine, when properly designed, will reliably produce Purified Water meeting the specifications proposed for USP 23. Beginning with a softened feedwater of 450 S/cm, this study was able to demonstrate that Stage 1 on-line compliance was routinely achieved when adding caustic to the feed of a two-pass RO operating at recoveries of 70% to 75%.
This paper was presented at Pharmaceuticals Waters '95, Atlantic City, NJ, May 1995.