Figure 1. A Typical High Purity Water Plant
Figure 1 is a diagram of a typical high purity water plant.
Figure 1. A Typical High Purity Water Plant
The system can be subdivided into the following basic process stages:
The RO pretreatment stage is an area whose function is frequently not very well understood, especially in regard to the scope of the filtration being carried out at this point. Aside form the various types of chemical treatment found at this stage (which will not be discussed in this paper), the following types of filtration can frequently be found:
Virtually every high purity water plant is custom designed for different water purity requirements, water demand, and quality of the raw influent water. The pretreatment stages can be found in a variety of configurations. However, the pre-RO cartridge filters are almost invariably found immediately upstream of the RO membranes. An understanding of the function of the RO membranes and the necessary quality of RO influent will give a greater understanding of the need for and the placement of the pre-RO cartridge filters.
The principle of the RO membranes is that of the selective permeation of a fluid across a nonporous membrane. Since this permeation must overcome the osmotic forces of the solution, the energy required for the process is quite high. The RO membrane is a nonporous water0permeable membrane that divides the influent water stream into two streams: the permeate (product) and the retentate (reject) streams. The permeate stream is freed of particulates and many dissolved impurities. The retentate or reject stream is enriched with the suspended and dissolved solids. The form of energy used to drive this process is high pressure, usually between 200 and 600 psig. The geometry of the RO membranes can be either spiral would or hollow fiber. The materials of construction of the membrane can be either CA (cellulose acetate), polysulfone, or TFC (thin-film composite).
The mechanism used by an RO membrane to produce a purified effluent is different from the mechanism used by a porous cartridge filter. The principle of RO is that of separating the impurities into a defect stream, and no impurities are retained by the membrane itself. If the RO membrane behaves like a cartridge filter, physically retaining impurities, then the RO membrane is not functioning correctly.
RO membranes attract foulants by two means: chemical fouling and physical interception. Chemical fouling, the deposition of dissolved constituents on the membrane, is minimized by the proper chemical pretreatment of the influent water. Physical interception is caused by particulates impinging on the membrane and becoming embedded. This process is exacerbated by the very high pressures necessary for reverse osmosis. These high pressures will embed a particle deeply into the membrane, and deeply embedded particles are difficult to remove by cleaning. If any appreciable load of particulates is present in the influent water, the membranes will quickly become fouled.
The function of the pre-RO filters is to minimize the particulate load of the RO influent. The more particles that are removed by the pre-RO filters, the more trouble-free will be the performance of the RO membranes. Most RO membrane manufacturers specify a maximum silt density index (SDI) of 5 or less for the RO influent.
SDI
SDI is a measure of the particulate load on raw influent water. It is a measure of membrane pluggage at constant pressure. Figure 2 is a diagram of the fouling index test kit used to measure SDI.
Figure 2. Fouling Index Test Kit.
It consists of a 1/4-inch tube to tape into the water main, a shutoff valve, a pressure gauge and regulator, a 47-mm filter disc holder and an outlet flow nozzle. The pressure regulator should be adjusted to yield a constant 30 psig back pressure. A 0.45-micron CA analysis membrane is used. The SDI is measured from the decay of the time required to fill a 500-mL container according to ASTM D4189-82 as follows (1):
where:
%P30is the plugggage at 30 psig feed pressure,
T is the total elapsed time (usually 15 minutes),
ti is the initial time required to collect 500 mL of sample, and
tf is the time required to collect 500 mL of sample after test time T.
When the %P at T = 15 minutes is greater than 75%, then the SDI is calculated from the time required to collect a 500 mL sample at T = 5 minutes.
Frequently the SDI test membrane is observed to plug (where pluggage is defined as 20 seconds between drips) before an adequate running time has elapsed. Then the modified SDI is calculated as follows:
For example, when the test membrane plugs in 5 minutes, the Modified SDI is 20.
Absolute Versus Nominal Filter Rating
Filters are assigned ratings according to the size of the particles that they can remove. One must be careful to ascertain whether a filter has an absolute or a nominal rating.
A nominal filter rating has been defined by the National Fluid Power Association (NFPA) as "an arbitrary micron value indicated by a filter manufacturer. Due to lack of reproducibility this rating is deprecated" (2). A nominal filter rating is based upon the removal of some percentage of all particles of a given size or larger. It is rarely well defined and not reproducible. A nominal rating is based on weight removal efficiency rather than particle size removal. This means that the rating is based upon the percentage of the total weight removed rather than the number particles.
There are a number of problems associated with nominal ratings. First, test conditions greatly affect the results of the tests, and therefore results can differ from test to test. Second, a given nominal filter rating might mean that 98% of the weight of the contaminant is removed and 2% of the contaminant is passing through. This 2%could be composed of a few large particles, or it could be composed of many tiny particles. Finally, some manufacturers do not base their nominal rating on 98% removal by weight. Instead they may arbitrarily select 95%, 90% or even a lower percentage as their efficiency.
Absolute filters, on the other hand, are specified with an absolute rating. The Oklahoma State University F-2 filter performance test (3), which has been adopted by the American National Standards Institute (ANSI) as their approved procedure (4), has been modified for single-pass efficiency testing in water to determine the micron size above which the particles are quantitatively removed. Automatic particle counting is performed on-line both upstream and downstream of the test filter. Well-dispersed AC Fine Test Dust (AiChE), a siliceous contaminant, is used to challenge the filters for particle counting. The particle counting results are then used to determine the beta ratio and efficiency, which will be described below.
Beta Ratio and Efficiency
Beta ratios are used to evaluate filter performance for filters having a pore-size rating greater than 1 micron. Beta ratio, Bx, is defined as
where x is the particle size.
The beta ratio can then be used to calculate the removal efficiency defined as
For example, a filter having a beta ratio of w has an efficiency of 50%. A filter having a beta ratio of 10 has an efficiency of 90%.
Beta ratios and efficiencies of filters rated greater than or equal to 1 micron are determined using the modified Oklahoma State University "OSU F-2 Filter Performance Test." In this test, a system is equipped with two particle counters having a range of 1 to 100 microns. One counter is installed upstream of the filter and records the influent particle levels, and the second counter records the downstream effluent particle levels. These particle counters can be preset to count particles of six different diameters. For data on efficiencies with particles of less than 25 microns in diameter, the filter is challenged with a suspension of AC Fine Test Dust.
The relation ship between absolute filtration, nominal filtration, beta ratio, and efficiency can be observed in Table A, which shows the grades of Pall Profile filters and their removal ratings. Profile filters are assigned absolute ratings that correspond to the smallest size particle that exhibits a beta ratio of less than 5,000 during the F-2 test. The results of the F-2 test in Table A show that a large percentage of particles smaller than the Profiles filter's absolute rating are removed.
| Grade | 90% | 99% | 99.9% | 100% |
|---|---|---|---|---|
| Removal rating um at which percent efficiency equals: | ||||
| 005 | less than 0.5 | less than 0.5 | less than 0.5 | 0.5 |
| 007 | less than 0.5 | less than 0.5 | less than 0.5 | 0.7 |
| 010 | less than 0.5 | less than 0.5 | less than 0.5 | 1 |
| 020 | less than 0.5 | 1 | less than 1.5 | 2 |
| 030 | 0.7 | 1.8 | less than 2.5 | 3 |
| 050 | 1.5 | 2.5 | 4 | 5 |
| 070 | 3.5 | 5 | 6 | 7 |
| 100 | 6.5 | 7.5 | 9 | 10 |
| 120 | 7 | 9 | 11 | 12 |
| 150 | 8 | 10 | 13 | 15 |
| 170 | 8.5 | 11 | 15 | 17 |
| 200 | 9 | 12 | 18 | 20 |
| 300 | 12 | 16 | 24 | 30 |
| 400 | 15 | 20 | 30 | 40 |
| 700 | 25 | 45 | 70 | - |
| 900 | 35 | 50 | 90 | - |
It is frequently demonstrated that an absolute rated filter with an rating much larger than the competitor's filter nominal rating will give superior performance and will produce cleaner effluent
SDI and Pre-RO Filtration
The quality of the water being fed into the RO membranes has a critical effect on the RO membrane's performance and lifetime. SDI is the standard method to evaluate the suitability of the influent water to undergo reverse osmosis. Most RO membrane manufacturers will state a maximum value for SDI, usually 5 or less, for the influent water. Water that exceeds an SDI value of 5 is generally considered to be unfit for reverse osmosis, and using high SDI RO influent will usually void the RO membrane warranty. On the other hand, using low SDI water is observed to extend, sometimes double, the lifetime of the RO membrane. An added advantage to using low SDI RO influent is reduction of downtime and expense caused by the frequent cleaning needed to maintain RO membranes.
The quality of the source (usually municipal) water is observed to have a high degree of variability. IF the water is obtained from a well tapping a deep aquifer, the source can have a low SDI, from 3 to 5. When the water source is a lake, river, reservoir, or other surface source, the SDI can be quite high. Table B gives the SDI from various locations.
| City | Water Source | SDI |
|---|---|---|
| Albequerque, NM | Deep wells | 3-5 |
| Lubbock, TX | Lake Meredith | 6-20 |
| Dallas, TX | Mixed well/reservoir/river | 6-20 |
| Juarez, Mexico | Wells | 4 |
| Long Island, NY | Deep well | 1-2 |
| Central Indiana | Mixed well/reservoir/creek | 5 |
| San Diego, CA | Reservoir | 13 |
| Suburban Los Angeles | Colorado Delta/California Reservoir | 20+ |
| Santa Clara, CA | City | 3 |
| Portland, OR | City | 14 |
| Seattle, WA | City | 18 |
Raw city water must be treated to render it of sufficient quality to undergo RO. Pre-RO treatment can include softening, descaling, carbon beds, multimedia beds, etc. The only pre-RO treatment that is consistently effective in reducing the SDI of the influent water is the utilization of pre-RO filters. An ideal pre-RO filter has the following characteristics:
Types of Filters
Different types of filter construction are used in pre-RO filters. The major types of pre-RO filters tht may be encountered in the field are described below:
Membrane Filters: Membrane filters have a thin porous membrane of approximately 100 microns thickness. The membrane is usually supported by a nonwoven material, and mounted in the cartridge in a pleated configuration. Sealing of the cartridge in the housing is accomplished by either O-rings or gaskets. The high cost and low dirt capacity of many pleated filters place them at a disadvantage when they are considered for pre-RO applications.
Depth filters: The types of depth filters are:
String-wound filters: A length of cotton, polyester, or polypropylene twine is would around a supporting core structure. String-wound filters are usually sealed in a housing by knife-edged seals. The difficulties associated with string would filters include fiber shedding, unloading, variable or unreliable retention characteristics, and filter bypass.
Spun filament filters with uniform pore size: Filament are spun into a filter that provides a uniform pore size through the depth of the filter. The filter may have knife-edge seals, or the filter can be mounted in an industrial- or AB- style configuration. Filters of this type do not have optimal dirt holding capacity and retention characteristics.
Spun filament filters with a pore size gradient: The filter element has an inner (downstream) section containing filaments of a consistent diameter, providing constant pore size for absolute filtration. The inner section can be manufactured to provide absolute removal ratings as low as 0.5 micron and as high as 90 microns. The unique manufacturing technique provides the ability to change filter diameter instantaneously and continuously. This makes possible an outer (upstream) section in which filament diameter varies throughout. The upstream section contains pore sized varying over a range of 40 to 1, providing multiple levels of effective prefiltration in one element.
Bag filters: Bag filters consist of a nonwoven fibrous medium in a bag-like configuration. The bag material is usually thicker than a conventional membrane filter, but thinner than a depth filter. Sealing is with a ring compressing the material at the bag opening lip.
Test Methods
The filters were evaluated at various sites where water purification systems producing high purity water for semiconductor manufacture were located.
A portable filter evaluation system was used to test the quality of the unpurified water. The filter evaluation system consisted of a 10-inch filter housing connected to an SDI fouling index test jig. The plumbing of the filter evaluation system permitted the sampling of the water both upstream and downstream of the filter without interrupting the water flow throughout the filter.
The portable filter evaluation system was usually tapped into the water purification system immediately upstream the pre-RO filter housing. The SDI of the water influent to the pore-RO filters was measured, then the SDI of the effluent of several nominal and absolute rated filters (usually in the range of 1 to 20 microns) was measured. The SDI of the effluent of the on-line pre-RO filters was also determined.
SDI was measured according to ASTM D4189-82(1).
Results and Discussion
The quality of the water influent to the pre-RO filters encountered during the field studies can be generalized into three basic classifications: high SDI water, intermediates SDI water, and low SDI water. SDI water ratings were considered as follows: High is water having an SDI of 15 or greater. Intermediate is water having an SDI between 5 and 15. Low is water having an SDI lower than 5.
High SDI Water
Table C shows the results for high SDI water tested in Seattle, WA.
| Filter | SDI |
|---|---|
| Pre-RO influent | 17.6 |
| 0.5 micron (nominal) | 18.4 |
| 20 micron (absolute) | 11.3 |
| 15 micron (absolute) | 9.5 |
The water influent to the pre-RO filters had an SDI of 17.6. The 0.5 micron (nominal) filters that were on-line yielded effluent having an SDI of 18.4, indicating that the filters may be releasing contaminants into the RO membranes. When the pre-RO influent was passed through 20-, 15-, and 10 micron (absolute) filters, the resulting effluents had SDI values of 11.3, 9.5, and 4.3, respectively. Of the filters evaluated, the 10-micron (absolute) filter yielded effluent having an SDI less than 5, rendering the water suitable to be fed to the RO members.
Table D shows the results for high SDI water tested in Los Angeles, CA.
| Filter | SDI |
|---|---|
| Pre-RO influent | 20+ |
| 1 micron (nominal) | 19+ |
| 20 micron (absolute) | 17.1 |
| 15 micron (absolute) | 9.7 |
| 10 micron (absolute) | 4.4 |
The water influent to the pre-RO filters had and SDI of greater than 20. The 1-micron (nominal) filters that were on-line yielded effluent having an SDI of >19, indicating that the filters were doing little to lower the particulate load on the RO membranes. When the pre-RO influent was passed through 20-, 15- and 10-micron (absolute) filters, the resulting effluents had SDI levels of 17.1, 9.7, and 4.4, respectively. Of the filters evaluated, the 10-micron (absolute) filter yielded effluent having an SDI less than 5, rendering the water suitable to be fed to the RO membranes.
High SDI values result form obtaining the water from surface sources, such as lakes or reservoirs, having a high silt load. The silt in high SDI water is frequently very fine, and it can penetrate the pretreatment stages prior to the pre-RO filters. The proper selection of an absolutely rated pre-RO filter can reduce the SDI of the RO membrane feed-water to acceptable levels. However, the filter can have a short life due to the high particulate loading of the water. When dealing with high SDI water, the entire RO pretreatment system must be analyzed and optimized. Frequently, multistage filtration schemes must be considered to yield the most economical method to reduce the SDI to acceptable levels.
Intermediate SDI Water
Table E shows the results for intermediate SDI water tested in Portland, OR.
| Filter | SDI |
|---|---|
| Pre-RO influent | 11.8 |
| 5 micron (nominal) | 10.4 |
| 30 micron (absolute) | 10.8 |
| 20 micron (absolute) | 8.1 |
| 10 micron (absolute) | 3.8 |
The water influent to the pre-RO filters had an SDI of 11.8. The 5-micron (nominal) filters that were on-line yielded effluent having an SDI of 10.4, indicating that the filters were doing little to reduce the SDI of the water influent to the RO membranes. When the pre-RO influent was passed through 30-, 20-, and 10- micron (absolute) filters, the resulting effluents had SDI values of 10.8, 8.1, and 3.8, respectively. Of the filters evaluated, the 10-micron (absolute) filter yielded effluent having an SDI less than 5, rendering the water suitable to be fed to the RO membranes.
Intermediate SDI water is usually obtained when the originating water is mixed from several sources. These sources can include reservoirs, lakes, rivers, and wells. Since the mixing of the sources is usually done by the municipal water department, the end-user of this water can do little to control day-to-day quality of this water, which can be quite variable. The results of the testing shows that the proper selection of an absolute filter can provide RO influent having the required SDI of less than 5. The results of a study conducted in Dallas, TX (5), which has city water of variable quality, has shown that an absolute rated filter having a pore size gradient will yield consistently low SDI throughout the life of the filter.
Low SDI Water
Table H shows the results for low SDI water tested in Santa Clara, CA.
| Filter | SDI |
|---|---|
| Pre-RO influent | 3.0 |
| 1 micron (nominal) | 2.8 |
| 15 micron (absolute) | 0.6 |
| 10micron (absolute) | 0.7 |
| 7micron (absolute) | 0.4 |
The water influent to the pre-RO filters had an SDI of 3.0. The 1-micron (nominal) filters that were on-line yielded effluent having an SDI of 2.8, indicating that the filters were doing little to reduce the SDI of the water influent to the RO membranes. When the pre-RO influent was passed through 15-, 10-, and 7-micron (absolute) filters, the resulting effluents had SDI values of 0.6, 0.7, and 0.4, respectively. All of the absolute filters evaluated therefore yielded essentially equivalent SDI values, which were all below 1. The 15-micron (absolute) filter would be recommended since it not only has an effluent having an SDI value less than 1, but also has the most advantageous differential pressure versus flowrate and dirt-holding capacity.
The water influent to the pre-RO filters had an SDI of 2.7. The 5-micron (nominal) filters that were on-line yielded effluent having an SDI of 2.5, indicating that the filters were doing little to lower the particulate load on the RO membranes. When the pre-RO influent was passed through 20-, 15-, and 10-micron (absolute) filters, the resulting effluents had SDI levels of 1.6, 1.6, and 1.8, respectively. Of the filters evaluated, the 10-micron (absolute) filter yielded effluent having an SDI less than 1. In this case, the 10-micron (absolute) filter is recommended.
Low SDI water usually originates from wells rather than from surface sources. An additional advantage of using well water is the reduced likelihood of major variations occurring in the water quality with changes in the weather and seasons. Wells are frequently owned by the company using the water purification system, yielding a reduced cost when the price of buying water from the municipal water company is considered.
When the raw water has a low SDI value of 5 or less, some users questions the necessity of using pre-RO filtration. Pre-RO filtration is still necessary for the following reasons:
- An upset in the quality of the raw water can send a cloud of contaminants to foul the RO membranes. The pre-RO filters will prevent the contaminants from reaching the RO membranes.
- Low SDI water still contains particulates that can impact on the RO membranes, thereby necessitating more frequent cleaning of the RO membranes and shortening their life. This can be demonstrated by observing a pre-RO filter that has been used for tofilter low SDI water for several weeks. The filter will appear visibly coated with contaminants.
Conclusions
Field testing has demonstrated that filtration using filters having an absolute rating will yield RO influent having acceptable SDI values. SDI values values lower than 5 can be obtained by the proper filtration of high SDI and intermediate SDI water. The quality of water having SDI lower than 5 can also be improved by the proper selection of an absolute rated filter.
The observation of the performance of nominal rated filters in the filtration of the water influent to the RO membranes has demonstrated 5-micron (nominal) and finer nominal rated filters will do little to improve the quality of the water. The use of an absolute rated filter having several times the nominal numerical rating will usually give a superior performance in the filtration of RO membrane influent.
The results of testing using Dallas city Water (5) have shown that absolute rated filters having a pore size gradient were able to deliver effluent water having SDI values below 5 when used to filter water having intermediate to high SDI values. The absolute rated filters having a pore size gradient yielded effluent SDIs that did not increase with increase with increasing throughput. Increased service life has also been observed.
last modified June 20, 1998