Ultrapure water production: The inside story
By Peter Cartwright, P.E., Cartwright Consulting Co.
As both the quality and quantity requirements for contaminant-free water increase, the demands for innovative technologies and improved system designs are creating challenges and opportunities for the multitude of industries that require ultrapure water.
All water supplies contain contaminants. The kind of contaminant is hugely variable and no two water sources are identical with regard to the kind and concentration. What constitutes a contaminant is entirely dependant on the application; for drinking water, it is defined by the Safe Drinking Water Act, a regulatory document. For semiconductor rinsing, anything other than H2O is a contaminant and the concentrations must be as close to zero as possible.
As it is virtually impossible to make water free of any and all contaminants, the goal of a treatment process is to reduce the level as much as possible.
It is possible to classify contaminants by category to more easily address their removal (see Table 1). There is no shortage of water treatment technologies available. Some remove only a single class of contaminants, while others are more versatile. Each technology has strengths and weaknesses. No single technology will produce truly ‘ultrapure’ water.
As a result, the challenge is to design a system utilizing a combination of technologies to provide optimum contaminant removal to meet the particular ‘use-specific’ water quality requirements.
The pressure membrane technologies of microfiltration, ultrafiltration, nanofiltration and reverse osmosis are the most versatile and, hence, most widely used as the lynchpin of most ultrapure water production systems.
In particular, membrane technologies possess certain properties that make them unique when compared to other water treatment technologies. These include:
■ Continuous process, resulting in automatic and uninterrupted operation
■ Low energy utilization involving neither phase nor temperature changes
■ Modular design-no significant size limitations
■ Minimal moving parts with low maintenance requirements
■ No effect on form or chemistry of contaminants
■ Discreet membrane barrier to ensure physical separation of contaminants
■ No chemical addition requirements
Simply put, these technologies are continuous filters. The form of contaminant removed is a function of membrane polymer selection and its pore size. Although they all provide separation of contaminants from water, each performs a specific function and has specific advantages and disadvantages when compared to the others in a particular application.
The development in filtration technology known as ‘crossflow’ or ‘tangential flow’ filtration allows for continuous processing of liquid streams. In this process, the bulk solution flows over and parallel to the filter surface and, because this system is pressurized, water is forced through the filter medium and becomes ‘permeate.’ Turbulent flow of the bulk solution across the surface minimizes the accumulation of particulate matter on the filter surface and facilitates continuous operation of the system. Figure 1 compares the crossflow mechanism with conventional filtration.
Generally, microfiltration (MF) involves the removal of particulate or suspended materials ranging in size from approximately 0.01 to 1 micron (100 to 10,000 angstroms). Figure 2 depicts the mechanism of microfiltration (MF).
Ultrafiltration (UF) is used to separate materials typically smaller than 0.01 micron (100 angstroms). The removal characteristics of UF membranes can be described in terms of molecular weight cutoff (MWCO), the maximum molecular weight of compounds that will pass through the membrane pores. MWCO terminology is expressed in daltons. Basically, ultrafiltration is used to remove dissolved nonionic contaminants, while suspended solids are removed by microfiltration (see Fig. 3).
Nanofiltration (NF) is an intermediate process between ultrafiltration and reverse osmosis. The MWCO properties of nanofiltration membranes are in the range of 300 to 800 daltons (<10 angstroms). Ionic rejections vary widely depending upon the valence of salts; multivalent salts such as magnesium sulfate (MgSO4) are rejected as much as 99 percent, while monovalent salts such as sodium chloride (NaCl) may have rejections as low as 10 percent (see Fig. 4).
The reverse osmosis (RO) process removes all dissolved organic (nonionic) solids with molecular weights above approximately 100 daltons, as well as a high percentage of ionic materials. Because reverse osmosis membranes are not perfect (they will typically remove 95 to 99 percent of the ionic contaminants), they are generally used as pretreatment to a final ‘polishing’ deionization unit for high-purity water production (see Fig.5).
To be effective, membrane polymers must be packaged into a configuration commonly called a ‘device’ or ‘element.’ The most common element configurations are: tubular, capillary fiber, spiral wound, and plate and frame (see Fig. 6)
Manufactured from ceramic, carbon, stainless steel or a number of thermoplastics, these tubes have inside diameters ranging from 3/8 inch up to approximately 1 inch (10 to 25 mm). The membrane is typically coated on the inside of the tube and the feed solution flows through the interior (lumen) from one end to the other, with the permeate passing through the wall to be collected on the outside of the tube.
Capillary (hollow fiber)
These elements are similar to tubular elements in design, but are smaller in diameter, are usually unsupported membrane polymers and require rigid support on each end provided by an epoxy ‘potting’ of a bundle of the fibers inside a cylinder. Feed flow is either down the interior of the fiber or around the outside of the fiber.
This element is constructed from an envelope of sheet membrane wound around a permeate tube that is perforated to allow collection of permeate. Water is purified by passing through one layer of the membrane and flowing spirally into the permeate tube. It is by far the most common configuration in water purification applications.
Plate and frame
This element incorporates sheet membrane stretched over a frame to separate the layers and facilitate collection of the permeate, which is directed into a center tube.
From the perspective of cost and convenience, it is beneficial to pack as much membrane area into as small a volume as possible. This is known as ‘packing density.’ The greater the packing density, the greater the membrane area enclosed in a certain-sized device and, generally, the lower the cost of the membrane element. The downside of the high-packing-density membrane elements is the increased propensity for fouling (see Table 2).
Figure 7 illustrates a complete membrane processing system. In terms of function, it could also represent a single membrane element. Note that the ‘feed’ stream enters the system (or membrane element) and as the stream passes along and parallel to the surface of the membrane under pressure, a percentage of the water is forced through the membrane polymer producing the permeate stream. Contaminants are prevented from passing through the membrane based on the polymer characteristics. This contaminant-laden stream exits the membrane system (or element) as the ‘concentrate’ stream, also known as the ‘brine’ or ‘reject.’
The percentage of feed flow that passes through the membrane and becomes permeate is known as ‘recovery.’ Typically, for water purification applications, recovery is set below 85 percent. As recovery is increased (to decrease the concentrate volume), the concentration of contaminants in the concentrate stream increases significantly. This effect is mathematically developed and illustrated in Table 3.
The vast majority of membrane element device and system failures are caused by membrane fouling, which is usually the result of one or more of the following mechanisms:
■ Suspended solids in the feed stream due to improper feed water filtration
■ Precipitation of insoluble salts or oxides resulting from concentration effects within the membrane device
■ Biofilm caused by microbiological activity
These mechanisms cause the membrane surface to become coated with fouling materials that build up in layers. As the layer thickness increases, the flow rate across the membrane surface and immediately adjacent to it decreases, reducing local turbulence and encouraging more settling of suspended solids, which increases the fouling layer thickness-a vicious cycle.
With nanofiltration and reverse osmosis membranes, which reject ionic contaminants, fouling usually creates a phenomenon known as ‘concentration polarization.’ The fouling layers inhibit the free movement of the feed stream away from the membrane surface and, as salts are rejected from the membrane, their concentration at the surface is higher than in the bulk solution (that portion above the fouling layer).
Since ionic rejection is always a percentage of the salt concentration at the surface of the membrane, the permeate quality decreases as a direct result of concentration polarization and this phenomenon may actually indicate the presence of foulants before a reduction in permeate flow is detected. The increased salt concentration at the membrane surface also promotes precipitation of those salts whose solubility limit is exceeded as a result of concentration polarization.
For ultrapure water production, reverse osmosis is virtually always used and, as this membrane technology is the most susceptible to fouling, pretreatment is usually necessary.
Additionally, reverse osmosis by itself will not produce ultrapure water (by most definitions). As a result, most systems utilize additional technologies to polish the reverse osmosis permeate. This approach of breaking the system design down into components has resulted in the concept of looking at every system as the optimum combination of pretreatment, primary and post treatment technologies.
Pretreatment technologies are dictated by the raw water quality and limitations imposed by the reverse osmosis membrane polymer. If the raw water is prone to calcium carbonate scaling (positive Langelier Index), pretreatment should include one or more of the following: softening, acidification or dispersant addition. Excessive iron (above 0.3 ppm) can be removed with a manganese greensand filter or oxidation and filtration. If the turbidity is above 0.1 NTU, a backwashable multi-media filter should be used. Cellulosic reverse osmosis membrane polymers are sensitive to hydrolysis at a pH above 7.0; this requires that acidification be used with high pH water supplies.
Activated carbon is a pretreatment technology capable of removing residual chlorine, which is essential when thin film composite reverse osmosis membrane polymers are utilized. In those applications where cellulosic polymers are used, the activated carbon unit is normally placed downstream of the reverse osmosis unit. Activated carbon filters must be backwashed to remove accumulated particulate material and require periodic replacement of the filter media.
As stated above, reverse osmosis is usually the key technology utilized for ultrapure water production. To achieve the required ultrapure quality for the specific application, one or more of the following technologies are used.
Mixed bed deionization
Mixed bed deionization (DI) will ‘polish’ the purified water up to 18 megohm-cm resistance, the maximum ionic purity attainable in industrial systems. Because deionization is a batch process, consideration must be made for off-line regeneration.
Obviously, if the system is used continuously, another identical DI unit must be available to allow time for regeneration of the exhausted resin without total system shutdown.
Resin beds that sit idle for more than 48 hours at a time may contribute to microorganism problems in the water treatment system.
The newest development in high-purity water production is a technology known both as electrodeionization (EDI) and continuous deionization (CDI) (see Fig. 8). This process is basically a combination of electrodialysis (ED) and resin deionization (DI). The DI resins are enclosed between layers of ED membranes. The energy to effect separation is electrical, imparted to positive and negative electrodes. The DI resins do not adsorb ionic contaminants, but facilitate ion movement into the concentrate streams.
When fed RO permeate, EDI will produce 18 megohm-cm quality water. It is a continuous process and does not require regeneration.
In general, once the water has been treated to achieve the desired purity, it is directed to a storage tank, which is typically constructed of inert materials and is sized to hold anywhere from several hour’s to a full day’s requirement. It is typically either vented to the atmosphere with the tank protected from atmospheric contamination by a submicron vent filter, or it is sealed with a blanket of inert gas such as nitrogen. The storage tank receives water directly from the primary treatment system as well as water from the recirculation loop.
Because ultrapure water is extremely aggressive and will become contaminated by virtually anything with which it comes into contact, the distribution loop from the storage tank to the points of use generally requires technologies to continuously remove these contaminants. The technologies defined earlier are often utilized as part of this post treatment. As the recirculation rate in this loop is usually much higher (and more variable) than the production rate to the storage tank, the technology components must be sized accordingly.
Typically 0.1- or 0.2-micron filters are used to remove particulate materials and live bacteria. These can be either conventional ‘dead-end’ cartridges or crossflow membrane devices. It is essential that they be manufactured from materials that will not leach or slough off into the pure water stream.
Ultrafiltration often provides the final polish. With typical molecular weight cut-offs in the range of 5,000 to 100,000 daltons, UF is effective in removing most of the residual contamination in the system. Typically, ultrafiltration units are designed with recoveries of 95 to 98 percent, meaning that between 2 and 5 percent of the water flow is directed to the drain or recycled to the front of the system. Again, it is essential that all materials of construction in contact with the highly aggressive pure water be completely inert.
This unit is intended to reduce bacterial propagation throughout the storage tank and distribution piping. Although ultraviolet irradiation (UV) does not remove microorganisms-and there is some debate with regard to its ability to completely kill bacteria-it does inhibit bacterial growth and is an effective component of any high-purity water system.
Considered the most effective disinfectant available, ozone will also break down organic compounds, theoretically into their basic elements. It is so aggressive that special materials of construction must be utilized and it must be removed (usually with 254 nm UV) before the water can contact membranes or resins.
Today, the industries that are the largest consumers of pure water include: semiconductor manufacturing-for rinsing of electronic devices (computer chips, etc.); the power industry-for high-pressure-steam-generating boilers; the pharmaceutical industry-for manufacturing operations requiring USP or WFI water; hemodialysis-for preparation of dialysate solutions and rinsing artificial kidneys; and medical laboratories-for analytical and research activities.
Although each industry requires pure water that is ‘contaminant-free,’ the particular contaminants of concern and their acceptable residual levels vary according to the application.
As analytical techniques become increasingly more sensitive, it has become obvious that there is no such thing as water that is completely free from all contaminants. Also, as water is purified it becomes more and more aggressive and will start to dissolve most materials with which it comes into contact.
Recognizing the practicality of this situation, each industry has established pure-water quality requirements that constitute a compromise between performance and economic reality. Tables 4 and 5 provide examples of water quality standards or guidelines for the semiconductor manufacturing and pharmaceutical industries.
Figure 9 illustrates a generic design for a typical pure-water treatment system. Although the optimum configuration is a function of the factors previously discussed, this system is representative for most applications.
In ultrapure water production, the optimum design requires the following input:
■ Feed water quality
■ Ultrapure water quality requirements
■ Ultrapure water quantity requirements
Of critical importance are the knowledge, experience and capability to select and implement the appropriate technologies into a complete, comprehensive, reliable and economical system.
Peter S. Cartwright, P.E., specializes in both marketing and technical consulting in high-technology separation processes. He can be reached via e-mail at email@example.com.