Better RO results lie in design

Simon Gare of Ecolochem International examines potential causes of reduced performance from RO systems, methods of pre-treatment and factors upon which they are dependant

The aim of a reverse osmosis (RO) pre-treatment system is to condition the feed water for further processing, protect downstream equipment and minimise maintenance. One advantage of RO systems is that they reduce fluctuations in feed water quality, producing continuous and dependable product water. However, incorrectly designed pre-treatment can reduce the expected benefits of the technology. This normally results in lower than expected RO permeate flow or quality and often both.

In the last 10 to 15 years the take-up of RO technology has grown enormously, displacing more established processes to become a common and almost familiar part of any water treatment plant. Increasing numbers of users have proceeded hand-in-hand with the development of newer and better membranes, along with research into ever more diverse applications. The advantages for this are:

  • consistent product quality,
  • high availability,
  • water savings,
  • chemical savings,
  • manpower reduction,
  • ease of operation.

However, to reap the rewards offered by RO systems, correct design of pre-treatment is essential. Membrane systems, and RO specifically, are less forgiving in terms of incorrect specification and operation than more established technology such as ion exchange.

The aim of a pre-treatment system is to provide feed water at a quantity and quality that allows continuous operation of downstream equipment and protects that equipment from variability in the feed. Therefore, before designing an RO system an in-depth knowledge of the feed stream in terms of dissolved and suspended solids, total organic materials (TOC), potential for biological activity and variations in temperature is required. It is good practise to then obtain long-term data relating to the feed stream, to pick out unusual events that the RO system may face, such as peaks in suspended solids or TOC. Figure 1 shows the steps involved when designing an RO plant.

Identification of the end use and an available water source lead to the specification of the membrane type and required permeate quantity and quality. These factors then dictate the operational parameters of the RO system. These obviously have a huge impact on the system design and overall performance, but also produce the specification for any pre-treatment of the feed water that may be required to operate the RO. For make-up water treatment systems this may be simple. For process applications questions of compatibility and performance in non-aqueous solutions must also be considered.

Membrane options
RO membranes are manufactured from basically two materials, cellulose acetate (CA) and polyamide (PA). Alterations to their basic chemistries then give rise to a variety of sub classifications of membrane. Generic membrane types would include high rejection, low energy, low fouling and hot water membranes. The choice of one of these membrane types to meet the particular needs of an end-user, be that high salt rejection or the ability to hot water sanitise will require compromise in terms of membrane choice, system design and operational requirements.

Most large RO systems (greater than 100gpm) are designed around the familiar spiral wound membrane type. The dominance of the spiral configuration is due to its compact nature, because a large membrane area can be packed into a small volume creating cost savings in terms of pressure vessels, piping and pumps. This design increases the potential for operating systems at high recovery while saving energy via low-pressure drops. The industry standard 8" diameter, 40" long
element can pack a vast membrane area into small volume, typically 350-440ft² of active membrane into a volume of ~1.2ft³. Other configurations are available including flat sheet and hollow fibre.

The choice of membrane type and configuration are initially chosen based on characteristics of the feed solution and required permeate quality. By choosing a particular membrane type, particular feed water characteristics are demanded for consistent operation. With any commercially available membrane, the manufacturers detail very specific operating requirements both for system design and feed water quality. Regardless of membrane type - CA, PA or any of the sub categories - the guidelines detail the maximum allowable suspended solids in the feed water, generally stated in terms of a Silt Density Index (SDI). The dense packing of all spiral wound membranes means that should large or excessive levels of suspended solids be sent to an RO membrane they are likely to foul. The result is over pressuring and eventually membrane damage. Once fouling begins, cleaning of the membranes becomes very difficult. As a general rule, most membrane systems will not return to their original performance levels once fouling has occured4. Guidelines for SDIs from membrane manufactures5,6 state SDI15s of less than 5 for operation but SDI15s of less than 3.5 for continuous operation. The relevance of these statements will be dealt with in a later section.

Other feed water contaminants may adversely affect a membrane's performance or life-span, these factors can be generalised by looking at the strengths and weaknesses of particular membrane materials (See Table 1).

PA membranes are generally the membrane of choice in today's water treatment market. This is due to its high salt rejection and lower operating pressures when compared to the older CA membranes. Other advantages are their ability to withstand the extremes of pH and their higher temperature stability. However they do have disadvantages, primarily their susceptibility to attack by oxidants. While oxidants such as free chlorine can be easily removed from water sources, and therefore prevent membrane damage, PA membrane systems then become vulnerable to excessive biological growth. Bio fouling of RO systems is the most frequent cause of poor performance.

PA membranes also exhibit an anionic surface charge. This surface charge limits the use on cationic coagulants and flocculants. If they are introduced into the feed stream, near permanent fouling of the RO membranes will occur.

RO system design
Bearing in mind the operating requirements set out by the membrane manufacturer, a pre-treatment system must be designed to provide a suitable quality and quantity of feed water. However, the design of the membrane system itself can make the job of pre-treating the feed water potentially more challenging. Key factors in RO system designs are:

  • flux - permeate production per unit area,
  • cross-flow velocity - speed of the proportion of the feed stream that does not pass through the membrane,
  • recovery - percentage of the feed stream that is recovered as permeate.

Manipulation of these factors governs the rate of particulate fouling, risk of bio fouling and scaling potential. Therefore, good RO design will balance the key factors to minimise these negative results.

How and why these effects come about and how they can be minimised must be based on an understanding of what is occurring at the membrane surface. As water is forced through the RO membrane, the rejected salts and any solids present in the feed stream are concentrated in the concentrate and left behind. At the membrane surface the flow of water is almost static due to friction against the membrane. As the flow of water is almost static high concentrations of salts and particulate matter can build up, this forms a gel layer. This phenomenon is referred to as concentration polarization¹ (See Figure 2). The static water near the membrane surface is separated from the well-mixed concentrate stream at the boundary layer. It is due to these flow conditions that the salt concentrations can build up, as the only way to reduce the concentration in the gel layer is to diffuse away from the membrane surface, which is very slow.

The higher the flux and recovery of the system, fewer membranes are needed to produce the required permeate flow and therefore, smaller feed pumps. However this has the effect of reducing the cross-flow velocity. It is cross-flow that provides mixing within the concentrate stream and reduces concentration polarization by reducing the thickness of the boundary layer, i.e. scale-forming ions can be kept from reaching saturation levels and because there is enough energy in the concentrate, small suspended solids can be carried through the process without fouling it.

Without sufficient mixing within the concentrate, the permeate total disolved solids (TDS) would also be higher than expected. This is because the passage of dissolved salts through the membrane is concentration driven (the water flux is pressure driven) i.e. the greater the recovery, the higher the salt concentration at the membrane surface and therefore, more salt passes through the membrane into the permeate.

Table 1
Characteristics of two membrane materials
Cellulose actetate
Salt rejection
Driving pressure
150-250 psl
200-400 psi
Pre-treatment requirements
very strict
pH limits
Surface charge
Cleaning frequency
Frequent (weeks to months)
Infrequent (months to years)
Organics removal
No problem
No tolerance

It is clear then, that flux and recovery chosen at the design stage must be based on accurate knowledge of the feed water source and the likely quality following some form of pre-treatment. Choosing too high a flux will lead to particulate fouling on poorly treated surface water and higher than expected cleaning frequencies. This will reduce permeate quality, quantity and result in reduced membrane life. Guidelines for flux rates on various feed water sources are given in Table 2.

All major RO membrane manufacturers provide computer projection packages that can assist in the design of membrane systems and identify limits of a potential design. However these programs do not always plainly identify the feed water requirements that may be demanded by a potential design. The model may well show no errors but in practise the system may foul or scale very quickly if the user does not understand the effects of high recovery, low cross-flow and high flux.

Table 2
Suggested minimum and maximum flux rates for different water sources7
RO Permeate
1-14 GFD
14-18 GFD
18-30 GFD

The filtration spectrum (See Figure 3) illustrates some of the solid materials that are present in naturally occurring feed waters, particularly those of surface origin, and separates them in terms of their physical size. Filtration and various membrane's processes are shown. The range plotted reflects their pore size and therefore, the size of particles they will reject as they act as a sieve.

Almost all water sources require some form of pre-treatment because of suspended solids, particularly water from rivers and lakes. Even where water is treated to potable standards, SDI measurements may be below 5 but will not typically be below 3.5. Where RO is to be used in some form of process application or effluent recycling system, solids removal can be very difficult, due either to the nature of the solids and/or the mass of solids. There are many types of filtration systems but these can be grouped together based on the size of solids they are normally designed to remove (See Table 3).

For large solids such as gravel, sand and other large pieces of detritus, common removal processes would include settlement tanks, lagoons or grit screens. While all are well-established techniques, the first two simply rely on the weight of the particles being heavier than the water. The third acts as a sieve. Course solids do not usually cause fouling or plugging within RO systems because they are easily removed. It is smaller particles that require either very long settling times or some form of assistance to be removed, i.e. flocculation and coagulation. Figure 4 shows the time in seconds for particles of various sizes to fall through 1m of static water. It is possible to build settling ponds and lagoons to remove silt sized particles, but the physical size and civil engineering required for such a process would be prohibitive.

Settling down
From Figure 4 the settling time of silt is ~3 hours, for colloidal particles ~3 years. To improve the removal of small particles clarification can be used. This process is based on settling rates similar to settling ponds or lagoons, but the mass and the settling rate of the solids is increased by addition of inorganic or organic coagulant and/or flocculant. By their addition, smaller particles floc, or clump, together.

Table 3
Particle size removal verses potential unit process
>0.1 mm
RO Permeate

Dissolvee salts
Grit screens
Settling ponds and lagoons
Mutimedia filters
Micro filters
Ultra filters
Reverse osmosis

It is small particles, such as colloids that contribute to RO fouling. The term colloid can be used to refer to many different particles in water, from clay particles (<1,000µm diameter) to poorly ionised organic acids. It is for this reason that filtration systems based on settling, do not typically provide water with a suitable SDI for operating RO.

Coagulation can be added to improve performance, but continuous operation will see spikes in feed water quality which will lead to poorly clarified water being fed to down-stream processes. Carry-over from clarifiers is also a problem - either due to incorrect dosing of coagulant or high flow rates through the equipment. Control can be a major limitation on using a clarifier to pre-treat RO. A potential solution for the above is media filtration. This often allows clarifiers to catch floc carry-over and mop-up very fine floc or overfeeds of coagulant.

Media filters operate by sieving out particles of greater size than the spaces between the media grains. The filters can therefore be single media, where the media is all approximately the same size and density, or, two or more media types, of different size and density, can be combined in the same vessel. In this process the lighter and coarser material sits on top of the denser, finer material. In this way the filter will act to remove more than one size of particle. In all media filters the filtered solids act as an active filtration layer within the bed, further aiding the removal process. Typically, good SDI15 of 3 or less can be obtained and maintained from these systems.

Table 4
A list of contaminants that should be quantified to correctly design an RO pre-treatment system
Site Name
Feed water source
Location of sample point
Water ion analysis (mg/l as ion or CaCO3)
Other constituents
Metals: eg. Fe+++, AI+++, Mn++, Zn++, Cu++, Sn++
Other ions: ef. NH+4, Br¨
Gases: (C12, H2S, O2)
Suspended or colloidal matter (Silt Density Index (SDI))
Organic substances (TOC, BOD, COD)
Biological activity (TBC)


It is possible to use multimedia filters alone on many feed water sources where the feed concentration of suspended solids, including coagulant addition, is less than 100ppm. Once the loading in the feed water has exceeded this level the run length, or time between backwashes, will be too short. This would give rise to excessive pressure drops and potentially poor quality filtered water.

Screens, clarifiers and multimedia filters are established processes which, when designed correctly, can give excellent results in terms of suspended solids removal and reduction of colloidal species such as humic acids and colloidal silica. Clarifiers are effectively large tanks and for industrial flow rates of over a 1,000gpm become intensive in terms of space and civil engineering costs. Also, all filtration processes will give rise to a waste stream containing the filtered solids and any coagulant that was used. Media filters may generate 3-5% wastewater compared to the treated water throughput. If this water cannot be returned to the feed water source (typically not if a coagulant has been added), the derived sludge must be dewatered, either by settlement or some mechanical process.

The last five years have seen the development of alternatives to clarifier and filter systems, these are based on membranes. The types of materials used are extensive, including polypropylene, alumina and cloth. The configurations and modes of operation in which they are used are also numerous.

Membrane based filter systems can offer absolute guarantees regarding the filtered water quality as they have fixed pore sizes. However, flux rate, cross-flow and cleaning frequency are very important in the design of these systems. Many of the systems available do foul and have high cleaning frequencies or redundancy to cope with this. Maintaining permeate flow is, therefore, always a concern. Pilot testing of these installations is highly recommended before proceeding with a full-scale design.

Case Study One
Figure 5 shows a simple process flow diagram for the pre-treatment of this system. The most challenging aspects of the feed water to this particular system were the potential biological activity and the possibility of ammonia within the feed water source. The solids in the feed are present due to carry-over from the clarifier within the activated sludge process that provides the water. Any pre-treatment system used would require a high dose of biocide to kill the bio solids to prevent bio fouling within the RO and for removal of the solids by coagulation filtration.

For this application chlorine was originally chosen as the biocide. However, due to the presence of ammonia the demand was in excess of 100ppm on occasions. For this reason chlorine dioxide was used instead. The dosage was significantly reduced (~20ppm) and the fast acting and extremely aggressive nature of ClO2 gave good quality, sterile water to the RO system. Suspended solids following the primary filtration are typically 1 nephelometric turbidity unit (NTU), the filters having a backwash frequency of 2 in 24h. Following the polishing filters SDI15s of 3.8 to 4 are maintained.

Following the installation of this plant, significant variation in the feed water quality was noted. This resulted in poor performance of the RO and unscheduled cleaning. The solution to the problem was installation of turbidity monitors to give an early warning when solids were likely to cause a problem and the provision of a secondary low solids water source, which was used to backwash the filters when the solids loading was high, therefore keeping the filters clean.

Chemical Methods
Chemical pre-treatment to RO systems is primarily concerned with the prevention of scaling due to the presence of poorly soluble salts and the minimisation or control of biological growth. The source of the feed water and the membrane type dictate the chemical additions necessary. Table 4 is a potential list of contaminants that should be quantified to correctly design an RO system and its required pre-treatment.

Chemical additions should always be kept to minimum where possible, over addition will increase the feed TDS to the RO system and therefore the RO permeate TDS. Table 5 lists common foulants within RO systems, their potential sources and possible pre-treatments. From Table 5 it should be noted that over addition of some chemicals within the RO pre-treatment system might also harm the RO's performance. Inorganic coagulants based on iron or aluminium and organic polymers are good examples of this.

Scaling of RO is closely linked to the system's recovery and cross-flow. As a general rule, the higher the recovery of the system, the more likely it is that scaling will occur. Sufficient control and monitoring of the RO and its pre-treatment is vital.

In systems that operate at low recoveries (65-70%) with little or no iron present, the addition of acid may be the only chemical pre-treatment necessary - this will prevent scaling of the RO system with calcium carbonate. The feed waters pH will typically be adjusted to ~6.0 at which point the majority of the carbonate within the water will be found either as bicarbonate or carbon dioxide - these do not form scales with calcium. Acid addition is a simple system to control using the pH of the water as feedback to meter the dosage of acid required. This potentially creates two problems. Firstly, the CO2 created will freely pass into the RO permeate requiring degassing downstream. Secondly, large doses of acid will increase the feed TDS to the system. If hydrochloric acid is used, this will increase the permeate TDS as the chloride added is poorly rejected by the membrane. If sulphuric acid is used, the permeate TDS will not be as badly affected, as sulphate is well rejected by RO membranes. However, if the dosage of sulphuric acid is high enough, precipitation of calcium sulphate becomes a potential risk.

As recovery increases to 75-80%, acid addition alone cannot control all of the species within the feed that are likely to scale. There are two potential alternatives to this problem:

  • remove the scale forming ions from the feed with some form of softening
  • use a scale inhibitor

Ion exchange softening is an excellent pre-treatment to RO systems. By removing calcium, magnesium, barium and iron II the formation of carbonate, sulphate and fluoride scales is prevented. Additional benefits from this treatment are the polishing filtration provided by the ion exchange beds further improving SDIs. This approach to pre-treatment is particularly suited to systems using PA membranes. As the membrane is stable over a wide range of pH, no acid addition is required when compared to CA membrane systems. Therefore, carbonate species can be rejected by the RO as NaHCO3, reducing the need for down stream degassing of CO2.

As with all ion exchange processes, the regenerant water from the process can be difficult to dispose of due to the high salt content. However, by softening the feed water to the RO the concentrate can occasionally be used as a secondary make-up water source to cooling systems. This improves the chemical efficiency of the RO system and allows for greater cycling within the cooling systems, thereby reducing fresh water needs.

The alternative to removing the divalent cations from the feed water is to use an inhibitor. Most inhibitors or antiscalents used for water treatment at this time can be classified as threshold inhibitors (Figure 6a shows normal scale formation, Figure 6b shows the method operation of threshold inhibitors). These materials are extremely good at preventing carbonate and sulphate scales. Using these, it is possible to operate at higher recoveries without significantly altering the feed water composition, as the dosage of antiscalent when compared to acid is extremely small. The need to de-gas CO2 from the permeate is reduced because the antiscalents do not lower the pH and therefore evolve CO2 from the alkalinity present. It is important to realise this when designing RO plants, as these materials do not prevent scaling from occurring but delay the formation of large crystals that form scales. However, the effectiveness of these chemicals is quite short-lived - potentially only 30 minutes. This effects the operation of the RO. While 30 minutes should be more than the hydraulic retention time of the feed water within the RO, when the RO shuts-down the concentrate may start to scale the membranes. RO systems that use these forms of scale inhibition should be designed to go through a flush sequence when shut-down, even under alarm conditions.

Table 5
Common foulants and scalents within RO systems and potential pre-treatments4
Sulphate scale
Feed water/acid addiction
Antiscalents/IX Softners
Metal hydroxides
Feed water / (Fe(ll) / Coagulant addition (Allll)
Oxidation of reduced species
IX Softners
Control of coagulant addition

Free water/free chlorine
Adsorption (GAC)
Addition of a reducing agent

Dissolved organics

Feed water / humic acids / polymers
Coagulant and filtration
Control of polymer addition
Bio films

Internal to RO
On line biocide addition
Off line sanitisation
Suspended solids
Feed water
Coagulation and filtration

Biological fouling of RO membrane systems is the most common cause of reduction in permeate flow and quality3. This is because most water sources that RO systems operate on are not sterile. Solids removal techniques such as clarification and filtration (media or membrane based) can remove bacteria and higher organisms from the feed stream by particle exclusion. This can be improved by the addition of a biocide to the feed stream to kill off any organisms present. At this point membrane choice is again a critical factor, on surface water sources with a high potential for biological fouling the continuous addition of an oxidising biocide may be beneficial. If a PA membrane has been chosen for this duty any residual of the oxidising biocide must be removed to prevent severe membrane damage. Typically, chlorine tolerances for PA membranes are quoted at around 1,000ppm/h6, therefore the membrane's rejection will drop below that quoted after 500h operation with a feed concentration of 2ppm. CA membranes can have less problems with biological fouling due to their tolerance of oxidants, but this must be measured against their lower salt rejection and pH range requirements.

Oxidant removal
Where a PA membrane has been chosen the removal of oxidants is therefore required. This can be done by either the addition of a reducing agent, such as sodium bisulphite or adsorption of the free oxidant, such as chlorine on to granulated activated carbon (GAC). Both options have pros and cons. The addition of a reducing agent relies on the correct dose being injected into the feed stream and adequate mixing and contact time of the reaction to occur. Over addition is quite possible because the residual of the oxidant may vary and the feed water TDS to the system will be increased. Most importantly, the system is vulnerable to mechanical failure of the dosing pumps, to avoid this GAC can be used as a polishing filter and absorber for oxidants.

GAC is well suited to pre-treating RO systems fed with low suspended solids and biological activity or waters which have already been treated for solids removal, particularly town's water sources. Town's water is typically supplied with a residual of a free oxidant, usually ~0.05ppm chlorine in the UK. SDI15s measured on raw town's water will typically reach 4-5, therefore polishing filtration will further increase the water quality to ~3.5. The solids found in drinking water supplies tend to be rust from the distribution system or fine silt.

Once oxidants have been removed from the feed water to a PA system there is generally no other form of continuous biological control. As a result, any organism that passes into the system is likely to be trapped within it. Microorganisms are capable of colonising almost every material or surface known to man3, therefore in within RO systems biological growth must be controlled. Out-of-control biological growth is typically referred to as bio fouling due to the presence of thick films of cells and their secretions. It is identified by increased pressure drops, reduced permeate flow and decreasing salt rejection. Where an oxidising biocide can not be added, such as in a PA system, non-oxidising biocides may be used. These are not added while the system is on-line due to the slow action of these materials - usually the RO will be placed in a rinse mode for this procedure. Examples of non-oxidising biocides that may be used are gluteraldehyde and isothiazolone. To control biological growth over extended periods it is important to vary the biocide used, this prevents the build up of colonies that are tolerant of the material used.

Case Study Two
Figure 7 shows simple process flow diagram of the pre-treatment system. The key factor controlling the design of this system was the silica content of the water. Due to the local geology the feed water was almost saturated with it. Silica is highly insoluble, the maximum concentration at pH of 6 to 9 being ~130ppm, therefore, the use of some form of scale inhibition was required. Threshold inhibitors on the market will guarantee prevention of silica scale up to a concentration of ~220ppm8. This would limit the recovery of the plant to ~58% which was considered an uneconomic use of the available resource. By increasing the pH of the feed water the solubility of silica can be enhanced, at a pH of 9.8 the solubility of silica is increased to ~240ppm. This allows the system to operate at a recovery of ~65%.

However at a pH of 9.8 any hardness present in the feed will combine with carbonate to form calcium carbonate and scale the system. Therefore it was necessary to soften the feed water to less than 1ppm total hardness as CaCO3. An advantage of using the softener system up-front of the RO is that colloidal silica present in the feed is filtered out to give SDI15s of <3.5.

This system has operated well producing high quality feed water to the customer, cleaning frequencies have been low and no evidence of bio fouling has been detected. Hardness leakage, or upsets to the softeners operation, are a potential weakness within this system, for that reason polishing softening was installed.
In exploring approaches to RO pre-treatment systems and some of the factors upon which they depend, such as water source, water availability, membrane type, downtime and environmental impacts, we have examined potential causes for reduced performance from RO systems and how in some circumstances it may be the pre-treatment system that introduces the foulant.

Two facts are certain, RO systems are still fouling and the business of RO cleaning chemicals is a growth area. This is due to poor or aggressive RO design and results in the inability of pre-treatment systems to supply water to the required quality.

To improve RO system design, good quality feed water data over an extended period should always be sorted out and attention paid to membrane manufactures' guidelines. This will produce more reliable systems and continue the rise of RO and other membrane technologies into the 21st century.

1) Course notes from Membrane Processes, short course by the School of Water Sciences, Cranfield University, 2000.
2) Betz Handbook of Industrial Water Conditioning, 9th Edition, 1991. USA.
3) Membranes and Micro Organisms, H.C. Flemming, Proceedings of Membrane Technology in Water and Wastewater Treatment 27th -29th March 2000, Royal Society of Chemistry.
4) Death, Taxes and RO Membrane Fouling, Nancy Mulhern, Osmonics Inc. 11/01/95 Water Technology.
5) Koch Membranes Catalogue, 2000. Koch Membrane Systems, The Granary, Stafford.
6) Osmonics Membrane Catalogue, 2000. Osmonics, Inc. Minnetonka, Minnesota. USA.
7) Important considerations in the design and construction of semiconductor water systems. Steven Gagonon, Feb 2000. Ultrapure Water. Pub Tall Oaks. USA.
8) Permacare International presentation to Ecolochem International, Oct 2000. Houston, Texas.
9) Reverse Osmosis Pre-Treatment of High Silica Waters, Simon Gare, Ecolochem International, Proceeding of the International Water Conference Oct 2000, Pittsburgh USA.


| CO2 | Cycling | gas | surface water | wastewater treatment


Click a keyword to see more stories on that topic, view related news, or find more related items.


You need to be logged in to make a comment. Don't have an account? Set one up right now in seconds!

© Faversham House Group Ltd 2002. edie news articles may be copied or forwarded for individual use only. No other reproduction or distribution is permitted without prior written consent.