Reverse Osmosis Membranes: Transport Models, Fouling Diagnosis and Operating Limits
What the membrane separates, and what it does not
Reverse osmosis is a pressure-driven membrane operation. Feed is pressurised on one side of a semi-permeable barrier; once the applied pressure exceeds the osmotic pressure of that solution, solvent crosses in the direction opposite to natural osmosis. The permeate is drawn off at low pressure, and a concentrated stream is left behind. Applied to seawater, the low-pressure side yields fresh water and the high-pressure side yields brine.
Full-scale elements are asymmetric or composite membranes wound as hollow fibres or spiral-wound leaves. The barrier retains inorganic ions, colloidal matter and high-molecular-weight solutes, which is how potable water is produced from a saline source; the same barrier will pre-concentrate solutions of macromolecular organics. Because the process is mechanically simple and its specific energy consumption is low, it expanded quickly over two decades into seawater and brackish water desalination, boiler feedwater softening, wastewater reclamation and, combined with ion exchange, ultrapure water production. Dairy and fruit juice concentration, and the separation of biochemicals, are established extensions of the same principle.
It is worth being precise about the boundary. An RO membrane does not remove dissolved gases to any useful degree, it does not reliably reject small neutral molecules such as some organics, and it does not disinfect. Anything that passes is a design problem for the next unit, not a membrane fault.
Three models, and why the argument lasted a century
How solvent crosses a dense polymer film has been explained three ways, and each explanation still shapes how operators think about fouling.
The first is the preferential adsorption capillary pore model proposed by Sourirajan: water is preferentially adsorbed at the surface and migrates through pore-like regions of the material. Dry-state electron micrographs of such membranes show no pores at all, and that absence was used as an argument against the model. It is a weak argument, because a wet membrane specimen cannot be examined in an electron microscope without preparation that collapses the very structure in question. The absence of visible pores proves nothing about the wet state.
The second is the solution-diffusion model, which denies pores outright. Solvent dissolves into the membrane material, diffuses across the film under a chemical potential gradient and desorbs on the far side. No convective channel exists, so selectivity is a solubility and diffusivity property rather than a sieving property.
The third is the dry-closed, wet-open model put forward by Deng Yu and co-workers in the 1980s and 1990s, and it reconciles the first two. A dry membrane contracts and densifies, its pore structure closes, and nothing is resolved under the microscope. A wet membrane swells, solvent fills and opens the pore network, and transport becomes far faster than pure diffusion would allow. Combining the two states gives the modern description, and it also explains why a membrane that has dried out after service rarely recovers its original flux.
Flux is set by net driving pressure, not by feed pressure
The solvent permeation rate, expressed as the liquid flow capacity N, follows N = Kh (delta p minus delta pi), where Kh is the hydraulic permeability coefficient, delta p is the hydrostatic pressure difference across the membrane and delta pi is the osmotic pressure difference between the two sides. Kh rises slightly with temperature, which is why a plant running in summer delivers more permeate at the same pressure and why cold feed in winter has to be compensated with either higher pressure or more area.
For a dilute solution the osmotic pressure is given by pi = iCRT, where i is the number of ions produced by dissociation of the solute, C is the molar concentration, R is the molar gas constant and T is the absolute temperature. Two consequences matter operationally. First, osmotic pressure is proportional to molar concentration, so a divalent salt contributes more per unit mass than sodium chloride and a concentrated brine demands far more pressure than the same water at the inlet. Second, temperature enters twice: directly through T and indirectly through Kh, so a temperature excursion shifts both the driving force and the permeability.
Identify the foulant before buying a cleaning chemical
Cleaning only works if the deposit has been identified. Before any chemical is selected, the membrane condition should be assessed and the contaminant classified, so that the cleaning agent matches the actual fouling and scaling rather than the last plant's experience.
Six lines of evidence are worth collecting: operating data from the system, the potential fouling and scaling constituents of the feedwater, the deposit collected on the membrane filter of the SDI instrument, the deposit on the cartridge filter elements, the internal surface of the pipework and both ends of the elements, and, where necessary, a cut-open element analysed directly. When fouling is severe, dye testing, microscopic examination, Fourier-transform infrared spectroscopy (FTIR), scanning electron microscopy (SEM) and energy-dispersive X-ray analysis (EDX) will locate both the cause and the position of the fault.
In most plants the simpler route is faster. Once fouling is confirmed and chemical cleaning is justified, open the end plate of the pressure vessel and look at how material has accumulated in the gap between end plate and element. That single observation usually identifies the foulant and therefore the cleaning programme. Where pretreatment media such as quartz sand or activated carbon have escaped, or where colloidal, organic or biological fouling is present, the front end is worst affected and the evidence sits at the inlet of the first element. Biological fouling smells fishy; burning a scraped sample of the slime gives the burnt-hair odour of protein, which is a reliable field test.
The cleaning sequence, and the interlocks around it
The general sequence for cleaning elements is fixed: flush the pressure vessel with clean, chlorine-free permeate for a few minutes; prepare the cleaning solution in the cleaning tank using permeate, with a formulation chosen for the membrane type; circulate for one hour or another preset period; drain and flush the tank, then refill with clean permeate for rinsing; flush the vessel again; and finally run the system with the permeate discharge valve open until the permeate is clean, foam-free and free of cleaning agent, which normally takes 15 to 30 minutes.
Two rules sit above that sequence and are broken more often than any other. The concentrate valve must be fully open before the system starts, and must be closed only slowly and gradually until design recovery is reached; starting with it closed will damage elements within seconds. And the permeate line valve must never be closed at any time, whether before start-up, during normal running, during flushing or during cleaning.
Loading and start-up follow the same logic. Do not open the sealed packaging before the element goes into the vessel, store elements cool and dry and out of direct sunlight, and confirm the system has been cleaned before elements are inserted. Lubricate O-rings and the brine seal with silicone grease or a 50% glycerine-water solution; never use oil, grease, petroleum jelly or any petroleum-based compound. As elements go in one at a time, take up the clearance at the end plate with shims, which stops movement during start-up and shutdown and reduces leakage at the interconnectors. A new system is flushed thoroughly after installation, for at least 24 hours where the permeate is destined for drinking water, and high-pressure operation is preceded by a low-pressure flush through the soft-start arrangement or a variable-frequency drive to expel air.
The operating window that keeps the warranty intact
Feedwater turbidity should stay below 1.0 NTU, or SDI15 below 5, with a feed temperature below 45 °C, and the feed must contain nothing that can damage the membrane physically or chemically. Free chlorine in the feed to the elements must never exceed 0.05 mg/L at any time; above that level the membrane suffers irreversible oxidative damage, and no cleaning will bring the flux back.
Cleaning is triggered by performance, not by calendar. After a period of normal operation the elements accumulate suspended and sparingly soluble matter, and the response point is a 10% drop in system performance under standard conditions, or visible scaling and plugging. Periodic flushing and chemical cleaning restore performance and extend service life. During short shutdowns, flush every five days and close the valves afterwards, so that neither scaling nor plugging has time to set.