There are no holes in a reverse osmosis membrane. It is not a filter, and that explains almost everything about it

Almost every reverse osmosis purifier in an Indian home throws away more water than it delivers, strips out calcium and magnesium along with everything else, and is sold under a word that does not describe what it does.

It is not a filter. A filter has holes, and things too big to pass through them stay behind. The working layer of an RO membrane has no permanent holes at all. It is a dense, continuous film of polymer, and water gets through it by dissolving into the plastic and diffusing out the other side.

That single structural fact explains the reject water, the electricity bill, why chlorine destroys the membrane, why every few years a "new membrane material will solve water scarcity" headline appears and nothing much changes, and whether you should have bought the unit at all.

First, what osmosis actually is

Put fresh water on one side of a membrane that passes water but not salt, and salty water on the other. Water moves spontaneously from the fresh side to the salty side — not because anything is pushing it, but because mixing is the more probable state and the salt cannot cross to do the mixing itself.

The salty side rises until the pressure of the extra column of liquid stops the flow. That pressure is the osmotic pressure of the solution, and it is a real, measurable, and surprisingly large quantity. For seawater it is around 27 bar — roughly twenty-seven times atmospheric pressure, or the pressure at the bottom of a water column nearly 280 metres deep.

Reverse osmosis is exactly what the name says. Apply pressure to the salty side that exceeds its osmotic pressure, and you push water back the way it did not want to go, leaving the salt behind. Real seawater plants run at somewhere around 55 to 70 bar to get useful flow; brackish groundwater, being less salty, needs perhaps 10 to 20.

So the first honest statement about desalination is that its energy cost is not an engineering failure. It is a consequence of a thermodynamic quantity that belongs to seawater itself.

Solution-diffusion: separation without holes

Now the part that is routinely got wrong, including in passing on this site's own article about municipal water treatment, where RO sits in a list of processes described as filtering by pore size. It belongs in that list for its effect and not for its mechanism, and the distinction matters.

A modern RO membrane is a thin-film composite. Its top layer — the part that does the separating — is an aromatic polyamide roughly a hundred to two hundred nanometres thick, formed in place by a reaction between two monomers meeting at the interface between a water phase and an organic phase. Beneath it sits a porous polysulfone support, and beneath that a fabric backing for strength. Everything below the top hundred nanometres is scaffolding.

That polyamide layer is dense. It has free volume between polymer chains — transient gaps that open and close as the chains move — but no fixed channels. Separation happens by solution-diffusion: water molecules dissolve into the polymer at the high-pressure face, diffuse through it down a gradient, and emerge on the far side. Salt ions do the same thing, far less readily, because they are charged, strongly bound to their hydration shells and poorly accommodated by the polymer.

Selectivity therefore comes from how well each species dissolves and diffuses, not from whether it fits through an opening. This is why RO rejects sodium and chloride ions — which are a fraction of a nanometre across, smaller than any pore you could sensibly build — at better than 99%. A sieve with holes fine enough to block those ions would block water too.

Filtration asks "is it bigger than the hole?" Reverse osmosis asks "does it dissolve in this plastic and move through it?" The second question has a completely different set of answers, which is why RO removes things no filter could and lets through a few small neutral molecules that a naive pore model says it should have stopped.

That last point is a real operational consequence: small uncharged molecules such as boron, and some low-molecular-weight organics, are rejected far less well than salt, because charge is doing much of the work and they do not have any.

Why better membranes cannot make water much cheaper

Here is the most important and least understood fact in this field.

Separating pure water from seawater has a thermodynamic minimum work — an amount of energy below which no process of any kind can go, however cleverly designed. For seawater at typical recovery ratios, that minimum is somewhere around 1 kilowatt-hour per cubic metre of product water.

A modern large seawater RO plant uses roughly 3 to 4 kilowatt-hours per cubic metre. So the industry is already operating within a small factor of a hard physical limit, and most of the remaining gap is not in the membrane at all — it is in pump efficiency, piping losses and pre-treatment.

Which leads to the conclusion that ought to be applied to every membrane headline. A more permeable membrane lets you push the same water through less membrane area, so it reduces the size and capital cost of the plant. It does not reduce the pressure you must apply, because that is set by osmotic pressure, and it therefore does not meaningfully reduce the energy bill. Permeability was not the binding constraint.

The gains that genuinely transformed desalination came from somewhere else entirely: energy recovery devices. The brine leaving an RO train is still at high pressure, and a pressure exchanger transfers that pressure directly to incoming feed water instead of throwing it away. This is the main reason seawater RO fell from around 8 kWh/m³ a few decades ago to today's 3 to 4. The membranes improved too, but the plumbing is what moved the number.

The two things that actually run the plant

Concentration polarisation is the first. Water crosses the membrane; salt does not. So salt accumulates in a thin layer right at the membrane surface, where its concentration — and therefore its osmotic pressure — is higher than in the bulk feed. The membrane experiences worse water than the tank contains. This is why RO is operated in crossflow, with feed sweeping tangentially across the surface rather than being pushed straight at it, and why flow velocity is a design parameter rather than an afterthought.

Fouling is the second, and it is the real operating cost. Four kinds arrive together: mineral scaling as sparingly soluble salts exceed their solubility in the concentrated brine; colloidal fouling from fine particles; organic fouling from natural organic matter; and biofouling, a living biofilm growing on the membrane.

Biofouling is the hardest because it is alive and regrows from survivors. It is also where a genuinely counterintuitive operational rule comes from: polyamide membranes are destroyed by chlorine, so feed water must be dechlorinated before it reaches the membrane — and then the product water is often chlorinated again afterwards for distribution. A plant spends effort removing the very disinfectant that would have controlled the biofilm, and then manages the biofilm by other means. Most of the equipment in a large RO plant is pre-treatment, and most of the operating trouble is fouling.

Whether you should have bought one

For an Indian household the calculation is more specific than the marketing suggests.

RO wastes water by design. A domestic unit typically discards one to three litres for every litre it delivers, because small membranes and low-pressure pumps mean low recovery — you cannot concentrate the reject too far without driving up pressure and scaling. Large plants achieve better ratios with staging and much higher pressures. The reject is not sewage; it is simply saltier feed water, and it can be used for floor washing, flushing or non-sensitive irrigation rather than sent down the drain.

It removes what you might have wanted. Calcium and magnesium go with the sodium, which is why many units add a remineralisation cartridge and why long-term consumption of fully demineralised water has been discussed in public-health guidance. In practice, diet supplies the great majority of dietary minerals, so this is a reasonable thing to address and not a crisis.

And often it is the wrong appliance. The most common mistake is buying RO on a high TDS reading. Total dissolved solids measures dissolved ions in aggregate; it says nothing about whether any of them are harmful. Water at 500 ppm of calcium bicarbonate is hard and harmless; water at 200 ppm containing arsenic is dangerous. If the actual problem is microbiological, ultraviolet treatment or ultrafiltration handles it without the energy, the waste or the demineralisation — and as our article on fluoride and arsenic argued, the useful first step is a laboratory test of the specific water, not an assumption about what a number means.

Where nanomaterials come in, honestly

Membrane research is full of nanomaterials and it is worth being precise about what they can and cannot change.

Graphene oxide laminates, carbon nanotube membranes and aquaporin-mimetic channels have all demonstrated remarkable water permeability in the laboratory — in some cases orders of magnitude above polyamide. The physics is real: water moving through an atomically smooth channel encounters much less resistance than water diffusing through a polymer.

Three things temper it. First, as established above, permeability is not the binding constraint — energy is set by osmotic pressure, so a hundredfold more permeable membrane does not give hundredfold cheaper water. Second, the requirement is not a good membrane but a good membrane over thousands of square metres with essentially no defects, because a single pinhole leaks unrejected salt and one defect can dominate a module's performance — the same "worst flaw" logic that governs brittle strength and shielding apertures. Third, polyamide membranes have decades of accumulated knowledge about fouling behaviour, chlorine tolerance, cleaning chemistry and lifetime, and a new material must re-earn all of it.

What nanomaterials are genuinely contributing today is more modest and more real: anti-fouling surface modifications, thin-film nanocomposite membranes with additives that improve flux or chlorine resistance, and better pre-treatment. Those attack the operating cost, which is where the money actually goes.

Why it matters for students and researchers

Reverse osmosis is a superb teaching case for the discipline of asking which constraint actually binds. A great deal of membrane research optimises permeability because permeability is measurable and publishable, while the field's economics are governed by osmotic pressure, energy recovery and fouling. A student who learns to ask "if this were a thousand times better, what would it change?" will assess claims in any field more usefully.

It also illustrates the permeability–selectivity trade-off that recurs across separations: materials that let water through faster usually let salt through faster too, and the interesting work is in moving the whole trade-off curve rather than sliding along it.

The open problems are practical and pressing. Fouling-resistant surfaces that survive cleaning cycles would change plant economics more than any flux improvement. Boron and small neutral organics remain poorly rejected and are increasingly regulated. Brine disposal is a genuine environmental problem at scale, and concentrate management — including approaches that recover salts as products — is an active area. And low-pressure, low-recovery household systems are under-engineered relative to how many of them India installs every year.

Frequently asked questions

Does an RO purifier really waste water, and can I do anything about it?

Yes, typically one to three litres rejected per litre produced in a domestic unit, because low pressure and a small membrane limit how far the feed can be concentrated. The reject is not dirty water — it is just saltier than what went in — so the practical answer is to plumb it into a storage container and use it for washing floors, flushing or watering non-sensitive plants. Units with higher recovery exist and generally use a booster pump.

Does RO remove essential minerals, and is that dangerous?

It removes calcium and magnesium along with the salts it is there to reject, and public-health guidance has considered the long-term consumption of fully demineralised water enough for many units to include a remineralisation stage. The honest framing is that diet supplies the overwhelming majority of these minerals, so this is worth addressing sensibly rather than treating as a hazard.

My water has high TDS. Do I need RO?

Not necessarily, because TDS is a measure of how much is dissolved and not of whether any of it is harmful. Hard water with a high calcium content reads high and is not a health risk; water with a modest reading can contain arsenic or fluoride and be dangerous. Test for the specific contaminants that are common in your area and choose treatment for those, since a microbiological problem is better handled by UV or ultrafiltration than by RO.

Why does my RO membrane need replacing, when nothing is touching it?

Because it fouls and degrades. Scale and biofilm accumulate on the surface and reduce flux; cleaning recovers some of it and not all. Oxidants, particularly chlorine that slipped past the pre-filter, chemically damage the polyamide layer and reduce salt rejection permanently. A rising reject ratio, falling output or worsening taste are the usual signs, and pre-filter maintenance is what determines how long the expensive membrane lasts.

Will graphene membranes make desalination cheap?

They will not, for a reason that has nothing to do with how good graphene is. The energy required is set by the osmotic pressure of seawater, which is a property of seawater, so a more permeable membrane reduces the area of membrane needed rather than the energy needed. That reduces capital cost, which is worth having. The remaining challenge is making defect-free membrane area at industrial scale, since a single pinhole undoes the selectivity of everything around it.