What Reverse Osmosis and Deionisation Are For

Why this matters

Reverse osmosis and deionisation are the only common treatments that take dissolved solids out of water rather than trading one ion for another or filtering out what is undissolved. That makes them the right answer to a small set of problems and an expensive, sometimes damaging answer to everything else. The two field errors are symmetrical. A shop fits a softener where the customer needed dissolved solids actually gone, and the spotting or the carryover never stops. Or a shop fits purified water into a system built for tap water, and finds out that water this clean is chemically hungry and will take metal, mortar and elastomer to feed itself.

Before opening any membrane housing or resin vessel, isolate it, relieve the pressure through its vent to a drain receptor, and confirm the pressure is at zero on a gauge rather than assuming the pump stopped. A membrane housing is a pressure vessel with stored energy in it, and an end cap released under residual pressure travels. Where a system is regenerated on site with acid and caustic, take glove class, eye and face protection and any respiratory control from the safety data sheet for each specific product, never add one product to a vessel containing the other, and never combine an acid with a chlorine-bearing product, because that combination generates chlorine gas, which is an inhalation hazard that gloves and goggles do nothing about. Most small shops should be exchanging tanks rather than regenerating on a truck for exactly that reason.

Two machines, two mechanisms

Reverse osmosis pushes water through a semi-permeable membrane with pressure, against the osmotic pressure the dissolved salts create. Water passes; most of the dissolved ions do not. The output splits in two: permeate, which is the purified stream, and concentrate, which carries away everything that was rejected. It is a bulk separation, and it is the cheaper way to remove most of a large dissolved load.

Deionisation is ion exchange without the sodium trade. A strong acid cation resin in the hydrogen form swaps every cation for H+; a strong base anion resin in the hydroxide form swaps every anion for OH-; the H+ and OH- combine into water. Nothing is added, so the output can approach pure water. It is a polishing technology, and the resin is consumed in proportion to the ions it removes, which is why running raw water straight into deionisation is the expensive mistake and why RO ahead of DI is the standard train.

The practical division of labour: RO does the heavy lifting on a large dissolved load; DI takes the small remainder to a level RO cannot reach.

What sets the limit on an RO, and it is not the permeate

The number that decides how an RO behaves is recovery, the fraction of feed that leaves as permeate. Everything rejected leaves in what remains, so the concentration factor on the concentrate side is 1 divided by (1 minus recovery), assuming essentially complete rejection, which is close enough for a first pass and is why a marginal case gets checked against the membrane maker's projection software. At 50 percent recovery the concentrate runs about twice the feed concentration; at 75 percent, about four times.

That is where scaling happens. The membrane is not fouled by the feed water, it is fouled by the concentrate at the tail of the pressure vessel, so hardness, alkalinity and silica limits apply to the concentrated stream, not to the report the customer handed you. Take a feed at 308 mg/L hardness as calcium carbonate and 22 mg/L silica. At 50 percent recovery the concentrate sees roughly 616 mg/L hardness and 44 mg/L silica. Push to 75 percent recovery to cut the drain volume, and the concentrate sees roughly 1,232 mg/L hardness and 88 mg/L silica. The membrane maker's own limits, applied at the concentrate concentration and at the operating pH and temperature, decide whether that needs softening ahead of the unit, an antiscalant, or neither. Silica limits in particular move with temperature and pH, so a figure quoted for one condition does not transfer to another.

The direction runs both ways and both halves are worth stating. Raising recovery cuts the volume going to drain and raises the scaling and fouling risk at the tail. Lowering recovery protects the membrane and sends more water down the drain, at 50 percent recovery roughly one gallon to drain for every gallon produced.

Three other conditions govern an RO in service:

Free chlorine destroys the common thin-film composite polyamide membrane, cumulatively, measured in exposure over time rather than in a single event. Where the supply is chlorinated, dechlorination ahead of the membrane with carbon or a bisulfite feed is not an option, it is part of the machine.

Output falls as the feed gets colder. Flux through the membrane drops as the water gets colder, on the order of 3 percent per degree Celsius as a linear approximation over the normal operating range; the membrane maker's temperature correction factor is the value to use. A unit sized on summer well water and installed to run on winter well water can miss its rated output by a wide margin with nothing wrong with it.

Pressure and flux set fouling as much as chemistry does. Running an element above its rated flux concentrates solutes at the membrane face faster than they diffuse away, which is a fouling mechanism independent of what the bulk water reads.

What deionisation adds, and what it costs

RO permeate from a supply at 700 microsiemens per centimetre typically lands somewhere in the region of a few percent of the feed conductivity, which is tens of microsiemens rather than hundreds. For most equipment protection duties that is finished water. For the duties that need better, DI takes it down toward the theoretical ceiling of pure water at 18.2 megohm-cm resistivity at 25 C, which is 0.055 microsiemens per centimetre.

Two things ride along. Carbon dioxide passes an RO membrane largely unrejected, arrives in the permeate as dissolved gas, and forms carbonic acid, which is why RO permeate commonly reads acidic and why a DI bed downstream sees a carbonate load it did not expect from the conductivity number alone. And DI capacity is consumed in proportion to ions removed, so every mg/L the RO fails to remove is bed life you buy back at the polishing stage. A poorly performing RO shows up as a DI bed that exhausts early, not as a purity failure, because the polisher covers for it until it cannot.

The requirement sheet, filled in

Purity is specified backwards, from the failure you are preventing. This is the sheet worth filling in before anyone quotes a machine, completed here for one real case: a shop wanting spot-free final rinse water for a wash bay, fed from the municipal supply described above.

The sheet does three things a quote does not. It forces the target to be a measured number rather than a promise, it exposes the drain volume before installation instead of after, and it makes the downstream materials a decision rather than whatever was on the truck.

Two entries deserve their reasoning spelled out.

Storage is normally the real answer to peak rate. RO output is a slow, steady flow; a wand is a short, high-rate demand. Sizing a membrane array for the peak rate is the expensive way to solve a problem that a tank and a delivery pump solve. Where the polisher is DI, the tank sits ahead of it or after it depending on how much the water picks up while stored, and stored purified water does pick things up. Two kinds of pickup, and only one is chemical: purified water carries no disinfectant residual and no competing load, so a warm stored tank grows organisms readily and a high-pressure wand is an aerosol generator fed from that tank. Size and site it so it turns over, keep it closed and cool, and where the rinse is delivered as a spray that tank belongs in the building's water management programme.

The air gap on the concentrate is not paperwork. A drain line submerged in a receptor is a cross-connection, and a cross-connection between a drain and a treated water system is a public health event rather than a plumbing detail. The adopted plumbing code in your jurisdiction sets both the air gap requirement at the drain and the backflow assembly class at the supply.

What you have made is hungry water

Water with almost nothing dissolved in it has almost no buffering and a strong tendency to dissolve whatever it touches until it has something in it again. Purified water attacks copper and galvanized steel, leaches from concrete and mortar, and carries dissolved carbon dioxide that makes it mildly acidic. It also runs through elastomers and gasket materials that were selected against tap water without complaint until they are not.

So the material decision is part of the treatment decision, in the same breath rather than as a follow-up. Plastic piping, appropriate stainless where mechanical strength is needed, and no galvanized anywhere on the permeate side. Where purified water must feed a system built for tap water, remineralisation or blending is the correction, and it is a design step with a target rather than a splash of untreated water at the tank.

This is also the answer to the most common misuse of these machines. Purified water is not a general upgrade to a system's water. Feeding RO permeate into a closed hydronic loop, for instance, buys nothing that a clean fill and a proper inhibitor do not already buy, and it hands you an aggressive fill in a system full of mixed metals. Match the purity to the duty.

Conductivity is the instrument

Because these processes remove conductive species, conductivity is both the product specification and the diagnostic. A permeate conductivity trending up at constant feed conductivity, recovery and temperature is a membrane losing rejection or a seal passing feed into permeate, and the trend appears long before anything visible does. A DI bed's exhaustion is a resistivity curve that holds flat and then falls away sharply, which is why an inline meter with an alarm beats a calendar.

Record feed conductivity alongside permeate conductivity every time, because the ratio is the real measurement and a permeate number alone moves whenever the supply moves. Normalise for temperature, or at least record it, since conductivity itself reads higher in warmer water for the same dissolved load.

References

• Membrane manufacturer projection data and element limits for recovery, flux, feed temperature correction, chlorine tolerance, and concentrate-side scaling limits at operating pH and temperature
• NSF/ANSI 58 for certified point-of-use reverse osmosis systems where a drinking water claim is made
• OSHA 29 CFR 1910.1200, hazard communication, for safety data sheet availability and the product-specific protection required for acid and caustic regenerants
• The adopted plumbing code in your jurisdiction for backflow protection at the supply connection and the air gap requirement at a treatment drain termination
• See related: What a Softener Removes and What It Leaves Behind; The Water Chemistry That Attacks a System