Osmosis, reversed
Put fresh and salty water either side of a semi-permeable membrane and nature moves fresh water toward the salt. That is osmosis, and the strength of the pull is the osmotic pressure, roughly 0.7 bar for every 1 000 mg/L of dissolved salts. Reverse osmosis overpowers it: apply more pressure than the osmotic pressure and water is forced the other way, through a membrane whose polyamide surface passes water molecules while rejecting dissolved ions. Only the excess pressure above osmotic drives production. That is why feed TDS sets the energy bill, and why seawater (≈ 25 bar osmotic) needs 55+ bar while brackish water runs at a fraction of that.
Crossflow: one stream in, two streams out
An RO element is not a dead-end filter. Feed water flows across the membrane surface under pressure; a fraction permeates through as permeate (product), and the remainder sweeps the rejected salts along and exits as concentrate (reject). This crossflow keeps the membrane surface from immediately blinding with everything it rejects. The concentrate stream does the cleaning, so throttling it too far defeats the design.
The three numbers
Rejection (%): how much of the dissolved salt the membrane holds back.
Rejection = (1 − permeate TDS ÷ feed TDS) × 100
Modern brackish elements reject 99%+ of NaCl at standard test conditions; what you measure in the field is system rejection, always a little lower and drifting with temperature, pressure and age.
Recovery (%): the fraction of feed converted to permeate.
Recovery = permeate flow ÷ feed flow × 100
Recovery is a design choice, and it prices everything: at 50% recovery the concentrate leaves at twice the feed concentration; at 75%, four times. Higher recovery wastes less water but concentrates the scaling risk and raises the osmotic pressure the pump must beat at the tail end. Small systems typically run 30 – 50%; well-pretreated industrial trains 70 – 80%.
Flux (L/m²·h): permeate rate per membrane area. Designers hold flux to the membrane maker's guideline for the water type (roughly 20 – 30 LMH on clean brackish feeds, less on dirty ones) because excessive flux drags foulants onto the surface faster than crossflow can sweep them away.
What follows from the physics
- Temperature moves everything. Water permeates ~3% faster per °C, but salt passes faster too. A plant makes more, slightly saltier, permeate in summer; that is physics, not a fault.
- Pressure is a consequence, not a setting. The pump provides whatever the design flux and recovery demand against the feed's osmotic pressure; rising required pressure at constant output is the classic fouling/scaling signature.
- RO removes almost everything dissolved: salts, hardness, most organics, and (as a barrier) microorganisms. Dissolved gases like CO₂ pass straight through, which is why permeate pH drops and downstream polishing exists.
- There is always a concentrate stream. Where it lawfully goes is a design question to answer on day one, not at commissioning.
Rule of thumb: subtract osmotic pressure from feed pressure and you have the pressure making water; choose recovery and you have chosen the concentrate strength. Sizing, fouling and energy questions all come back to those two calculations.