An RO membrane is a synthetic semi-permeable film with a 0.1–0.3 µm dense skin on a 100–300 µm porous support, rejecting dissolved salts, colloids, microbes, and organics when feed pressure exceeds osmotic pressure [S1][S5].
Upstream feedstocks: polyamide, polysulfone, and additives
Upstream material flow starts with petrochem-derived monomers: m-phenylenediamine (MPD) and trimesoyl chloride (TMC) for TFC polyamide skins, plus polysulfone (PSf) and polyethersulfone (PES) for the porous support, with the MPD-TMC interfacial polymerization step performed on the cast PSf/PES substrate [S1][S3].
RO membrane element operating envelopes typically span feed TDS 500–45,000 mg/L, applied pressure 10–80 bar (seawater at the upper end), and recovery 35–85% depending on feed salinity and element geometry [S1]. The dense skin rejecting layer performs the actual ion separation, while the 100–300 µm porous sublayer provides mechanical strength against the 10–80 bar feed differential [S5].
A 2022-01 published review confirms that membrane parameter selection (permeability A, salt passage B, structural parameter S) determines whether an entire desalination plant succeeds or fails, reinforcing how tight the upstream monomer and substrate tolerances must be [S1]. For an engineering view of the broader water-treatment membrane supply picture, see the 2026 water-treatment membrane supply risk field map.
Downstream plant categories: seawater, brackish, process, and reuse
Four downstream segments buy RO elements at scale: seawater desalination (SWRO, 800–1200 psi operating), brackish water (BWRO, 150–400 psi), ultrapure process water for pharma/semiconductor, and wastewater reuse for industrial parks, with the largest installed base still in municipal and industrial process water [S1][S3].
Pharmaceutical and food-process users impose additional downstream constraints: FDA 21 CFR indirect food contact listings, USP purified water quality, and 3-log virus reduction per membrane validation, none of which the upstream polyamide chemistry can satisfy without formulation and post-treatment controls [S1].
Selection criteria: feedwater, recovery, and energy
Engineers selecting an RO element weigh feed TDS, target recovery, specific energy consumption (kWh/m³), and chlorine tolerance, since TFC polyamide skins are limited to <0.1 mg/L free chlorine continuous exposure and require activated-carbon or bisulfite dechlorination upstream of the membrane [S1].
RO process trains also sit downstream of the membranes covered in the industrial valve and flow meter product families, where high-pressure feed valves (typically 1.0–1.5× pump dead-head) and feed/concentrate flow meters (often magnetic or Coriolis on the concentrate) are co-specified against the same operating envelope as the elements themselves.
The four main RO process options compared on three decision criteria: SWRO (recovery 35–45%, energy 2.5–4.0 kWh/m³, capex high); BWRO (recovery 70–85%, energy 0.4–0.8 kWh/m³, capex medium); high-recovery closed-circuit RO (recovery 90–95%, energy 0.6–1.2 kWh/m³, capex high); and nanofiltration-as-pretreatment hybrids (recovery 80–90%, lower NF rejection of monovalent ions) [S1].
Who it is for, who it is not
RO is the right call where dissolved solids must drop below 500 mg/L, where pathogen log-removal must exceed 4-log, or where a single membrane step must replace multi-stage ion exchange; it is the wrong call on high-temperature feed (>45 °C, where TFC polyamide hydrolyses) or where heavy metals drive scaling that no antiscalant can suppress [S1][S3].
Users who should not default to RO: small-flow sites below 5 m³/d (capex per m³ amortises poorly), feedwater with high oil/grease load (fouls the dense skin irreversibly), and applications where the lighting-equipment-and-electric-lamps cooling loop or boiler makeup demand ultra-low conductivity under 0.1 µS/cm and where mixed-bed polishers downstream are already budgeted (so RO becomes redundant).
Pretreatment and post-treatment, the real failure gates
Three failure modes kill more RO plants than membrane chemistry: biofouling (controlled by 0.5–1.0 mg/L DBNPA or 0.2–0.5 mg/L non-oxidising biocide shock), colloidal fouling (controlled by 5 µm cartridge or UF prefilter, SDI₁₅ below 3), and scaling by CaCO₃, BaSO₄, or SiO₂ (controlled by antiscalant dosing and Langelier Saturation Index <0) [S1].
Post-treatment stage, often missed in spec sheets, is dominated by calcite contactors for hardness/stability and by degassing membranes or forced-draft decarbonators for CO₂ stripping after the high-pressure array, with energy-recovery devices (ERDs, isobaric or turbine-type, recovering 0.7–1.2 kWh/m³) sitting between the concentrate throttle valve and the booster pump [S1].
Limits, constraints, and what spec sheets quietly admit
Pretreatment quality, not membrane chemistry, sets the cleaning interval: 6–12 months for SDI₁₅ <3 and <0.1 ppm Fe, but 30–60 days for SDI₁₅ above 5, which is why procurement teams now co-source cartridge filters, antiscalant, and CIP chemicals as a single upstream/downstream bundle rather than as three separate line items [S1][S2].
Sourcing signals and 2026 trackable nodes
Two trackable signals for Q3-Q4 2026: (a) MPD/TMC spot pricing in eastern China, where one Hangzhou-based RO specialty chemical supplier is currently listed for the upstream membrane-agent chemistry, and (b) availability of 8040 seawater elements with 99.8% nominal rejection at 55 bar feed, since seawater element supply has historically been tighter than brackish [S2][S4].
Specifiers comparing supply-chain risk in 2026 should cross-read this upstream/downstream view against the 2026 RO/NF/electroactive membrane stock and sourcing signals article and the 2026 polyamide/polyester alternatives note, since both piece the same monomer-substitute and finished-element picture from a different angle.