Evaluating the impact of emerging contaminants on membrane performance in ultrapure water production for semiconductor manufacturing: a PRISMA-directed meta-analysis

A single chip fabrication plant uses 4 to 10 million gallons of ultrapure water a day, water so clean that its organic carbon has to stay below one part per billion. As transistors shrink below five nanometres, even parts-per-trillion contamination can ruin a wafer. The membranes that make this water were designed before anyone worried about per- and polyfluoroalkyl substances (PFAS), pharmaceuticals, microplastics or the organics washed off the wafers themselves. The evidence on how they cope was scattered across dozens of papers.

We gathered all of it, 31 peer-reviewed studies from 2002 to 2025, and analyzed them under the PRISMA protocol (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) that medicine uses for systematic reviews. Reverse osmosis rejects more than 99 percent of long-chain PFAS and pharmaceuticals but only 93.8 to 99.2 percent of the short-chain PFAS, which is where the problem now sits. Nanofiltration ranges from 86 to 98 percent depending on the compound. Ultrafiltration handles microplastics well but loses 15 to 45 percent of its throughput when the water carries the organics used in chip polishing.

For the industry, this is a design table: which membrane holds up against which contaminant, and where the short-chain PFAS slip through. For me it was the first time I ran a formal evidence synthesis end to end, with a registered search strategy, inclusion criteria and extraction, rather than a narrative review. The subject is semiconductors, but I use the same method now on water and health questions.