How to Select PE, PS and DF Process Membranes

Start with the separation objective

Define which component should be retained and which should pass through the membrane. Molecular weight is a starting point, but molecular shape, charge, salt composition, viscosity and concentration also influence separation. Establish whether the valuable product is in the retentate or permeate before selecting a grade.

PE, PS and DF cover different, partly overlapping separation duties. A nominal molecular-weight cut-off is not an absolute boundary or a guarantee of product rejection.

Compare the membrane families

PE ultrafiltration uses PES chemistry and a 34 mil feed spacer. PE1–PE4 reference cut-offs are 1,000, 3,000, 5,000 and 10,000 Da. Final PE5 performance data remains to be confirmed. PS process nanofiltration spans approximately 80–400 Da across the available grades.

DF separation nanofiltration spans approximately 300–3,000 Da: DF1 400–600, DF2 300–500, DF3 800–1,000, DF4 600–800, DF5 approximately 1,000 and DF6 2,000–3,000 Da. Around 1,000 Da, PE1, DF3 and DF5 are candidates for comparative testing. Reference flows may use different test pressures, so a direct ranking from catalogue flux is inappropriate.

Account for pretreatment and salt passage

Cell debris, suspended solids and colloids may require microfiltration or ceramic ultrafiltration before nanofiltration. Identify the salt species and the stream that must meet the salt specification. Diafiltration can be evaluated alongside concentration, with purity, product yield and water use assessed together. Further desalination may require another process stage, such as reverse osmosis.

Validate a route to scale-up

Screen two to four adjacent grades with the same actual feed and comparable operating conditions. Continue beyond initial flux to evaluate concentration, diafiltration, fouling and cleaning recovery. Record product mass balance and repeatability. Pilot design should use sustained performance and hydraulic requirements; increasing membrane area alone does not establish a reliable scale-up.