Cleanroom HVAC and HEPA/ULPA filtration

The intuition that a filter is a sieve — that it stops everything larger than its holes and passes everything smaller — is wrong for HEPA media, and almost every counter-intuitive property of cleanroom filtration follows from that.

Three capture mechanisms and a minimum between them

HEPA media is a mat of randomly oriented glass fibres with pore spaces much larger than the particles it removes. Capture happens when a particle contacts a fibre, by three mechanisms with opposite size dependence. Large particles cannot follow the air’s curving streamlines around a fibre and hit it by inertial impaction; effectiveness rises with size and velocity. Mid-size particles follow the streamline but pass within one particle radius of the fibre and touch it — interception. Small particles undergo Brownian motion, wander across streamlines and strike a fibre by diffusion; effectiveness rises as size falls.

Because two mechanisms improve with increasing size and one improves with decreasing size, total efficiency has a minimum where all three are weak. That most penetrating particle size is roughly 0.1–0.3 µm for typical media and face velocities, and it moves with velocity: run the filter slower and it shifts. This is why “99.97% at 0.3 µm” is not a spec point in the ordinary sense but a statement about the hardest case. The same filter is more efficient at 1 µm and more efficient at 0.05 µm. ULPA grades push the same physics further, to 99.999% at the most penetrating size. It is also why leak testing is done by scanning with an aerosol photometer or particle counter at the face: an installed system fails at the gasket, the frame or a pinhole, not through the media.

Filters remove particles. They do not remove gases or vapours — solvent, ozone, ammonia and volatile process residues pass through untouched, and control of those is a chemistry and dilution problem, not a filtration one.

Cascade and air changes are recovery arguments

A room’s classification is a particle count, but the design quantities behind it are pressure differential and air-change rate, and both are best read as statements about time. A pressure cascade fixes the direction of leakage through the inevitable gaps: air moves from the cleaner side to the dirtier one, so the cascade declares what is being protected. Reverse it and a room protects the corridor from the product instead — which is the correct choice for a containment suite and the wrong one for aseptic filling. EU GMP Annex 1 treats the cascade and its monitoring as a design requirement rather than an operating preference.

Air-change rate matters because the room is diluting a continuous source — people. Under near-perfect mixing, concentration decays exponentially with the number of air changes, so doubling the rate roughly halves the time to return to the class limit after a disturbance. ISO 14644-3 describes recovery testing directly: raise the particle concentration deliberately, then measure the time to fall back by a defined factor. That measured recovery time, not the nominal air-change number, is the honest description of a room’s ability to absorb an event.

Unidirectional airflow and cabinet inflow — the local protection over an open process — are a separate argument, covered on biosafety cabinets and laminar flow hoods and not repeated here.