Semiconductor Cleanroom HVAC: Cleanroom Classes and Pressure Cascades

Why semiconductor air is unforgiving

A wafer with a 3-nanometre feature size has no tolerance for a particle measured in micrometres. One stray particle, one spike in relative humidity, or one airborne molecular contaminant (AMC) can ruin a batch that took weeks to reach that step. This is why semiconductor cleanroom HVAC is not an afterthought bolted onto a factory — it is the production system that makes the process possible.

Two ideas carry almost all of the design weight: cleanroom class, which sets the maximum particle count a zone is allowed to hold, and pressure cascade, which keeps contamination from drifting from a dirtier area into a cleaner one. Get either one wrong and the fab’s yield drops before the first tool is qualified.

Cleanroom classes: ISO 14644-1 at the fab level

Semiconductor fabs are specified against ISO 14644-1, which grades cleanrooms by the maximum number of airborne particles per cubic metre at a given size. The grades that matter in a modern fab are ISO 5 through ISO 8, with ISO 5 and ISO 6 reserved for the photolithography and etch bays and ISO 7 to ISO 8 covering metrology, support and service areas.

What the class does not tell you is how to hit it. Reaching ISO 5 typically demands unidirectional airflow at 0.3–0.5 m/s across the entire critical zone, which means fan-filter units (FFUs) or a tightly engineered laminar-flow plenum — not a conventional ducted diffuser. ISO 6 and ISO 7, by contrast, are usually achievable with high-grade recirculating AHUs plus a final HEPA or ULPA stage, provided the air-change rate is high enough.

A useful planning rule: cleanroom class and air-change rate move together. A zone targeting ISO 6 will often need 60–90 air changes per hour, while an ISO 7 zone may need only 30–60. Fan energy follows air changes, so over-specifying a class you do not need is an ongoing operating cost, not a one-time capital line.

Pressure cascades: the direction air is allowed to move

Class sets the ceiling; pressure decides the flow. In a fab, air must always move from the cleanest zone toward the less-clean zone, and never the reverse. That is what a pressure cascade is: a stepped pressure profile, typically 10–15 Pa between adjacent zones, with the photolithography bay held at the highest positive pressure and the service chase or loading dock held negative relative to everything around it.

Three details decide whether a cascade actually works:

• Supply/return balance per zone. Pressure is created by supplying slightly more air than a zone exhausts. A cascade fails the moment a return damper drifts or a door is left open.
• Air locks and gowning rooms. Every transition between classes needs a buffered space that lets personnel pass without short-circuiting the pressure step.
• Monitoring, not just setpoints. Differential pressure across every critical boundary should be live-monitored and alarmed, because a cascade that drifts silent is worse than none.

In practice the pressure ladder often runs something like this: photolithography (+45 Pa) → etch/process (+30 Pa) → clean corridor (+15 Pa) → service chase (0 Pa) → ambient. The exact numbers follow the facility, but the direction of the gradient is non-negotiable.

Filtration, temperature and the AMC problem

Particle filtration in a fab is staged, never single-stage. The typical chain is a pre-filter (G4/F5) to protect the coil, a medium filter (F8/F9) in the AHU, then a final HEPA or ULPA at the point of delivery. Each stage exists to protect the one downstream of it and to extend the life of the expensive final filter.

Temperature and humidity are held far tighter in a semiconductor fab than in a general cleanroom. Photolithography in particular is sensitive to thermal expansion: a temperature swing of half a degree across a reticle can distort the pattern. The usual targets are ±0.5 °C for temperature and ±3% for relative humidity in the most critical zones, which requires precision cooling with tight capacity control rather than a conventional chiller cycling on and off.

Increasingly the harder problem is AMC — airborne molecular contamination. Ammonia, amines and siloxanes at parts-per-billion levels can fog lenses and poison resist chemistry. Addressing AMC means chemical filters — activated-carbon or chemically impregnated media — placed ahead of the final HEPA, and, where outside air is the source, a make-up air AHU with gas-phase filtration on the fresh-air path.

Where the HVAC design effort goes

For a fab being planned now, the HVAC decisions that most affect yield and running cost are, in order of leverage:

1. Lock the cleanroom class map early. Agree which bay is ISO 5, which is ISO 6, and resist the urge to make everything ISO 5 — every extra air change is fan energy paid forever.
2. Design the cascade before the ductwork. Pressure steps, air locks and exhaust placement come first; the AHUs are sized to serve them, not the other way around.
3. Specify precision cooling for the critical zones. ±0.5 °C is not a chiller-cycling problem; it needs close-coupled capacity control.
4. Budget for AMC filtration up front. Retrofitting gas-phase filters into an operating fab is far more expensive than designing them in from the start.

The equipment list — recirculating AHUs, FFUs, dry-cooling or chilled-water precision units and make-up air handlers — is the visible part. The engineering value is in the class map and the pressure ladder that make those machines behave as one controlled environment.