Load-lock pump-down: sizing for semiconductor tools
A load lock exists so the process chamber never sees atmosphere. Wafers enter a small antechamber, that antechamber is pumped from atmospheric pressure to process vacuum, and only then does the gate to the process chamber open. Everything about the pumping set is dictated by two demands that pull in opposite directions: evacuate fast enough not to throttle throughput, and do it without adding a single particle or a trace of hydrocarbon.
This note covers how load-lock pump-down time is actually calculated, why the naive formula misleads, why dry pumps are not negotiable, and the venting detail that governs particle counts more than pump choice does.
1. Pump-down time, and where the textbook formula stops working
The standard relation for evacuating a fixed volume is:
t = (V / S) × ln(p₀ / p₁)
where V is chamber volume, S the pumping speed, p₀ the starting pressure and p₁ the target. Rearranged to size a pump for a required cycle time:
S = (V / t) × ln(p₀ / p₁)
Take a 30 litre load lock going from 1,000 mbar to 0.1 mbar in 60 seconds:
S = (0.03 / 60) × ln(1000 / 0.1) = 0.0005 × 9.21 = 4.6 × 10⁻³ m³/s ≈ 17 m³/h
That result is optimistic, and treating it as an answer is the most common sizing error in this application. Three effects push the real requirement several times higher.
The ultimate pressure floor
The logarithm implies any pressure is reachable given time. It is not. Use:
t = (V / S_eff) × ln[(p₀ − p_ult) / (p₁ − p_ult)]
As p₁ approaches the pump's ultimate pressure, t goes to infinity. A pump whose ultimate is 0.05 mbar will approach 0.1 mbar asymptotically and waste most of the cycle in the last decade. Specify a machine whose ultimate sits at least a decade below the target.
Conductance
The pump's rated speed is not what reaches the chamber:
1 / S_eff = 1 / S + 1 / C
C is the conductance of the line. In the molecular flow regime, for a pipe of diameter d and length l in centimetres, conductance in litres per second is approximately:
C ≈ 12.1 × d³ / l
The cubic term is the point. Doubling pipe diameter multiplies conductance eightfold; halving the run only doubles it. Short and fat beats long and narrow, every time. A generously sized pump behind a long narrow line is money spent on a bottleneck.
Outgassing and leakage
Below roughly 1 mbar the gas load stops being "the air that was in the box" and becomes what is coming off the walls. The base pressure a system settles at is:
p_base = (Q_leak + Q_outgas) / S_eff
Water is the dominant desorbing species on any surface that has seen room air, and desorption falls roughly as 1/t — which is why the last decade always takes disproportionately long, and why a chamber left open to the cleanroom pumps down far more slowly than one that was vented dry.
2. Why the pump must be dry
Load locks are pumped by a turbomolecular pump backed by a dry primary. The backing pump is where contamination risk lives, and it is not a preference — it is a hard requirement.
An oil-sealed rotary vane pump can backstream hydrocarbon vapour up the foreline toward the chamber, particularly at the low throughputs typical near base pressure. On a wafer surface hydrocarbon contamination causes adhesion failures, gate-oxide defects and unrepeatable etch and deposition rates. The damage is invisible at the point it occurs and shows up later as yield loss, which makes it the most expensive class of contamination there is.
Dry screw pumps compress the gas between rotors that contact neither each other nor the housing. There is no lubricant and no operating fluid anywhere in the swept volume, so the failure mode does not exist. That is the reason the semiconductor industry moved to dry primaries and has not looked back.
3. Particles: the venting detail that matters more than the pump
This is where most load-lock specifications are weakest, and where the largest, cheapest gains are found.
Particle counts inside a load lock are governed less by the pump than by how fast gas moves during pump-down and venting:
- Turbulent flow lifts particles. Rapid pumping at the start of the cycle, while pressure is still near atmospheric, stirs the chamber and puts settled particles back into the gas — where they land on wafers. A soft-start valve that limits initial pumping speed through the viscous regime measurably reduces counts.
- Venting is worse than pumping. Gas rushing in is more disruptive than gas leaving. Restrict the vent, and never vent straight into the chamber against a wafer.
- Vent with dry nitrogen, not cleanroom air. This single change gives a large reduction in particle density, and it also removes the water load that would otherwise be adsorbed onto every interior surface and slow the next cycle. It improves both the particle count and the pump-down time.
If a load lock has a particle problem, examine the valve timing and the vent gas before assuming a bigger pump is the answer.
Recommended Dr.Vakuum machines
A load lock needs two stages: a turbomolecular pump for the process-vacuum regime and a dry primary to back it and handle the roughing.
Turbomolecular pumps (T Series) — ultimate pressure below 1 × 10⁻⁷ mbar, covering the high-vacuum stage.
| Model | Suits |
|---|---|
| T600 | Small load locks and transfer chambers |
| T1200 | Standard load locks |
| T1600 | Larger chambers, higher gas loads |
| T4200 | Process chambers and high-throughput tools |
Dry screw pumps (DSI Series) — oil-free compression, 0.05 mbar ultimate, for backing and roughing.
| Model | Flow (50/60 Hz) | Suits |
|---|---|---|
| DSI150 | 150 / 180 m³/h | Backing a single turbo |
| DSI300 | 300 / 360 m³/h | Load lock roughing plus backing |
| DSI630 | 630 / 756 m³/h | Multi-chamber tools |
| DSI1200 | 1,200 / 1,440 m³/h | Cluster tools and central roughing |
Size the primary on the roughing duty — the volume and the cycle time you need — and the turbo on the process pressure and gas load. Then check the foreline conductance before ordering either, because that cubic term decides whether the pump you bought is the pump you get.
Sizing help. Send us chamber volume, target pressure, required cycle time and foreline geometry, and our engineers will work the duty point through with you. Talk to an engineer.
See also: how the turbo or cryopump and its backing pump are matched in High vacuum for coating and analytical instruments; the back-end plants that receive these wafers in Semiconductor back-end (OSAT); and the hard-disk cleanrooms next door in HDD and precision electronics cleanroom vacuum.