Atomic layer deposition (ALD) builds films a fraction of a monolayer at a time by alternating precursor and reactant exposures. The chemistry assumes each pulse delivers a controlled, saturating dose. That assumption fails when the precursor leaves the ampoule inconsistently—because the liquid is too cold, the headspace pressure is unstable, or a section of delivery tubing falls below the condensation point.
CMP pads, sputter targets, and plasma ash endpoints are different consumable and process stories. This piece is about precursor delivery hardware for ALD (and closely related pulsed CVD) as part of the process window.
What the tool is asking the gas box to do
Many liquid precursors are held in temperature-controlled ampoules or bubblers. Vapor is drawn or carried into the manifold through heated lines so it remains vapor until it reaches the chamber injectors. If ampoule temperature drifts, vapor pressure changes and pulse size changes with it. If a heated line has a cold fitting or a failed trace zone, precursor can condense, then re-evaporate in bursts—producing dose spikes that look like random thickness or composition noise.
Carrier-gas flow, valve timing, and purge efficiency still matter. They do not replace thermal control of the liquid and the path.
Maintenance and MoC habits that protect the dose
- Treat ampoule temperature controllers and line heat-trace zones as process-critical instruments: calibrate, alarm, and trend them with the same seriousness as chamber pressure.
- After any gas-box rebuild, verify continuity of heat tracing and insulation on every joint in the precursor path—not only on the ampoule body.
- Change precursor lots and ampoule fill levels under MoC when the delivery method is sensitive to liquid level or surface area.
- Correlate thickness or compositional SPC shifts with gas-box temperature and flow trends before rewriting the ALD recipe table.
OEM manuals differ by chemistry and hardware generation; the plant window still belongs to that tool family. Where delivery temperature logs are missing, thickness drift investigations start with incomplete evidence.
Film growth is chemical. The dose starts in the ampoule and the heated line, not on the recipe summary screen.