MFC calibration basics for process engineers
Zero check, span verification, calibration intervals, and common failure modes that affect gas flow accuracy in CVD and etch tools.
By Semiconductor Tools Editorial Team · Last updated: 2026-06-14
Field verification of MFCs confirms that commanded sccm matches delivered mass flow within vendor tolerance. Calibration drift, wrong K-factors, and downstream leaks are among the most common root causes when film thickness or etch rate trends appear without obvious chamber faults.
Verification procedure outline
- Zero check: Gas off, line purged, valve closed—reading should be <0.5% FS or vendor spec.
- Span check: Verify at 10%, 50%, 100% FS against reference meter or tool-built verification port.
- Record: Indicated vs measured, compute % error, trend over PM cycles.
- K-factor audit: Confirm gas name in software matches physical gas and vendor K-factor table.
Worked example — Document verification with MFC Gas Flow Unit Converter
PM verification on Ar carrier MFC, FS = 5000 sccm. Setpoints tested: 100, 1000, 5000 sccm. Reference meter agrees at 1000 sccm; log mass-flow equivalents for records.
- 100 sccm Ar → converter reports 0.1784 g/min, 0.0107 kg/hr.
- 1000 sccm → 1.784 g/min, 0.107 kg/hr.
- 5000 sccm → 8.92 g/min, 0.535 kg/hr.
- If reference at 100 sccm reads 112 sccm (+12%), relative error 12%—unacceptable for a 100 sccm setpoint even if within ±1% FS (±50 sccm) spec. Action: service MFC or swap to smaller FS device for low-flow steps.
Worked example — Downstream leak vs MFC error
MFC passes span check, but chamber pressure higher than recipe. Total MFC flow 200 sccm, chamber reads 18 mTorr vs target 15 mTorr.
- Pumping Speed: at 200 sccm and 18 mTorr, implied S ≈ 141 L/s. At 15 mTorr target, need S ≈ 169 L/s for same Q—or Q is higher than 200 sccm equivalent.
- Throughput Converter: extra 3 mTorr at S=141 L/s ≈ extra 0.42 Torr·L/s unexplained load—check downstream leak between MFC and chamber, not MFC calibration first.