Report ID: ET-09
Tool: Etch chamber family C (anonymized)
Symptom: CD mean shift + edge-heavy map after “pressure tweak”
Format: Designed experiment write-up. Not a brainstorming slide.
1. Objective
Determine whether chamber pressure setpoint (±8%) explains the CD shift, or whether RF hours / season kit / ESC temperature dominate.
2. Factors (coded)
| Factor | − | + | Notes | | --- | --- | --- | --- | | A Pressure | −8% | +8% | Contested knob | | B RF hours bucket | Low | High | Same chamber | | C ESC temp | Nom | +3°C | Often ignored |
Full factorial 2³ on short-loop wafers; then confirmation on product-like stack.
3. Results (compressed)
Main effect A (pressure) small.
Main effect B (RF hours) large.
Interaction B×C visible on edge dies.
Pressure narrative failed. Consumable/seasoning narrative held.
A loud opinion about pressure is not a main effect.
4. Confirmation lot
High RF-hours chamber at nominal pressure reproduced the defect signature.
Low RF-hours chamber at +8% pressure did not.
Decision: freeze pressure tribal fixes; trigger kit/season playbook by RF-hour gate.
5. What we almost did wrong
- Changed pressure on all chambers “to match the golden.”
- No split lot.
- No coded factors.
- MES note: “optimized.”
That path invents tribal recipes that fight each other at the next PM.
If it is not coded, it is not an experiment—it is a mood.
6. Actions
| Action | Owner | | --- | --- | | RF-hour threshold → auto advisory | Equipment | | ESC temp logging mandatory in run cards | Process | | Pressure changes require ET-style DOE ID | Process + Yield | | Close ET-09 with maps in QMS | Yield |
Adjacent fences
Ion-implant dose walks own Faraday chains. Wet-bath notebooks own titration lifetime. High-NA stitching owns litho seams. AMC hunts own airborne chemistry. ET-09 owns causal honesty on an etch rumor. Do not retune all chambers because one map looked ugly.
Sign-off
Process ____ Equipment ____ Yield ____
Without three signatures, ET-09 is a PDF, not a control.