Start with the event and the available data
Growth A has RHEED recordings from two measurement views and a process log covering a 60-minute run. Active deposition runs from minute 10 to minute 55. An engineer notices that the pattern looks different around minute 32. Open Analysis, select New Notebook, and enter:Investigate the change in Growth A around minute 32. Compare the RHEED behavior before and after the event with a suitable reference growth. Check whether the change persists, whether it appears in both measurement views, and whether it coincides with a recipe transition or a measurement-setting change. Show the evidence and identify what is missing.Click Create Notebook. Confirm the notebook uses the intended growth and reference. See Creating a Notebook for the full instructions.
Separate a sustained change from a brief excursion
Ask for a timeline of similarity to the reference during equivalent process phases, comparing each measurement view with the corresponding reference view. Higher similarity means the patterns resemble one another more closely under the selected comparison.
Fictional similarity traces for two measurement views. A brief excursion recovers near minute 18; the shift near minute 32 persists.
Show frames just before and after minute 32, using the same image scaling. Does the change come from the diffraction features, the background, or image brightness? Show the actual frame timestamps and any gaps.A sustained shift is worth investigating even if a detector based only on neighboring windows starts treating the new state as typical. Keep the original reference visible while examining the later part of the run.
Compare intervals with the same process meaning
For this example, compare minutes 27–31 with minutes 33–37. Both intervals are inside the same recorded deposition step. Leave out the transition itself, and inspect whether the result holds over other nearby intervals.
These observations narrow the investigation. They do not identify a cause. Stable recorded settings do not rule out an unlogged camera change, sample motion, or an incomplete process log.
Check clock alignment before interpreting the timing of another signal. If a process log’s timing uncertainty is larger than the event you are comparing, record that limit rather than claiming which signal changed first.
Challenge the leading explanation
Use the observations to ask a more discriminating question:The shift persists in both views without a recorded recipe or exposure change. Compare a process change with sample motion or a measurement artifact. What evidence supports each explanation? Which additional observation would best distinguish them?For example, a simultaneous change in diffraction-feature positions could motivate a geometry check. A coincident change in a recorded process value could motivate a process investigation. Neither coincidence alone establishes causation. If other relevant growths are available, check whether the same feature appears at the same recipe phase. Compare equivalent conditions, rather than every event at the same elapsed minute.
Record the finding and the next check
A useful handoff for this example is:Growth A shows a sustained RHEED change near minute 32 in both recorded views. The selected before/after intervals belong to the same deposition step, and the exposure log shows no change. Inspect feature positions and the instrument event log next to distinguish a process change from a geometry or measurement issue. The cause remains unresolved.Save the notebook with the reference selection, interval definitions, and missing evidence. Record confirmed findings on the sample or project; follow your team’s procedures for any operational change.
Adapt the investigation
For example, a pressure excursion during a repeated exposure step could be investigated against earlier exposures and the recorded valve events. The same sequence applies: locate the change, establish context, challenge the explanation, and choose the next check.