PVD Process Monitoring & Virtual Metrology

Context and questions

Infineon’s public release contains anonymized physical vapor deposition (PVD) process inputs and 17 target measurements for each of two metal-layer processes, AlCu and WTi. I used AlCu as the primary analysis and fitted the workflow separately on WTi.

I asked two related questions: which recorded runs deserve statistical review, and how well can the released targets be predicted from process inputs?

The release omits physical units, inverse scaling, timestamps, equipment identifiers, specifications, and fault labels. Those omissions limit how I can interpret the results.

My contribution

I built the data audit, preprocessing, PCA monitoring, excursion analysis, and virtual-metrology workflows in Python. I compared baseline models with a Random Forest, repeated the analysis on WTi, and documented sensitivity checks and interpretation limits.

Data checks and decisions

I checked the four CSV files against the published checksums, then audited shapes, missing and nonfinite values, duplicates, and X/Y row counts. No join key is provided, so input/target pairing is positional.

Each process has one all-zero target record whose meaning is unknown. I retained these records in input-based monitoring, excluded them from primary target-based modeling, and tested their influence separately. Training-only feature screening removed two near-constant AlCu channels.

Two analysis paths

Process-state monitoring

I fitted standardization and principal component analysis (PCA) on training rows only, retaining the fewest components needed to explain at least 90% of input variance: 34 for AlCu and 35 for WTi.

Hotelling T² measures distance within the retained PCA space; Q/SPE measures variation the model does not reconstruct. I used training-set 99th-percentile thresholds to flag statistical excursions and contribution analysis to identify anonymous channels for review.

17-target virtual metrology

I predicted all 17 released targets directly. Linear Regression and Ridge provided baselines. A fixed Random Forest was retained after a predeclared training-only improvement check. Its configuration transferred from the primary AlCu analysis to WTi, where the model was fitted again using WTi training data.

Held-out evaluation and profile grouping

The primary evaluation used a fixed 80/20 random split. Scaling, PCA fitting, model selection, and tuning used training data rather than held-out outcomes.

Some rows have exactly equal 17-target profiles. I repeated the evaluation with each identical-profile group kept entirely on one side of the split. This tests sensitivity to profile repetition without assuming those rows are duplicate wafers. The grouped AlCu analysis used its own training-only model gate.

Grouping increased AlCu RMSE by 6.4%, indicating some optimism in its ordinary-split result. WTi improved under grouping. Neither split establishes performance on future production runs: the release has no chronology for a time-based test.

Findings and engineering limits

The Random Forest retained predictive signal under both split designs. Monitoring flagged runs and ranked channels for review. However, comparisons between flagged and unflagged runs showed no clear shift in the released output summaries: the clustered 95% intervals included zero.

• Statistical alerts require interpretation. The empirical thresholds are review aids, not validated fault labels or production control limits. Anonymous channels do not establish physical root cause.
• Released-scale errors are not physical metrology. Without units and inverse scaling, they cannot be reported as film-thickness tolerances or used to rank AlCu against WTi physically.
• Missing context limits qualification. There is no known-good period for chronological SPC, no specifications for Cp/Cpk, and no fault labels for validated fault detection. WTi is a second public dataset, not an external fab qualification.

Artifacts and sources

The repository contains the numbered analysis scripts and reproduction instructions. The technical report and results guide include figures, per-target errors, and sensitivity tables.

Dataset: Amina Mević, Andreas Laber, and Senka Krivić, Advanced Process Control and Statistical Process Control Data for Thickness Prediction of AlCu and WTi Metal Layer in Semiconductor Manufacturing, licensed under CC BY 4.0. This page summarizes my public v1.0 project documentation, reviewed September 9, 2026.