JMP-based lithography line-width CD control, SPC, ANOVA, and process capability analysis using NIST wafer data.
This project uses JMP to analyze lithography line-width critical-dimension variation using the NIST LITHOGRA.DAT process-control dataset.
The goal was to identify whether CD variation was mainly associated with:
- Cassette-to-cassette shifts
- Wafer-position effects
- Site-level within-wafer non-uniformity
- Measurement-sequence behavior
Source: NIST/SEMATECH e-Handbook of Statistical Methods
Dataset: LITHOGRA.DAT
Response variable: Raw_LineWidth
Total observations: 450
Dataset structure:
- 30 cassettes
- 3 wafers per cassette
- 5 measurement sites per wafer
- Sites: Top, Left, Center, Right, and Bottom
- Distribution and summary statistics
- Boxplots by wafer site, wafer position, and cassette
- Sequence run chart
- Wafer-within-cassette summary
- ANOVA using Standard Least Squares
- Individuals and Moving Range control chart
- X-bar and S control chart by cassette
- Cp/Cpk and Pp/Ppk process capability analysis
-
Mean line width: 2.532 µm
-
Cassette and wafer site were statistically significant: p < 0.0001
-
Wafer position was not statistically significant: p = 0.8871
-
The ANOVA model explained approximately 80.8% of the observed variation
-
Cassette-level mean CD was unstable in the X-bar chart
-
Within-cassette standard deviation was comparatively stable in the S chart
-
Under assumed specification limits of 1.5–3.5 µm:
- Cpk = 0.723
- Ppk = 0.465
- Observed nonconformance = 133,333 PPM
Cassette-level mean shifts and wafer-site non-uniformity were the dominant sources of lithography CD variation.
Wafer position within the cassette was not a significant contributor.
The process was not statistically stable at the cassette-mean level and was not capable under the assumed specification limits.
- Investigate exposure dose and focus uniformity
- Review resist coating and development uniformity
- Check for cassette-to-cassette tool or process drift
- Review metrology repeatability and measurement-location effects
- Repeat capability analysis using official product specification limits
The NIST dataset does not provide official LSL, USL, or target CD values. Assumed specification limits were used only to demonstrate the JMP process capability workflow and should not be interpreted as official product requirements.