# TCAD DOE and surrogate optimization replay

- Version: 2026-08-22-v90
- Experiment: SA-DEV-TCAD-OPT-004
- Public routes: /simulations/tcad-doe-surrogate-optimization-replay, /learn/tcad-doe-surrogate-optimization-literacy, /labs/tcad-doe-surrogate-optimization-literacy
- Dataset: tcad-doe-surrogate-optimization-replay-v90.csv
- Rows: 6144
- Execution: precomputed browser CSV replay only

## Purpose

This replay teaches how TCAD-style DOE, response surfaces, Gaussian-process uncertainty, surrogate screening, extrapolation guards, and validation posture should be read before any optimization claim is reused.

## What Is Included

- Original synthetic rows across device context, design lever, DOE plan, surrogate family, objective priority, and validation posture.
- Scores for mesh stability, calibration anchor strength, sample efficiency, surrogate uncertainty, extrapolation risk, constraint violation, objective fit, and pareto readiness.
- Decision gates that route the row to mesh, calibration, extrapolation, constraint, uncertainty, owner-review, bounded-teaching, or bounded-reference states.
- Link-only source spine from public official or rights-holder pages.

## What Is Not Included

- No live TCAD, SPICE, process, electrothermal, RF, field-solver, or optimizer execution.
- No TCAD deck, mesh, structure file, material file, PDK, compact model, process recipe, solver transcript, calibration dataset, measured silicon, or vendor benchmark.
- No foundry-node ranking, product recommendation, qualification, signoff, or production optimization claim.

## Replacement Evidence Needed

Engineering reuse needs owned solver input decks, exact tool versions, mesh convergence records, material/model provenance, measured I-V/C-V or other calibration anchors, validation residuals, uncertainty budgets, DOE hull checks, constraint reviews, and device-owner approval.