Skip to content
SASemiAgoraSemiconductor knowledgeResearch services
Menu
Research services한국어 안내LearnTraceCompareReproducePlanSearch
ExploreKnowledge AtlasDetailed mapSimulationsDomain mapGuided journeysLabs
WorkspacesScenario ComparisonReproducibility RegistryExperiment PlannerResearch Project BuilderClaim-Evidence WorkbenchData Contract
TrustEvidence Knowledge GraphMetric OntologyCitation ProvenanceSource WorkspaceEvidence BrowserPublic Release LedgerSource PolicyCorrections
MoreSearchPracticeCalculatorsResourcesResearch atlasRelease provenanceReading RoomResearch servicesAbout SemiAgora
SASemiAgoraSemiconductor knowledge & engineeringOpen run card

LESSON PACKET / SA-DEV-TCAD-OPT-004

Optimization needs a surrogate boundary.
Device / TCAD

A TCAD optimization result is useful only when DOE coverage, mesh stability, calibration anchors, surrogate uncertainty, extrapolation risk, physics constraints, and owner review stay in one bounded packet.

Open simulationOpen replay lab ->
packetSA-DEV-TCAD-OPT-004runSimulation ObjectlearnEvidence SequencemethodMethodslimitsBoundaries

EVIDENCE SEQUENCE

What the learner should inspect first.

This surface connects the daily-scan source signal to an original SemiAgora simulation object without copying source text, code, figures, PDK files, or process recipes.

01

Start with device context, design lever, DOE plan, surrogate family, objective priority, and validation posture before reading a pareto-readiness score.

Open the run card to inspect the saved rows, methods, limits, and export files behind this step.

Inspect run card ->
02

Compare mesh stability, calibration anchor strength, sample efficiency, surrogate uncertainty, extrapolation risk, constraint violation, objective fit, and pareto readiness together.

Open the run card to inspect the saved rows, methods, limits, and export files behind this step.

Inspect run card ->
03

Use the decision gate to decide whether the next safe check is mesh convergence, calibration, uncertainty, extrapolation, constraint, owner review, bounded teaching, or reference-bounded evidence.

Open the run card to inspect the saved rows, methods, limits, and export files behind this step.

Inspect run card ->

METHODS

Precomputed model contract.

The goal is useful study scaffolding and paper-planning discipline, not calibrated production evidence.

method

Original synthetic TCAD DOE and surrogate-literacy table based on public DEVSIM, NIST, SciPy, scikit-learn, Synopsys, and arXiv vocabulary.

The simulation object keeps the same statement beside downloadable JSON and CSV evidence.

method

No TCAD deck, mesh, structure file, PDK, process recipe, solver transcript, source code, documentation example, measured silicon, vendor table, or optimizer run is copied.

The simulation object keeps the same statement beside downloadable JSON and CSV evidence.

method

The browser filters saved CSV rows and exposes the model card with no-live-solver, no-calibration, and no-product-claim boundaries.

The simulation object keeps the same statement beside downloadable JSON and CSV evidence.

LIMITS

Useful, but deliberately bounded.

No account, upload, payment, comment, newsletter, live solver, facility workflow, or product-release surface is enabled.

non-claim

No live TCAD, SPICE, process, electrothermal, RF, field-solver, optimizer, uploaded deck parser, or server-side model.

Keep this boundary attached when reusing the card for lesson writing, figure planning, or research notes.

non-claim

No PDK, TCAD deck, process recipe, measured I-V/C-V, calibrated compact model, foundry-node ranking, product recommendation, qualification, or signoff claim.

Keep this boundary attached when reusing the card for lesson writing, figure planning, or research notes.

non-claim

Real work needs owned solver decks, mesh convergence, model provenance, measured calibration anchors, DOE hull checks, uncertainty budgets, and device-owner review.

Keep this boundary attached when reusing the card for lesson writing, figure planning, or research notes.

SOURCE ANCHOR

Link-only public source spine.

SemiAgora uses the source as a signal and citation anchor, then writes original summaries and original toy data with explicit limitations.

DEVSIM, NIST, SciPy, and scikit-learn source spine

https://devsim.org/index.html

Open source ->