Inspect two to four public replay slices side by side. SemiAgora exposes the numeric table only after metric identity, grain, evidence class, source hash, and unit conversion pass together.
Model boundary. Original dimensionless teaching replay only; no physical value, recipe, equipment action, calibration, qualification, reliability, or production claim.
Replacement evidence slots (20)
accountable run identifier, wafer ownership, review owner, and disposition owner
revisioned chamber, source, shield, collimator if present, and wafer-stage configuration
owned target identity, backing, geometry, bonding state, erosion map, use history, and change history
base-vacuum, residual-gas, leak-check, pump, throttle, and chamber-history evidence with uncertainty
owned process-gas identity, composition, flow, pressure, and transient evidence with uncertainty
power, voltage, current, waveform, match, and arc-event evidence with calibration and uncertainty
magnetic-field, source-state, cooling, and target-surface evidence with traceability
wafer position, orientation, motion, rotation, temperature, and backside-contact evidence
calibrated incident energy, angular distribution, flux, and ion-neutral fraction evidence or validated equivalents
revisioned chamber, target-to-wafer, shield, feature, and pattern-density geometry
deposition rate, thickness, center-edge, azimuthal, wafer-to-wafer, and chamber-to-chamber maps
top, middle, bottom, and sidewall cross-section measurements with uncertainty and sampling locations
continuity, sheet-resistance, contact-resistance, adhesion, and interface evidence where applicable
resputter, redeposition, overhang, pinch-off, void, seam, and open-defect evidence with definitions and detection limits
film-stress and wafer-curvature evidence with film thickness, substrate properties, uncertainty, and sign convention
roughness, density, composition, texture, phase, grain, impurity, and interface evidence appropriate to the claim
within-wafer, wafer-to-wafer, lot-to-lot, target-life, maintenance-cycle, and tool-matching variation evidence
model version, source assumptions, geometry, parameters, deterministic posture, sensitivity, residuals, and applicability limits
independent model-to-measurement review, rights review, limitations review, and accountable approval
08
Limitations
These statements are included in every permitted export.
The notebook compares selected original dimensionless teaching proxies; it does not establish physical equivalence, calibration, accuracy, causality, process fitness, or scientific truth.
A shared display unit does not make different metric contracts, families, grains, evidence classes, or model boundaries comparable.
The four source records provide evidence ownership and links; their bodies, numbers, equations, apparatus details, and results are not copied into or used numerically by V125.
The sixteen selected slices are a deterministic teaching subset of public representative states, not a complete replay dataset or statistically representative sample.
No live solver, background job, arbitrary code, upload, private dataset, account, remote project, recommendation, ranking, qualification, or signoff is provided.
Researcher-facing businesses remain deferred to a separately verified V200 contract.
DESCRIPTIVE EVIDENCE BOUNDARY
Comparable does not mean predictive, causal, or production-ready.
The notebook preserves selected dimensionless teaching proxies and public lineage. It does not rank scenarios, infer desirability, validate a process, recommend a recipe, or replace owned measurements, uncertainty, calibration, and accountable review.