Proxy
An original normalized, synthetic, or surrogate teaching state that supports bounded comparison only.
A proxy is not a physical prediction, calibration result, measurement, or causal proof.
20 V123 edgesCausal chain navigator / V123
Trace where semiconductor evidence can travel, and where it must stop.
Seven fixed paths connect public process, material, device, circuit, metrology, packaging, and SRAM artifacts. Every handoff names its assumption, limitation, evidence class, and missing measured replacement.
These are curated associations for review. They do not infer or prove physical, statistical, or production causality.
- Curated paths
- 7
- Evidence nodes
- 32
- Inspectable edges
- 25
- Required stops
- 7
Evidence legend
Four evidence classes, shown in text.
Color is only a secondary cue. The class label and claim boundary remain visible on every path.
Derived
A deterministic calculation or extraction from stated inputs, equations, and conditions.
A derived result inherits every input and model assumption and is not measured evidence.
3 V123 edgesReproduced
An output regenerated with a documented public model or tool under the displayed conditions.
Reproduction establishes the stated run result, not measured-silicon agreement or transfer to another context.
2 V123 edgesMeasured
A rights-clean observation with method, conditions, uncertainty, source, and metric contract attached.
A measurement supports only its declared grain and conditions; it does not automatically prove cross-domain causality.
0 V123 edgesPath directory
Seven paths. Every bridge is inspectable.
Open a path to inspect its ordered nodes, exact handoff fields, source artifacts, and final evidence stop.
process-geometry / device-metric / circuit-behavior / system-yield-proxy / stop
Contact resistance to read timing
Trace a bounded contact-module state into Rc/Rsd, delay sensitivity, and an SRAM read-timing question without claiming a qualified module or yield result.
Where does a contact-state proxy stop being defensible as it moves toward product behavior?
- Nodes
- 5
- Edges
- 4
- Proxy / derived / reproduced
- 4 / 0 / 0
- Stops
- 1
process-geometry / process-geometry / device-metric / circuit-behavior / stop
Roughness and local CD to margin
Follow a profile or local-CD proxy toward device variability and circuit margin while preserving the missing LER, calibration, and yield evidence.
Which geometry observations are required before a profile change can support a device or SRAM claim?
- Nodes
- 5
- Edges
- 4
- Proxy / derived / reproduced
- 4 / 0 / 0
- Stops
- 1
material-interface / material-interface / device-metric / stop
Deposition and interface evidence
Trace surface activation toward nucleation and interface electrical evidence, then stop before an unsupported circuit consequence.
What must replace the public deposition proxies before interface capacitance can enter a circuit claim?
- Nodes
- 4
- Edges
- 3
- Proxy / derived / reproduced
- 3 / 0 / 0
- Stops
- 1
process-geometry / material-interface / device-metric / circuit-behavior / stop
Thermal budget to delay question
Follow normalized anneal budget through activation and diffusion proxies into device and FO4 sensitivity, then stop before Vmin or yield claims.
Which thermal, device, and circuit conditions must stay attached across the handoff?
- Nodes
- 5
- Edges
- 4
- Proxy / derived / reproduced
- 4 / 0 / 0
- Stops
- 1
measurement-decision / measurement-decision / measurement-decision / system-yield-proxy / stop
Metrology uncertainty to decision risk
Trace study design into variance components, guardband logic, and false-pass or false-fail risk without qualifying a gauge or disposing material.
How does measurement-system uncertainty change what a decision proxy can support?
- Nodes
- 5
- Edges
- 4
- Proxy / derived / reproduced
- 3 / 1 / 0
- Stops
- 1
process-geometry / device-metric / circuit-behavior / stop
Packaging parasitics to system margin
Follow a bounded package RLC budget into electrical response, then stop before a product bandwidth, timing, SI, or qualification claim.
What extraction and measurement evidence is missing between a package proxy and system behavior?
- Nodes
- 4
- Edges
- 3
- Proxy / derived / reproduced
- 1 / 2 / 0
- Stops
- 1
device-metric / circuit-behavior / system-yield-proxy / stop
SRAM cell to array-read evidence
Trace reproduced cell and bitline artifacts into a sense-decision question, then stop before array Vmin, coverage, yield, or qualification.
Which model and measurement handoffs are required to move from a cell replay to an array claim?
- Nodes
- 4
- Edges
- 3
- Proxy / derived / reproduced
- 1 / 0 / 2
- Stops
- 1
Interpretation boundary
Association is not causality.
The navigator presents curated associations, assumptions, evidence classes, and replacement needs. It does not prove physical or statistical causality.
No V123 edge is measured causality. The measured class is defined for future rights-clean replacement evidence and remains at zero edges in this release.
No public path authorizes process changes, equipment control, wafer disposition, qualification, signoff, product ranking, or production release.
Review evidence provenance