Screening Study and Touschek

简体中文 · Previous: Equilibrium chain · Next: Nonlinear dynamics

Screening Study orchestrates existing formal workflows. It does not introduce another physics implementation or result store; it places project, target, dependencies, warnings, and evidence in one state graph.

Stages and states

The main chain is Validate Project → Radiation Slice Model → Equilibrium, branching to Matched Distribution and Local Momentum Acceptance.

State	Meaning
Not ready	Project or target snapshot is incomplete
Ready	Every predecessor and qualification gate passes
Running	A formal one-shot workflow is active
Completed	Parser, hash, identity, and Engine evidence passed
Completed with warnings	An explicitly allowed warning remains visible
Blocked	A prerequisite, convergence, warning, or compatibility gate failed
Stale	Evidence belongs to another target/content/physics revision
Failed	Workflow or evidence validation failed
Client-aborted	Local client stopped; no durable remote-cancel claim

After restart, Completed stages are reconstructed from verified stores, not UI memory.

Execution buttons

• Run next available stage: first runnable stage in dependency order.
• Run to equilibrium: run to equilibrium, reusing only exact reverified radiation evidence.
• Run available downstream screening: start eligible Matched and/or LMA siblings.
• Abort client: stop active one-shot clients; this is not durable cancellation.

Before remote execution, Studio displays the target, exact identity, actual route, interruption semantics, and fixed physics boundary.

Corrected Touschek v0.7

This is the only route enabled for new Touschek submissions. It is a durable CPU screening job with submit, continued execution after exit, reconnect, cancellation, and four-artifact fetch.

Submission requires a complete v0.2 project, valid equilibrium on the exact target/revision, LONGITUDINAL_LINEAR_APPROXIMATION=pass, explicit positive bunch population, explicit positive geometric εy, and compatible v0.7 Engine capability.

Click Submit corrected screening in the Touschek panel. During execution, use Reconnect / fetch result or Cancel request. A completed view reports screening lifetime, positive/negative acceptance contributions, formula/cross-code audit, and all four artifact identities.

Its permanent scope is SCREENING ONLY, formula-level cross-code validated, sparse physical aperture, not a machine prediction, and not total beam lifetime. Gas scattering, quantum lifetime, IBS, impedance, errors/correction, and real machine state are excluded.

Why v0.6 is historical only

The v0.6 touschek-lifetime/0.2 used a wrong Equation 42 normalization and is permanently marked:

KNOWN FORMULA BUG · SUPERSEDED · HISTORICAL FAILURE EVIDENCE

New Submit and replay are disabled. Previously submitted jobs retain Get/Watch/Reconnect/Cancel/ListArtifacts/FetchArtifact. Their four JSON artifacts enter a separate quarantine with physicsVerified=false; the wrong scalar can never become a formal result card.

For the frozen diagnostic benchmark, corrected local evidence is about 783379.643 s, PyAT about 783380.428 s, and old v0.6 about 5286.457 s. These diagnose one fixed benchmark and are not a lifetime for your project.

Example: equilibrium to v0.7

1. Open a reviewed v0.2 project with explicit beam population and vertical emittance.
2. Select the v0.7 target.
3. Produce valid lattice-derived equilibrium with the longitudinal-linear check passing.
4. Open Screening Study and confirm corrected Touschek is Ready.
5. Submit and record the tsj7-* job ID.
6. Detach or close Studio; reopen and Reconnect later.
7. Inspect formula audit, project/revision, Engine source, and four artifacts before the screening lifetime.

If the gate is Blocked, read every detail. Never fill LMA with RF bucket height to bypass it.