Macroflow Physics Review

For PSPTA-related changes, verify that the change aligns with the current phase of the execution plan before accepting.

Skip if the change is Gate 0 or Gate 1 only.

Step 2 — Classify the change

Map changed files to the required acceptance gate:

Reference: docs/validation/acceptance-gates.md

Step 3 — Check invariant risk surfaces

For any change in the scientific core, apply the Lester constraints (docs/theory/lester-2023-key-claims.md):

1. Does this change the velocity field structure?

   • Divergence-free is necessary but not sufficient for invariant geometry.

2. Does this change invariant construction quality?

   • Near-invariance, independence, refinement stability.

3. Does this change tracking behavior?

   • Newton convergence, projection accuracy, particle exit conditions.

4. Does this change macrodispersion interpretation?

   • Moment computation, ensemble statistics, α_L / α_T output.

5. Could this create spurious transverse macrodispersion?

   • In the smooth, locally isotropic, purely advective regime, positive transverse macrodispersion is NOT automatically physical.

Step 4 — Require evidence, not intuition

For each affected gate, verify the author provides:

Gate 1 (build + smoke)

• Configure succeeds
• Build succeeds
• Tests pass
• Smoke run completes: ./build/*/macroflow3d_pipeline apps/config_pspta_small.yaml

Gate 2 (operator integrity)

• ctest -R operator_tests passes
• ctest -R validate_slepc_eigensolver passes (if eigensolver touched)
• Residual norms reported
• Before/after comparison if behavior changed

Gate 3 (PSPTA local integrity)

• v·∇ψ1 and v·∇ψ2 residuals inspected
• Independence / degeneracy signal checked
• Newton failure counts reported
• Particle status summary (active / exited / failed)
• No unexplained metric regression

Gate 4 (helicity-free regime)

• Controlled pure-advection test case
• Transverse spreading explicitly characterized as physical / numerical / unresolved
• Invariant confinement behavior checked

Gate 5 (macrodispersion)

• Baseline comparison with prior known-good run
• α_L(t), α_T1(t), α_T2(t) before/after
• Config and build artifacts recorded
• Run is reproducible

Step 5 — Block or approve

Block if:

• required gate evidence is missing
• residuals materially worsened without explanation
• Newton failures exploded
• positive transverse macrodispersion is claimed as physical without control
• diagnostics were weakened silently

Approve only when:

• evidence matches the gate requirements
• remaining uncertainty is stated explicitly
• no silent behavior changes

Special caution areas

• Near-stagnation regions (invariants and Newton become ill-conditioned)
• Functional dependence collapse (two modes becoming nearly dependent)
• Projection masking (numerically stable but geometrically leaky)
• Diagnostics drift (renamed, filtered, or weakened metrics)

Report template

Use the template from docs/validation/acceptance-gates.md:

## Scientific change report

### Scope
- files changed:
- intended effect:

### Commands run
- configure:
- build:
- tests:
- smoke:

### Evidence
- gate(s) satisfied:
- metrics inspected:
- before/after comparison:

### Remaining risk

Related

• docs/validation/acceptance-gates.md
• docs/theory/lester-2023-key-claims.md
• docs/theory/beaudoin-de-dreuzy-2013-key-claims.md
• src/physics/particles/pspta/AGENTS.md
• src/numerics/AGENTS.md
• docs/runbooks/autonomy-policy.md