Convergence checklist

A printable one-pager: run through this list before reporting any HELIX result as "converged".

Pre-flight checks

• [ ] continuous=True for any pre-RFQ (DC) beam.
• [ ] Ki=1 on any LEBT FIELD_MAP that has SC compensation.
• [ ] emit_nx, emit_ny are normalised (β·γ·ε), not geometric.
• [ ] emit_z is in deg·MeV (TraceWin native).
• [ ] Lattice path in .lgproj resolves on the current system (or use the relative-path fallback in parse_tracewin).

Particle count

• [ ] Run with n_particles ∈ {2k, 5k, 10k, 20k}; verify σ_x at end of lattice plateaus by 10k.

PIC grid

• [ ] Grid nx=ny=nz ≥ 64 for tests, ≥ 96 for production.
• [ ] grid_extent ≥ 5σ. Reduce to 3σ only if the beam is pinched and you've verified tail truncation is safe.
• [ ] kernel="cic" for RF-frequent lattices, "tsc" for long no-RF transports (BTL).
• [ ] green_kind="igf" always.

Step density

• [ ] step1_per_m ≥ 50 for drift / quad sections, ≥ 100 for cavity sections.
• [ ] For RFQ sections, n_steps per cell auto-picked or ≥ 20.

SC convergence

• [ ] Run with nx ∈ {32, 48, 64, 96} and verify σ_x at end plateaus by 96.

Validation

• [ ] TW parity — for production lattices, compare σ_x, σ_y end values to TraceWin partran. Expect ~3 % residual (PIC kernel calibration).

• 5 % residual ⇒ check convergence again, or check for parameter mismatch (Ki, continuous, ε normalisation).

Cross-references

• Convergence guide → models
• Reading results
• TraceWin parity

← Known limitations · Continue to Surrogates → Overview →