From a FLASH current sheet to kinetic pressure measurements¶
The fresh Simjecture 0.7.0 investigation supports the local pressure-isotropy prediction within its tested scope. Independent claim and report reviews both accepted the retained evidence. Two fresh particle realizations remain below the 0.20 threshold after adding the declared numerical and sampling allowances.
The source and measurement panels locate the actual kinetic problem. The magnetic field and pressure maps use fresh seed 6101. The control panel includes both fresh seeds; rightward bars add the declared sensitivity allowances. They are empirical diagnostic envelopes, not statistical confidence intervals. Download figure PDF.
What connects the two solvers¶
The worker selects a current sheet from a fresh \(512^2\) FLASH trajectory at \(t/\tau_A=0.8001\), then transfers its density, scalar pressure, bulk flow, magnetic field and current into WarpX. A discrete-curl reconstruction preserves a solenoidal magnetic field; species drifts reproduce current and center-of-mass flow. The measured sheet half-width is \(1.0114d_i\), spanning 8.22 FLASH cells.
This is a one-way initialization of a collisionless local kinetic problem. FLASH’s scalar resistivity is not continued in PIC, and kinetic fields do not feed back into FLASH. The declared physical mapping uses \(m_i/m_e=25\) and initial \(T_i/T_e=1\); it does not infer a unique physical plasma from normalized MHD.
The worker freezes the source selection, mapping, measurement and decision rule before collecting claim evidence. Local species mean flow is subtracted from particle velocities before constructing the pressure tensor. The observable is
averaged over a fixed \(4\times4\) interior grid and \(t\Omega_{ci,\mathrm{ref}}\in[0.4,0.9]\). Missing or underpopulated bins invalidate the measurement rather than being discarded.
What the tests found¶
Species |
Seed 6101 |
Seed 6102 |
Maximum mean + allowance |
Threshold |
|---|---|---|---|---|
Ions |
0.04939 |
0.04779 |
0.14747 |
0.20 |
Electrons |
0.04733 |
0.05205 |
0.15448 |
0.20 |
The controls test source refinement, a \(384^2\to768^2\) kinetic mesh change, a halved timestep, 32 versus 64 particles per cell, a larger patch at matched cell size, finer measurement bins and independent particle seeds. The additive allowance also retains sampling, flow-shear and relativistic-pressure corrections. The independent reviewer reproduced the fresh means from native tensors and checked control outputs and identities before accepting the finite prediction.
The Debye scale remains underresolved, electric-field noise changes with numerical settings, and boundary-flux quadrature is approximate. The pressure controls and allowances support this threshold decision; they do not establish continuum convergence or a generally valid isotropic MHD closure.
Execution and independent review¶
The numerical study used Simjecture 0.7.0, commit
cfca558cb30868d7bc12af7a9b636303f7bc500b. GPT-6.1 Sol at high effort served as
worker, director and reviewer in separate contexts. It retained 25 successful
execution receipts and three failed receipts, including a fresh FLASH source
and two prospective evidence-stage kinetic runs.
The initial two-hour allowance ended before claim review completed. Report review also exposed a response-format mismatch: an accepted response included nonblocking comments in a field reserved for blocking issues. The operator subsequently authorized 30 minutes for review only. A case-owned controller used unchanged v0.7.0 review transport and validators, preserved the original numerical deadline and records, and completed both reviews in 469 seconds. No new simulations were run in that session. The scientific verdict came from the independent reviewer, with the controller and authorization retained in the audit.
Inspect and reproduce¶
From the repository root:
uv run python scripts/verify_research_audit.py demos/flash_warpx_kinetic_patch/fresh_v070_record
uv run --extra flash-demo python scripts/render_fresh_plasma_examples.py
The public audit preserves original receipts, review packets, sources and compact outputs. Large solver states are explicitly omitted; integrity verification and figure regeneration do not replace full numerical replay. Fresh investigations need qualified FLASH and WarpX capabilities and a configured agent backend. FLASH source and binaries remain operator-supplied.
The historical two-hour study and its continuation remain available with their unresolved outcomes. Those records are unchanged.