Scope — pmid-41708617

Paper: Vörös Z. et al., "Turbulent dynamo in the terrestrial magnetosheath", Nat Commun 2026. PMID 41708617 · PMCID PMC13032041 · DOI 10.1038/s41467-026-69469-y. Manuscript ref NCOMMS-25-30764A.

Code: irfu-matlab (https://github.com/irfu/irfu-matlab, also zenodo:11550090) — a third-party MATLAB toolbox for MMS space-physics analysis. The authors' own figure-generating MATLAB scripts ship in the Zenodo data package (see below). Per BRIEF rule P16, applying / reading the shipped scripts on the shipped data is a fully valid reproduction.

Data: zenodo:10.5281/zenodo.17780770 → single archive CodesData.zip (13.8 MB, sha256 ba3f588ad7d1081b48f323d1861df3e3a1a7e0ad7a6a6902128fe55582f3431f). Contents:

codes/Figure{1,2,3,4}_script.m      MATLAB plotting scripts (irf_plot etc.)
data/Figure{1,2,3,4}_data.mat       MATLAB v7 .mat, pre-computed derived quantities
readme.txt

The .mat files contain the already-derived physical quantities as irfu TSeries/EpochTT objects plus plain numeric arrays. The scripts only plot them — there is NO heavy pipeline that runs from raw MMS CDFs in this package; the heavy MMS→derived-quantity step (c_4_grad gradients, rotate_tensor, resampling) was done by the authors and its outputs shipped. Raw MMS brst CDFs (fgm, fpi dis-moms/dis-dist, 2015-11-30) live at the LASP SDC.

In scope (pipeline-derived, reproducible from shipped data, NO MATLAB needed)

Decode the v7 .mat with scipy.io.loadmat (these are NOT v7.3/HDF5, so the HDF5-MCOS decoder from pmid-35105885 is not needed; scipy reads v7 directly). Then recompute / verify the reported scalar results:

• C1 — Equation 1 internal consistency. Eq.1: d lnB/dt = bb:∇Vi − ∇·Vi. Data ships dBts (measured d lnB/dt), bbgradVts (bb:∇Vi), divVts (∇·Vi). Verify dBts ≈ bbgradVts − divVts pointwise (this is the paper's central balance, Fig.1d / Fig.2d).
• C2 — anisotropy of magnetic length scales l∥/l⊥ = 40 ± 20 (Results / Fig.2c). From lpar, lperp TSeries arrays in Figure2_data.mat.
• C3 — mean ion gyroradius ⟨ρi⟩ ≈ 95 km (paper) — NB the shipped Fig.2c script annotates "~99 km". Record both; flag the 95-vs-99 mismatch.
• C4 — mean electron gyroradius ⟨ρe⟩ ≈ 0.9 km (paper & script agree).
• C5 — Case I (folded field, I1 = 2015-11-30T00:24:06.8–00:24:12.8): cs(d lnB/dt) decreases by −6 ± 0.7 s⁻¹ (Fig.3). From btimes_sum over I1.
• C6 — Case II (mirror mode, I2 = 2015-11-30T00:25:22.9–00:25:27.2): cs(d lnB/dt) increases ~+6 ± 0.5 then decreases −9 ± 0.9 s⁻¹ (Fig.4). From btimes_sum over I2.

Out of scope (not attempted, and why)

• Raw MMS CDF → derived gradient quantities (c_4_grad on the tetrahedron, rotate_tensor, 150 ms barycentre resampling). The authors' derived outputs are shipped; re-deriving from raw L2 CDFs would require pulling the full brst MMS dataset + a MATLAB irfu-matlab run — the hard ~20%, explicitly skipped.
• Prandtl-number estimate Pm ∈ (400, 3600) — a scaling argument from l∥/l⊥ and assumed diffusivities, not a direct pipeline output; checked only for plausibility.
• Pitch-angle spectrograms (Fig.3/4 panels) — qualitative figure panels, no scalar to grade; need iPDistpitch PDist objects + irf_spectrogram (MATLAB).
• Instability threshold curves (Fig.2b) are analytic formulae (Hellinger/Gary), not data — out of scope as a "reproduction".

Method

scipy.io.loadmat decode of the 4 shipped v7 .mat files, on «our HPC» (compute node, pip venv with numpy+scipy). Data staged on «infra»; only small derived scalars + this scope come back to «host». Compute is trivially light (<1 s, <15 MB).