# Charge-fraction prediction, 2026-10-28

Four slices solved for the defense figure `defense/figures/cfrac_pred.png`,
archived here beside the solves that were already in `wires2d/vtu_files`.

| index | z (mm) | phase in the cell | relative azimuth | anode-cathode gap |
|---|---|---|---|---|
| 10 | 0.0000 | midpoint | 7.500 deg | 6.220 mm |
| 10 | 19.6900 | crossover | 0.037 deg | 4.339 mm |
| 11 | 19.7940 | crossover | 0.003 deg | 4.339 mm |
| 12 | 29.8500 | quarter | 3.794 deg | 4.908 mm |
| 13 | 39.9200 | midpoint, next cell | 7.435 deg | 6.247 mm |
| 14 | 44.7820 | three eighths | 5.625 deg | 5.535 mm |
| 15 | 49.9800 | three quarters | 3.699 deg | 4.925 mm |

Indices 10 and 11 were already here; 12 to 15 were added by this study.
Each index has an `elfield_anasen_*` (physical solve, anodes at 660 V) and an
`elfield_weight_*` (Ramo weighting solve, cathode 1 at 1 V) on the same mesh.
The archived 10 and 11 ran at 650 V, the new ones at 660; only ratios are
used, and the extraction reads the electrode potential off the wire interior
rather than assuming it.

`dotproduct.py` was not run on the new slices: it writes another ~1 GB per
slice and only the anode surfaces are needed.  `probe_fields.py` in the code
below computes the same product straight from the two field VTUs.

## Do not take E.E_w from the nodal "electric field" array

`FluxSolver` recovers the gradient element by element and a degenerate element
puts a spike on a node.  The meshes for 12 to 15 carry ~960 nodes above
1e6 V/m where meshes 10 and 11 have none, so the extremum over a wire's
surface nodes lands on a spike: a re-solve of z = 19.794 returned -1.526e12
where the archived value is -2.569e8.  Take it from the potential instead, as
`surface_field.py` does.  That is stable to five figures across meshes.

## The code

`scratch/cfrac_fem_pred_2026-10-28/` in the analysis repository
(github.com/vs19g/ANASEN-Analysis_private, branch devel_vignesh).  `run_all.sh`
reruns everything; `collect_slices.py` and `map_function.py` read the VTUs
here.  The CSVs in this directory are their output, so the numbers can be read
without re-solving.