Paper 1 Preliminary Results

2026-09-11 — corrected source candidate (analytical check only)

The editable D3 scanned-plate Design was corrected and saved through the GUI: Gaussian s-type 13, 5 MeV electrons, local +Z, emission Z = -0.005 cm, 2 cm FWHM, X scan -48.5 to +48.5 cm, 321 trapezoidal nodes, totfact = 1. The source restriction was removed (reg omitted) to avoid conditioning the Gaussian on the narrow source air cell. The 3 cm RAMAL preview support remains a display/containment parameter and is not a physical cutoff.

The existing immutable Q21 edge-loss Case and Result below are unchanged and do not represent this revised Design. No new PHITS result is claimed here.

python3 tools/check_paper1_scan_profile.py compares the unconditioned source mixture with the exact uniform-line/Gaussian convolution. It samples 10,001 X coordinates over the scan interval plus four sigma at each end. Error is the maximum absolute density difference expressed as a percentage of the continuous interior plateau (1 / scan length), not relative error in tails.

Nodes	Spacing (cm)	Maximum profile error (% of plateau)
21	4.85	127.813578
41	2.425	31.948897
81	1.2125	4.449213
161	0.60625	1.045594
321	0.303125	0.257952

These comparisons all use the corrected inset interval. For Q321, the analytical normal-incidence interception of the 100 x 10 cm plate is 99.971236%, leaving approximately 0.028764% geometric spill. This calculation ignores upstream scattering and is not PHITS transport evidence. The inset must be called reduced-spill, not strictly no-spill: an unbounded Gaussian always has tails. This remains a one-dimensional line scan with Gaussian Y distribution, not uniform irradiation of the full plate face.

Q321 is a refined preliminary candidate, not an accepted scientific threshold. Next: create a new immutable Case from this Design, verify the incident profile and downstream transport in PHITS, and compare source-resolution sensitivity before D3/D5 acceptance. Multi-source normalization remains per primary source electron; old downstream fields must not be relabelled as corrected results.

This file records traceable preliminary calculations used to refine the open Paper 1 decisions. These Results are not D3 acceptance evidence and are excluded from the production Dataset unless a later approved protocol explicitly says otherwise.

2026-09-10 — D3 scanned-plate transport discovery

Status: completed and integrity-verified; preliminary transport discovery only. Not D3 acceptance. Not eligible for D7 acceptance or the production Dataset.

Provenance

Record	Value
Project	`industrial-eb-x-ray-dose-prediction`
Design	`d3-industrial-scanned-plate-preliminary`
Case	`preliminary-transport-discovery-not-d3-acceptanc-001`
Job	`job-001` — `Preliminary transport discovery run 1`
Attempt / Result	`a0001` / `result-0001`
Execution	local PHITS; return code `0`
Start / completion	`2026-09-10T13:14:35Z` / `2026-09-10T13:41:01Z`
Random seed	`1639214267`
Histories	`10,000` per batch, `10` batches, `100,000` total
Field	`103 x 20 x 40` cells; `82,400` voxels
Quantity	ambient dose equivalent H*(10), PHITS multiplier `-200`
Normalization	`pSv/primary-source-electron` at the source plane

The immutable Case contains a 5 MeV monoenergetic electron source, a local Gaussian spot with 2 cm FWHM and declared 3 cm support diameter, and 21 scan centres from -50 cm to +50 cm. The uniform continuous scan is approximated with trapezoidal weights: 0.025 at both endpoints and 0.05 at the 19 interior nodes. The weights sum to one and totfact=1.

The component is a rectangular 100 x 10 x 0.08 cm natural-tantalum plate, followed by 0.2 cm water and 0.2 cm Type 304 stainless steel. The water and steel thicknesses are interim Petwal-derived transport assumptions, not an industrial thermal or mechanical design. The endpoint footprint spill is intentional: scan centres reach the plate edges while the declared spot support extends another 1.5 cm laterally. This Result therefore represents an edge-loss diagnostic, not a representative full-interception baseline.

Execution and transport findings

PHITS completed all ten batches in 1567.14 s total CPU time.
PHITS reported zero geometry warnings, zero geometry-error terminations, zero recovered geometry errors, and zero lost particles.
The requested low-energy neutron tables resolved to the available 50c JENDL-4 family. PHITS reported that 18040.50c lacks gamma-ray-production cross sections.
PHITS reported missing 1001.20u photonuclear data and used its physical model for the corresponding hydrogen reactions. This is a declared transport-applicability limitation, not an execution failure.
The PHITS production summary recorded 404 photons and 202 positrons. It did not list produced neutrons.

Field observations

The verified combined field maximum is 1.5471 pSv/primary-source-electron; RAMAL-EBX reports 5.337% relative standard error at that maximum-dose voxel. Particle-page maxima read directly from the persisted VTK field are:

PHITS page	Particle	Maximum (`pSv/primary-source-electron`)	Nonzero voxels
`p1`	all	`1.5471`	`81,856`
`p2`	electron	`1.5377`	`6,958`
`p3`	photon	`0.038964`	`81,814`
`p4`	positron	`0.00014793`	`466`
`p5`	neutron	`0`	`0`

Many low-dose voxels remain poorly converged; one inspected voxel had 27.23% RSE, and the uncertainty field reaches 100% in scored voxels. The Q21 source quadrature is also unconverged. Consequently, this field must not establish D7 statistical thresholds, converter efficiency, accepted converter thickness, or a production reporting range by itself.

Consequences and next comparisons

Retain this immutable Result as pipeline, transport-resolution, and edge-loss evidence only.
Create a paired no-spill preliminary Design by retaining the 100 cm plate and contracting scan-centre travel to -48.5 cm through +48.5 cm. This changes only the edge-interception condition and is preferred before adding another plate-size variable.
Compare endpoint-spill and no-spill fields before freezing revised D3.
Then test scan quadrature convergence with Q21, Q41, and Q81 using a predefined field/profile difference measure.
Repeat transport checks at the upper Paper 1 energy boundary before drawing any conclusion about neutron relevance. The observed zero-neutron page at 5 MeV is not evidence for the entire 5–7.5 MeV domain.
Use the same immutable VTK field for the existing 1D profiles and exploratory XY/XZ/YZ 2D slices. The 2D view supports physical slice position, particle-page selection, linear/log10 values, and the aligned RSE field, but still needs fixed cross-Case colour ranges and provenance-bearing exports before it is a paper figure or cross-Case comparison.