[ryoji-info]

Adds three outstanding rows to papers/ERRATA.md against the deposited Part I v6.1 (10.5281/zenodo.23758795). Each row quotes the deposited sentence (page numbers of the v6.1 PDF) and the primary source it was checked against, fetched 3 November 2026. They come out of the 3 November 2026 review session on surface plasmon polaritons (its papers lens, verdicts 5, 9 and 16).

• I-41 — Table 3 'Plasmon mode' (p.10): "required for confinement and gate programmability [4]". Programmability does not require screening, and [4] does not say it does: Yoshioka et al. gated the unscreened mode with a THz-transparent ZnO top gate ("investigate plasmons free from the gate screening … while tuning the carrier density"); Tu et al. 2020 (Commun. Mater. 1, 7) programmed cavity size, position and frequency through a patterned ZnO back gate of ≈90 % transmission over 1.3–10 THz; Ju et al. 2011 tuned ribbon plasmons by in-situ electrostatic (ion-gel) doping. Accurate wording: required for confinement and for ≈1-V, sub-µm metal-gate programmability (and for the gate-referenced source short of §9.1).
• I-42 — §3.2 (p.8) and §9.5 (p.39): "unquantified by any cited work" / "no cited work quantifies the coupling efficiency". Ref. [4] reports a lower bound η > 0.35 for the unscreened (ZnO-gated) plasmon and η ≤ 0.03 for the acoustic (Au-gated) one, at 4 K; its abstracts credit the absence of a momentum mismatch, its body names the CPW–graphene impedance mismatch (smaller under the ZnO gate) as the limit. What no work quantifies is coupling into the acoustic mode at 300 K on this stack — §3.3 and Table 6 already say that.
• I-43 — §4 (p.11): "gate-metal ohmic loss, not included in τ, adds to the budget at small d [41]". [41] (Woessner 2015) is a back-gated, mid-IR hBN(7 nm)/graphene/hBN(46 nm)-on-SiO₂ study with no metal top gate and no gate-metal loss; Principi 2018 (PRB 98, 035427) and Dias 2018 (PRB 97, 245405) do treat it. Recorded as unquantified: the manuscript gives no gate thickness or resistivity.

Also touched: README.md and CITATION.cff each lose the phrase "(none outstanding since the October 2026 deposits)". The manuscripts, PDFs and DOI metadata are untouched. python ledger_sweep.py reports 0 undisposed hits; pytest tests/ passes (50). CRLF preserved in all three files.

Convention flag. ERRATA.md's own rule opens rows for promoted notes, and the v6.1 block of REVISION-HISTORY carried that revision's four unbound corrections in the block rather than as rows. These three have no promoted note; the new preamble says so plainly and treats the merge as the maintainer's decision to carry them as rows. If they belong elsewhere (a dated note, or the next revision block), say so and they move.

What did not check out against the brief. (1) The ≈3 % figure for the Au-gated acoustic device is not "secondary press only": it is in [4]'s main text ("η can be as large as 0.03", propagation-length-corrected). (2) Both abstracts attribute the 30–35 % to the absence of a momentum mismatch; the impedance-mismatch attribution is the body's explanation of why the ZnO device does better and what limits both — the row carries both statements. (3) Tu et al.'s ZnO gate is a patterned back gate (dual back gate with the Si substrate), and the journal title ("Active spatial control of terahertz plasmons in graphene") differs from the arXiv title. (4) Ju et al. 2011's body is paywalled (no arXiv, OSTI or Semantic Scholar copy found): the "3–6 THz" figure is unverified and the row quotes only the abstract and the acknowledged ion gel. (5) Dias et al. is arXiv:1804.06478. The image-current estimate for the gate loss (Q_metal = ωL_k/R_s ≈ 1.6 × 10³ for bulk Au at 1 THz and n = 10¹² cm⁻², ≈0.4 % extra damping at τ = 1 ps; ≈23 % for a 15-nm Pd gate) is kept out of the row as briefed — derived in the review, not adopted.

🤖 Generated with Claude Code