What I verified, and how
I rebuilt triangulene (C22H12) from honeycomb coordinates and confirmed D3h numerically: C3, three in-plane C2 axes at 30°, σh, S3 and σv all map the 34-atom set onto itself. Fixed-atom counts are E 34, C3 1, C2 6, σh 34, S3 1, σv 6 — the C3 count of 1 confirms a carbon sitting exactly on the axis. Reducing Γ_3N (dim 102 = 3N) and removing translations (e′ + a2″) and rotations (a2′ + e″):
Γ_vib = 11a1′ + 10a2′ + 22e′ + 3a1″ + 8a2″ + 10e″
This checks exactly against planarity: in-plane 11 + 10 + 2(22) = 65 = 2N−3; out-of-plane 3 + 8 + 2(10) = 31 = N−3; total 96 = 3N−6.
Hückel on the 22-centre π system gives exactly two eigenvalues at x = 0. Projecting that 2-D NBMO space onto the D3h operations returns characters (2, −1, 0, −2, 1, 0) — e″, exactly. So the frontier configuration is (e″)², and since E″⊗E″ = A1′ + A2′ + E′ with antisymmetric part A2′, the ground state is ³A2′, a triplet, with ¹E′ and ¹A1′ above it. This is what the paper's own reference [1] (Pavlíček 2017) established.
Three things this makes wrong
1. The premise has an empty solution set. I enumerated all 21 state-symmetry pairs. Every distinct-irrep product falls in {A2′, E′, A1″, A2″, E″} — and Γ_vib contains all of them, the scarcest being a1″ at 3 modes. For identical 1-D irreps the product is A1′, coupled by 11 totally symmetric modes. There is no pair of electronic states of triangulene in D3h whose linear interstate coupling vanishes across all 96 modes. The paper's stated design target — "identify when the product contains no available mode" — is unmeetable for the molecule it names. At 34 atoms the vibrational manifold is simply too rich; this rule can only bite for very small or very highly symmetric systems, which is precisely the scope question the paper never asks.
2. The mechanism is inverted. The paper says that if symmetry forces the coupling to vanish, "the intersection is lifted to an avoided crossing." Degeneracy requires ΔE = 0 and V = 0 — two conditions, giving a seam of dimension 3N−8 = 94. If symmetry annihilates V along every mode, one condition remains and the seam grows to 3N−7 = 95. The degeneracy is not lifted; it becomes a higher-dimensional glancing (Renner–Teller-type) intersection that is geometrically easier to encounter. An avoided crossing is what you get when V is non-zero — the same-symmetry case, coupled by the 11 a1′ modes — which is the opposite of the paper's premise. The central inference of the paper runs the wrong way.
3. The protective geometry guarantees the intersection. The lowest excited singlet is ¹E′, degenerate, and [E′⊗E′]_sym = A1′ + E′, so it is first-order Jahn–Teller active along all 22 e′ modes. A symmetry-required conical intersection sits exactly at the D3h point the paper offers as protection. Here, raising symmetry creates the crossing rather than removing it.
What is right
The stated condition Γa ⊗ Γq ⊗ Γb ⊇ A1 is correct as written. The paper does not fabricate: it says plainly that the calculations are "proposed, not reported," and I found no invented measurements or benchmark runs. Under this rubric an honest proposal is legitimate, and that keeps this out of the fabrication category. The limitations section names Herzberg–Teller reactivation and substituent-induced distortion candidly.
What is missing
The rule is never stated. No irreducible representation is named anywhere in the paper. The sentences "we enumerate," "tabulating the expected qualitative ordering," and "specifying functionals, active spaces, and the geometric criteria" describe artefacts that do not appear — no functional, no basis set, no active space size. This is a description of a paper rather than the paper, and the one original increment (the triangulene tabulation) is exactly the part withheld.
Spin is also never mentioned. With a triplet ground state, decay out of the singlet manifold is spin-forbidden and governed by spin–orbit coupling, on which a spin-free selection rule is silent. Magnitudes matter here: for a hydrocarbon SOC is small (ζ_2p ≈ 28 cm⁻¹, S–T elements ~0.1–1 cm⁻¹), but so is the residue after "forbidding" a linear term — second-order coupling of order λ²/ΔE ≈ 0.01 eV ≈ 80 cm⁻¹ for typical λ ≈ 0.1 eV and ΔE ≈ 1 eV, still ~2 orders above SOC. "Forbidden at linear order" does not imply "suppressed," and no magnitude comparison is offered anywhere.
What would fix it
Drop triangulene, or drop the emitter framing — triangulene dimerises and had to be made by tip manipulation on a surface, per the paper's own reference [1]. Then: state the rule; publish Γ_vib and the frontier irreps; replace "forbidden" with a ranked scarcity argument, since the a1″ channel at 3 of 96 modes is the only genuinely thin one and is a defensible if weak result; redo the codimension analysis correctly; and treat the E-state Jahn–Teller problem head-on. One CASSCF/NEVPT2 MECI optimisation on the parent would be worth more than the entire proposed tabulation.
Scores
Novelty 3. The condition is textbook — and the paper cites the source of it, Köppel–Domcke–Cederbaum 1984, as its own reference [3]. The only claimed increment is the triangulene application, which is absent from the text and which, when I carry it out, returns the empty set. Against the low anchor ("no demonstrated gain"), this is below bar.
Rigour 3. No fabrication and honest scoping keep it above 2, and the selection-rule condition it does state is correct. But the inference drawn from it is physically inverted, the premise is vacuous for the named system, and the open-shell ground state is never reconciled with the two-state closed-shell picture. Real gaps a competent peer would not pass.
Clarity 4. The prose is clean and well-organised, but this rubric anchors clarity on reproducibility, and nothing here is reproducible: zero computational settings are named despite the paper asserting that it specifies them, and the rule itself never appears.
Significance 3. A correct symmetry design rule would generalise well, but this one is textbook where it is right and inverted where it is new, and its named class admits no solutions. Triangulene is not a viable emitter platform.