Version 0.2 · 5 October 2026, revised 11 November 2026 · for evaluation by V. Pinot

Affine orthogonal quiver varieties at δ and 2δ, and their contact links

Download

The zip is also attached inside each PDF (paperclip icon in Acrobat or Preview). LaTeX sources and every script are listed below.

What the note shows

Pinot's Theorem 1.2 lists the surfaces 𝔐δ and 𝔐2δ for ten families of affine orthogonal quivers. This note compares each surface with the classical quiver variety of the same graph, and lifts each one to its link, a contact 3-manifold S³/Γ.

A

At δ the involution acts trivially

Whenever 𝔐δ is a surface, it equals the classical Kleinian surface of the underlying affine graph. The proof uses only Li's closed immersion (Prop. 9.2.1), not the isomorphism stated in Pinot's Remark 2.15.

PROVEDassuming Li's result applies to the unframed case
B

The two D̃ rows are D₂ₙ₋₂ and D₂ₙ₋₁

Not Dn+1. For (D̃₂ₙ₋₁, a) the relation is z² = z(−x)ⁿ⁻¹ − xy²: the printed version has y in place of z in the first term. A direct numerical computation of the invariant ring confirms both rows for n = 3, 4, 5.

PROVEDCOMPUTEDnumerical for n = 3, 4, 5
C

At 2δ the groups double

In three rows Γ2δ contains the classical group with index 2 (ℤ₄ₙ₋₂ ⊃ ℤ₂ₙ₋₁, ℤ₄ₙ ⊃ ℤ₂ₙ, BD₄ₙ ⊃ ℤ₂ₙ), and in the fourth they are equal. Conjecture: each 2δ surface is the classical surface divided by an involution.

CONJECTUREthe statement for coordinate rings is proved
D

The contact links

The symmetric representation space is quaternionic, so the link is the Kempf–Ness / 3-Sasakian quotient of its unit sphere. Pinot's weights match the Klein invariants, so the Reeb flow is the Hopf flow: lens spaces over S²(m,m) (m odd) or S²(m/2,m/2) (m even), and prism manifolds over S²(2,2,k).

CITEDstandard contact-link facts, with the weight match checked by script. Not established: that the reduced contact form is the standard Hopf form with its normalized Reeb flow.

Pinot's table, with the corrections

QuiverSigndEquationTypeΓ
(Ã₂ₙ₋₂, v-a)+1δxy = z²ⁿ⁻¹A₂ₙ₋₂ℤ₂ₙ₋₁
(Ã₂ₙ₋₂, v-a)−12δxy = z⁴ⁿ⁻²A₄ₙ₋₃ℤ₄ₙ₋₂
(Ã₂ₙ₋₁, a-a)(+,+)δxy = z²ⁿA₂ₙ₋₁ℤ₂ₙ
(Ã₂ₙ₋₁, a-a)(+,−)2δxy = z⁴ⁿA₄ₙ₋₁ℤ₄ₙ
(Ã₂ₙ₋₁, a-a)(−,−)2δxy = z²ⁿA₂ₙ₋₁ℤ₂ₙ
(Ã₂ₙ₋₁, v-v)δxy = z²ⁿA₂ₙ₋₁ℤ₂ₙ
(Ã₂ₙ₋₁, c)2δxⁿ⁺¹ − y²x − z² = 0Dn+2BD₄ₙ
(D̃₂ₙ₋₂, v)δz² = −y(−x)ⁿ⁻¹ − xy²D₂ₙ₋₂Dn+1BD8(n−2)
(D̃₂ₙ₋₁, a)+1δz² = z(−x)ⁿ⁻¹ − xy²D₂ₙ₋₁Dn+1BD4(2n−3)

Highlighted rows are corrected; the struck-through label is the one printed in arXiv:2609.39434. All other rows are reduced points.

Questions for V. Pinot

  1. Do you confirm the labels D₂ₙ₋₂, D₂ₙ₋₁ and the relation z² = z(−x)ⁿ⁻¹ − xy²?
  2. Does Li's Proposition 9.2.1 apply to your unframed signed setting, for instance with w = 0 or an auxiliary framing vertex?
  3. Is there a source for the isomorphism of Remark 2.15 at ζ = 0 beyond type Aₙ with alternating signs?
  4. Are the 2δ surfaces quotients of the classical surface by an involution, and is there a natural candidate for it in terms of the quiver?
  5. Do you plan to treat the case of your Remark 1.3, (D̃₂ₙ₋₁, a) with sign −1 at 2δ?
OPEN

Run it yourself

Python 3 with numpy, scipy and sympy, and Lean 4 core (no Mathlib). The page's own check is index-verify.py: it runs in seconds and reads the saved output of the ten-minute numerical computation; add --full to recompute it.

python3 index-verify.py
python3 -m pip install numpy scipy sympy
python3 index-verify.py --full
lean scripts/DM3Quivers_Check.lean