Complex hyperbolic geometry, Hodge theory and arithmeticity

Special subvarieties of non-arithmetic ball quotients and Hodge theory

Gregorio Baldi and Emmanuel Ullmo

48 pages100 claims indexed82 referencesManuscript 2022-05-19

Checked local corrections and remaining obligations

These are coordinating source checks, not accepted native arbitration or a complete repair of the paper. Locations use the frozen arXiv v3/v4 body.

Unitary standard representations and adjoint traces

In Theorem 4.2.2 (totreal), unitarity gives $\bar\rho\simeq\rho^\vee$, not $\rho\simeq\rho^\vee$. For $A=\operatorname{diag}(i,i,-1)$, determinant one and preservation of $\operatorname{diag}(-1,1,1)$ hold, but the traces of A and its inverse are $-1+2i$ and $-1-2i$. The auxiliary implication is false.

There is a short corrected adjoint calculation. If A preserves any nondegenerate Hermitian form, $\operatorname{tr}(A^{-1})=\overline{\operatorname{tr}(A)}$. On $\mathfrak{sl}_{n+1}$,

$$\operatorname{tr}(\operatorname{Ad}A) =\operatorname{tr}(A)\operatorname{tr}(A^{-1})-1 =|\operatorname{tr}(A)|^2-1\in\mathbb R.$$

Assuming the preceding construction supplies a polarized conjugate local system for every coefficient-field embedding, the same calculation at every embedding proves total reality of the adjoint trace field. This repairs the last algebraic step conditionally on that construction. It does not produce a totally real coefficient field for the standard complex representation. The descent, polarization and integral-structure claims in thmzvhs still need their own argument; switching to an adjoint representation requires spelling out that construction. The report therefore does not certify the full integral-VHS chain.

Centralizers and the monodromy proof

Corollary 5.2.5 (cor123) uses finite centralizer as a consequence of rank one. For the standard block SU(1,1) in PU(1,2), the classes of $\operatorname{diag}(z,z,z^{-2})$, $|z|=1$, form a nonconstant circle in the centralizer. Their projective ratios include $z^3$. Hence that implication is false. A compact normal factor can sometimes fix the period point, preserving an orbit equality while changing the group; the printed group equality does not follow from this observation alone.

Theorem 5.1.4 (thmmonodromy) has three directly checkable issues. Distinct embeddings of Q(sqrt(2)) have the same image field. Equality of embedded fields is therefore not the contradiction asserted in the proof. Intersection with a pair of factors cannot replace projection: the diagonal SL2 in SL2^3 intersects SL2×SL2×{1} trivially, while its pair projection maps onto each factor. Finally, semisimplicity does not imply that each normal kernel is finite or the whole factor: SL2×{1} is a proper positive-dimensional normal subgroup of SL2×SL2. The dichotomy needs simplicity.

For the global W=S_Gamma case there is a standard conditional correction. If a subdirect product of the absolutely simple conjugate groups identifies two factors by an algebraic graph, their adjoint characters agree pointwise on the same monodromy elements. Thus $\sigma_i(\operatorname{trAd}\gamma)=\sigma_j(\operatorname{trAd}\gamma)$ for every gamma concerned. If those traces generate the stated field, the embeddings agree, a contradiction. A connected-closure or finite-index passage must preserve that trace-field hypothesis. For arbitrary W, the field in the source is instead defined using Gamma intersect M_W, not just the actual image of pi_1(W); factor simplicity and this trace-field identification remain open. This is a repair route under explicit conditions, not a proof of the full theorem.

Special-locus comparison and the withdrawn corollary

In Proposition 5.3.3 (mainthmproof), projecting the asserted period-domain preimage to X gives only $\widetilde\psi^{-1}(D_{\widehat M})\subseteq M.w$. The reverse inclusion requires control of the other period coordinates on the entire M orbit, beyond equivariance under Gamma and containment of the lifted W. Neither equality of dimensions nor notation for the generic Mumford–Tate group supplies that control. Replacing a restriction from K_W by one from K also changes the full group: for K_W=Q and quadratic K, the diagonal copy of M has half the dimension of Res_{K/Q}(M_K), despite possible related orbit statements. The general comparison remains unresolved.

The official erratum withdraws the proof of Corollary 1.3.3/5.5.4 (cor1), using an analogous period-preimage identity without the algebraic base needed for fixed part. That proof is incomplete and the authors leave its truth open. This is not a counterexample. A scan of native references to cor1 and cor1proof finds no use as a proof input in the main finiteness argument. The unrelated comparison defects above must not be inferred merely from the erratum.

Dimensions, induction and the product-covering issue

In Proposition 6.1.4 (mainpropwithas), put n=dim X, d=dim D, h=dim D_H, and u=dim U. The codimension inequality is exactly $u>n+h-d$. If u≥m=d_H1, it suffices to prove $\sum_{i>1}(d_i-d_{H_i})>0$. The displayed inequalities alone do not imply this: equality of the two dimensions at each other factor does not imply both are zero. For example, n=2, m=1, d_2=d_H2=1 gives equality with nonzero nondistinguished dimensions. This is a counterexample to the bare numerical inference, not a lattice counterexample to the desired strict inequality. For genuine proper subball block subgroups one expects a strict dimension drop at an extra noncompact factor; that geometric claim and the orbit conventions must be proved for every permitted parameter.

In the downward induction the condition is membership exclusion from the finite sets of larger maximal subvarieties. A smaller subvariety contained in one of them is automatically unequal to it, so it passes this exclusion. The needed condition excludes containment in every such larger subvariety. Replacing it requires a definable containment/germ argument, not just changing a word.

Theorem 5.4.1 (nonarithas7) calls the full period square Cartesian. With an arithmetic quotient A\D, its pullback to S_Gamma generally has sheets indexed by the monodromy subgroup in A, not just the chosen X lift. On that selected component the dimension shift is correct: adjoining D_>1 of dimension q changes W to dimension dim W+q and the ambient to 2n+q, so the expected intersection is dim W−n. A componentwise version preserves the numerical reduction, but it still depends on the unresolved special-locus comparison. The product-base use in the commensurator proof needs its own weakly-special classification and surjectivity arguments; a one-factor rank-one statement is not that proof.

Literal definitions and maximality

The printed weak Mumford–Tate-domain definition allows M={1}, normal in MT(x). It gives D_0={x}. Since the first period coordinate is the identity, the fiber of psi over that quotient point is locally discrete; its zero-dimensional components are points. Under these literal definitions every point of S_Gamma is Z-special. They cannot all be Gamma-special torus points: there are only countably many K-defined tori, and a compact maximal torus in PU(1,n) fixes one ball point, while the ball quotient has uncountably many points. Thus the special-point comparison cannot be used with this literal weak definition. The intended Hodge-special notion needs correction. This does not produce a positive-dimensional counterexample to the main finiteness theorem.

Theorem 6.0.1 (thm123) omits “maximal”; its next paragraph explains why that condition matters. We record the mismatch and use the intended maximal statement for the main verdict. We have not constructed a new nonarithmetic higher-dimensional example refuting every stronger literal formulation, nor inferred one from the typo.

The new Sol allegations about general reductive quotients, cusp coordinates, unique algebraic models, and other peripheral statements remain an explicit unadjudicated inventory. They are not additional confirmed defects in this report.

Source record

Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.

Cite this review

If this review or its supporting reports inform your work, please cite it.

Igor Rivin. Review of Special subvarieties of non-arithmetic ball quotients and Hodge theory. Math Referee, .

Download .bib
View BibTeX
@misc{rivin2026review-human-baldi-ullmo-special-subvarieties,
  author = {Rivin, Igor},
  title = {{Review of Special subvarieties of non-arithmetic ball quotients and Hodge theory}},
  year = {2026},
  month = {October},
  date = {2026-10-23},
  howpublished = {Math Referee},
  url = {https://mathreferee.dimensionreducers.ai/papers/human-baldi-ullmo-special-subvarieties/panel/},
  note = {Review report}
}