Family 052 — From a split tangent bundle to a product of manifolds

← All digests

Family 052 · Algebraic and complex geometry
Importance 4/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

On a product $Y_1\times Y_2$, a tangent vector naturally splits into a direction along the first factor and a direction along the second. Suppose a complex manifold has exactly such a splitting at every point. Must the manifold itself come from a product?

There are two obstacles. First, the chosen directions must fit together into leaves, like coordinate slices. Second, local slices must continue globally without twisting or breaking down. Passing to the universal cover removes loop monodromy, but does not automatically solve the continuation problem.

Family 052 contains two complementary claims. M1 starts with two integrable tangent summands on a compact Kähler manifold and claims a compatible product decomposition of its ordinary universal cover. M2 starts with a rationally connected projective manifold and claims that every two-summand holomorphic splitting is automatically integrable. An earlier theorem then supplies a product of the original manifold itself.

In both statements, compatibility means preserving the particular tangent subbundles supplied at the start. An unrelated abstract product decomposition would not answer the question.

2. Background and essential definitions

Splitting directions and integrating them

The holomorphic tangent bundle $T_X$ consists of the complex tangent spaces of $X$. A specified splitting

$$T_X=E_1\oplus E_2$$

means each $E_i$ is a holomorphic subbundle and every tangent vector has a unique decomposition into the two summands. Let their positive ranks be $r_1,r_2$, so $\dim_{\mathbb C}X=r_1+r_2$.

A subbundle $E$ is integrable if the Lie bracket of any two local vector fields tangent to $E$ is again tangent to $E$. The bracket measures the infinitesimal failure of successive motions to commute. Frobenius’ theorem says integrability is precisely the condition for local submanifolds tangent to the chosen directions to exist.

These local submanifolds fit together into leaves. If both complementary summands are integrable, local coordinates can be separated into two blocks, with one block varying along each leaf. Locally the manifold is a product. The global question remains.

Compatibility is stronger than being a product

An isomorphism $X\simeq Y_1\times Y_2$ is compatible with the specified splitting when it identifies $E_i$ with the tangent directions of $Y_i$:

$$E_i\simeq\operatorname{pr}_i^*T_{Y_i}$$

as the actual subbundles of the tangent bundle, not only as abstract vector bundles.

Even a manifold already known to be a product can carry a different, nonintegrable tangent splitting. The manuscript recalls an example on an abelian surface times $\mathbb P^1$: adding noncommuting vertical vector fields to two translation directions gives a complementary subbundle whose bracket leaves that subbundle. Product topology alone does not force the specified directions to be product directions.

Covering spaces and rational curves

The ordinary universal cover $\widetilde X\to X$ is the simply connected topological covering equipped with its lifted complex structure. A product upstairs need not descend to a product downstairs: deck transformations can act on both factors together. Even a finite cover need not remove all such coupling.

A smooth projective variety is rationally connected if two general points lie on a rational curve, the image of $\mathbb P^1$. “General” means pairs in a nonempty algebraic open subset of $X\times X$. This is stronger than uniruledness, which requires a rational curve through a general individual point.

Rational connectedness supplies many moving rational curves and, in the smooth complex projective setting, simple connectedness. M2 uses this abundance of curves to control the Lie brackets of the tangent summands.

3. Development history and previous partial results

The Riemannian analogue is de Rham’s decomposition theorem: parallel orthogonal distributions on a complete Riemannian manifold give a product on its universal cover. A holomorphic splitting is weaker. Compactness makes a chosen metric complete, but does not make its given tangent subbundles parallel.

Beauville formulated the compatible splitting problem and proved major cases, including Kähler–Einstein manifolds and compact Kähler surfaces. For multiple summands, the compatible formulation involves integrability of relevant partial sums. In the two-summand setting here, integrability of both summands is the required local hypothesis.

Druel treated projective manifolds whose tangent bundle is a sum of line bundles. Brunella–Pereira–Touzet established splitting when one summand is a line bundle and its complement is integrable. Further work of Pereira–Touzet and Druel–Pereira–Pym–Touzet supplies product decompositions under flatness assumptions or the existence of a compact leaf with finite holonomy.

On the projective side, Höring proved automatic integrability for non-uniruled manifolds with split tangent bundle. He also proved that on a rationally connected projective manifold, integrability of one summand already gives a compatible product. The remaining rationally connected problem was therefore to force integrability without extra hypotheses.

Campana–Peternell obtained important low-rank Fano cases. Höring’s 2026 Fano-type theorem establishes algebraic integrability in a broad singular setting with Fano-type positivity, and gives an actual compatible product for smooth Fano manifolds. M2 claims to remove that positivity restriction in the smooth rationally connected case, with arbitrary positive ranks.

The family thus addresses two different missing steps: M1 globalizes an already integrable local product in general compact Kähler geometry; M2 supplies integrability from rational connectedness.

4. What the new manuscripts claim

M1: a compatible product on the universal cover

Let $X$ be a compact connected Kähler manifold of dimension at least two, and assume both positive-rank summands in $T_X=E_1\oplus E_2$ are integrable. M1 claims connected simply connected complex manifolds $Y_1,Y_2$ and a biholomorphism

$$\Phi:\widetilde X\xrightarrow{\ \sim\ }Y_1\times Y_2$$

such that

$$d\Phi(\pi^*E_i)=\operatorname{pr}_i^*T_{Y_i}, \qquad \dim_{\mathbb C}Y_i=r_i.$$

Here $\pi$ is the covering map. The factors are the universal covers of the intrinsic leaves through a common point. They can be noncompact, and the leaves use their own manifold topology.

Combining this with Höring’s earlier integrability theorem gives the same universal-cover conclusion for every two-summand splitting on a smooth non-uniruled projective manifold.

When $K_X$ is nef and big, M1 also claims that the canonical Kähler–Einstein connection preserves the specified summands on the regular canonical locus. This is the inverse image of the smooth part of the canonical model under the pluricanonical contraction. Nefness means nonnegative degree on every curve; bigness means maximal growth of pluricanonical sections. The additional parallelism is more than the product conclusion already obtained from the non-uniruled case.

M2: rational connectedness forces integrability

For a smooth connected rationally connected projective complex manifold, M2 claims that both summands of every specified holomorphic splitting into two positive-rank subbundles are integrable, with no further rank or positivity condition.

Applying Höring’s established product theorem then yields

$$X\simeq X_1\times X_2, \qquad E_i=\operatorname{pr}_i^*T_{X_i},$$

where both factors are smooth projective manifolds. This is a product of $X$ itself, stronger than an upstairs statement for a general compact Kähler manifold.

M2 also applies the result to general fibers of the rationally connected quotient of a uniruled compact Kähler manifold. It supplies integrability of the induced summands there, allowing earlier structure theorems to produce almost-holomorphic fibrations in the corresponding rational directions. When one ambient summand has rank two, those directions have rank two, one, or zero; the resulting conclusions differ accordingly. This is not an assertion that every uniruled manifold with a splitting is a global product.

5. Why the advance matters

Importance: 4/5. If correct, the family would resolve the two-summand compatible splitting problem under its integrability hypothesis and a central automatic-integrability case beyond known positivity or rank restrictions.

Tangent bundles describe infinitesimal geometry. A theorem recovering a global product from their decomposition turns local linear information into a structural classification tool. Compatibility ensures that the factors explain the original geometry rather than merely provide another way to write the same space.

The complementary hypotheses are useful in different regions of the classification. Rationally connected varieties have many rational curves; non-uniruled varieties do not. M2 uses the former abundance, while M1 combines with earlier work in the latter setting. Neither manuscript erases the separate difficulties for general intermediate uniruled geometry.

6. The main idea

M1: continue rectangles, not just individual paths

With integrable complementary directions, a local map of two complex variables can move in $E_1$ along the first variable and in $E_2$ along the second. M1 calls such a map a holomorphic box. It need not be injective or have full rank.

The manuscript first extends these maps across Hartogs figures, domains in several complex variables where analytic continuation can fill a missing region. A general meromorphic extension could have indeterminacy; the separate tangent conditions are used to exclude the exceptional behavior and make the extension holomorphic.

It then crosses mixed boundaries using area estimates from a Kähler form. Closedness of that form permits Stokes’ theorem, while bounded projections onto the two summands control the relevant directions. No assumption that a metric already makes the summands parallel or orthogonal is introduced.

These extensions combine two disc axes into a whole box and preserve additional holomorphic parameters. Finite grids then combine continuous leafwise paths into rectangles. Continuing both the local product map and its inverse, and passing to simply connected covers, is intended to produce the global compatible biholomorphism.

M2: use rational curves to kill the bracket obstruction

M2 begins without integrability. Given a rational curve $g:\mathbb P^1\to X$, it claims a tangent-adapted map

$$F:\mathbb P^1\times\mathbb P^1\longrightarrow X, \qquad F(s,s)=g(s),$$

whose two factor directions map into $E_1$ and $E_2$. The construction starts formally near the diagonal, reconstructs a rational map using section counts, and removes indeterminacy through the tangent identities.

On this box, the pullback of each summand comes from its corresponding factor. Along a factor curve, the complementary summand is therefore trivial. This is the geometric feature needed to test integrability in arbitrary rank.

Failure of integrability is measured by the projected bracket tensor

$$\bigwedge^2E_i\longrightarrow E_j, \qquad j\ne i.$$

Rational connectedness supplies moving curves and sections vanishing at one marked point while spanning the desired tangent directions at another. Applying the bracket tensor gives a section of a trivial bundle on $\mathbb P^1$. Such a section is constant; because it vanishes at the first point, it vanishes at the second. The spanning property then forces the tensor to vanish on an open set, and holomorphicity extends the vanishing everywhere.

This separates the construction valid for split bundles from the final step where rational connectedness is essential.

7. Scope, limitations, and open questions

M1 assumes integrability. It does not claim that every holomorphic splitting on every compact Kähler manifold is integrable; the recalled product example shows why that would be false.

M1 concludes a product of the ordinary universal cover, not necessarily of $X$ or a finite cover. M2’s actual product conclusion uses smooth rational connectedness and the earlier product theorem. Singular rationally connected spaces are not covered, and the source cites singular examples where automatic integrability fails.

Both main statements are formulated for two summands. More elaborate splittings require attention to integrability of partial sums; they should not be treated as an automatic restatement with more symbols.

The claims do not classify the individual factors or guarantee compact leaves in M1. Understanding intermediate uniruled cases and the possible deck-group actions remains part of the wider structure problem.

8. Sources and further reading

M1’s Theorem 1.1 is the integrable universal-cover statement; its corollaries and Section 7 give the projective and canonical-metric applications. M2’s Theorem 1.1 is automatic integrability, Corollary 1.2 invokes Höring for the actual product, and Section 5 states the rational-quotient consequences.

The primary works of Beauville and Höring linked above explain the conjectures and earlier cases. Background on foliations and the Frobenius theorem helps interpret integrability as the passage from directions to local slices.

Source manuscripts

• M1. OpenAI, Universal-cover splitting for compact Kähler manifolds (September 23, 2026).
• M2. OpenAI, Integrability of split tangent bundles on rationally connected manifolds (September 23, 2026).

Sources consulted on 12 October 2026 at repository revision fd4aeeb. Download the manuscript citations.