From nef canonical bundles to maps: abundance and uniform degrees

← All digests

Family 034 · Algebraic and complex geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

A line bundle can look nonnegative when measured on every curve yet have too few global sections to describe a map. The abundance conjecture predicts that canonical line bundles, which encode top-degree differential forms, are special: their numerical nonnegativity should produce an actual geometric fibration.

Family 034 combines fourteen manuscripts addressing this prediction and its quantitative refinements. There are three successive questions. Does a positive power have even one nonzero section? Does some power have sections generating the line at every point? Can one choose a degree depending only on dimension and specified boundary coefficients?

The answers claimed here range from projective varieties to compact Kähler spaces, which need not be projective. The leading Kähler theorem explicitly assumes logarithmic Iitaka subadditivity. A fourfold minimal-model paper is also expressly conditional. These dependencies are part of the mathematical statements, even though related manuscripts in the collection claim the needed inputs.

2. Background and essential definitions

Divisors, singularities, and adjoints

The canonical bundle $K_X$ of a smooth complex $n$-fold consists of holomorphic top-degree forms. A divisor is a formal sum of hypersurfaces; a rational divisor permits rational coefficients. A pair $(X,\Delta)$ adds an effective boundary $\Delta$, and its adjoint is $D=K_X+\Delta$.

On a singular space, a divisor need not itself define a line bundle. It is Cartier when it does so locally by one equation, and $\mathbb Q$-Cartier when some positive integer multiple is Cartier. The manuscripts concern actual multiples and their section spaces, not only their intersection classes.

The singularity conditions can be expressed on a resolution $\pi:Y\to X$: write

$$K_Y+B_Y=\pi^*(K_X+\Delta).$$

The pair is log canonical, or lc, when all coefficients encountered in these crepant boundaries are at most one; it is Kawamata log terminal, or klt, when they are strictly less than one. The latter is more restrictive. Divisorial log terminal, or dlt, is an intermediate controlled setting in which the coefficient-one boundary supports inductive arguments. Semi-log-canonical, or slc, also permits components joined along double loci, with lc behavior after normalization and inclusion of the joining divisor.

Three levels of positivity

For a projective variety, $D$ is nef if $D\cdot C\geq0$ for every curve $C$. It is pseudoeffective if its numerical class is a limit of effective divisor classes. On a compact Kähler space these notions use curvature classes and positive currents, generalized differential forms which can concentrate along subvarieties.

Nonvanishing means $H^0(X,mD)\ne0$ for some positive Cartier multiple. Semiampleness means that some such multiple is generated by global sections everywhere. One section can vanish along a large divisor; global generation requires, at each point, a section which does not vanish there.

Kodaira or Iitaka dimension $\kappa(X,D)$ is the largest dimension of a rational image defined by ratios of sections of multiples of $D$, with value $-\infty$ when there are no sections. For a nef divisor, numerical dimension $\nu(D)$ instead measures nonzero intersection powers. These two dimensions need not agree merely by definition. Abundance is precisely a bridge from numerical positivity to section geometry.

A minimal model is a birational model with nef adjoint and the required singularity comparison. A good minimal model has semiample adjoint. Birational means that dense open subsets are isomorphic. The minimal model program simplifies a variety through contractions and flips; finding a nef endpoint and proving it good are separate tasks.

Uniformity and the full Iitaka field

If $D\sim_{\mathbb Q}0$, some multiple defines the trivial line bundle. Its least positive trivializing degree is the index. Uniform index bounds must work across all pairs in a stated class.

For a rational divisor, define

$$V_m(D)=H^0\bigl(X,\mathcal O_X(\lfloor mD\rfloor)\bigr), \qquad K(D)=k\bigl(s/t:\ s,t\in V_m(D)\setminus\{0\},\ m>0\bigr).$$

The floor rounds each coefficient down; on a normal variety the indicated divisorial sheaf is interpreted reflexively when it is not a line bundle. The field $K(D)$ is the Iitaka field inside the function field $k(X)$. Recovering it in one degree is stronger than obtaining an image of the correct dimension: a proper finite subfield has the same transcendence degree.

3. Development history and previous partial results

The threefold story already separated the main difficulties. Miyaoka treated numerical dimension one, Kawamata completed canonical abundance for minimal threefolds, and Keel–Matsuki–McKernan established the logarithmic result. Fujino’s slc work addressed the additional problem of gluing sections across components. M4 §1 gives these original references and explains how their boundary methods enter its induction.

Birkar–Cascini–Hacon–McKernan established fundamental minimal-model and finite-generation results. These do not make every nef lc adjoint semiample automatically. In particular, finite generation alone does not supply a nonzero section when all positive section spaces vanish.

The Kähler theory required analytic replacements for projective tools. Höring–Peternell constructed threefold minimal models, Campana–Höring–Peternell treated canonical abundance, and Das–Ou proved log abundance for compact Kähler lc threefolds. M1 and M14 carefully distinguish these lower-dimensional inputs from the new higher-dimensional claims.

Uniform effectivity developed along another route. Fujino–Mori’s canonical bundle formula, Viehweg–Zhang’s results for surface bases, and Birkar–Zhang’s effective Iitaka theorem controlled degrees using additional fiber invariants. The new uniform claims seek to remove those extra dependencies under their specified coefficient assumptions.

For indices, Jiang–Liu established slc bounds through dimension three and important non-klt fourfold cases. M2 extends the dimension range, with substantial companion inputs. M8 also explicitly credits Zou’s theorem for the same zero-Kodaira-fiber conclusion that it proves by a different route. Its contribution should not be described as first establishing that particular statement.

4. What the new manuscripts claim

Abundance and its stated inputs: M1, M3, M4, M14

M1 assumes logarithmic Iitaka subadditivity for smooth projective complex fibrations with compatible reduced simple-normal-crossing boundaries. Under that assumption, every analytically nef $\mathbb Q$-Cartier adjoint of a normal irreducible compact Kähler lc pair with effective rational boundary is semiample, in every dimension. The conclusion concerns the given holomorphic line, with no unrecorded flat twist.

M4 claims projective rational-boundary log abundance over every algebraically closed characteristic-zero field. Over $\mathbb C$, it also claims canonical nonvanishing for smooth projective varieties and a stronger good-model induction. It states its logarithmic-subadditivity input explicitly; its sole use in the nonvanishing argument is an ordinary Albanese-fibration case, which the paper notes is also covered by earlier external theorems. The collection’s corresponding subadditivity claims are explained in family 033.

M3 concerns compact Kähler klt fourfolds with pseudoeffective actual adjoint and global strong $\mathbb Q$-factoriality: every rank-one reflexive sheaf on the whole space has an invertible positive reflexive power. This is not simply a local condition on each open chart. Its good-model conclusion assumes the stated orbifold Iitaka inequality, the specified ordinary fourfold program, and abundance after nonvanishing. The manuscript identifies the program as an independent companion input; its remaining conditional premises are stated rather than suppressed.

M14 supplies abundance after nonvanishing for normal compact Kähler klt fourfolds with effective rational boundary. It assumes analytic nefness and $\kappa(X,D)\geq0$, and requires neither projectivity nor $\mathbb Q$-factoriality. When $\kappa(X,D)=0$, it claims an actual trivial positive Cartier multiple. It does not itself manufacture the initial section required by its hypothesis.

Obtaining and extending sections: M5, M6, M7, M8

M5 claims that a minimal-singularity semipositive metric on the pullback of a nef projective klt adjoint has zero Lelong numbers everywhere on every projective log resolution. Lelong numbers measure logarithmic concentration of the metric’s singularities; zero does not mean the metric is smooth. A separate injectivity theorem in degree-one cohomology, for strictly interior rational SNC boundaries and zero-Lelong metrics at both endpoints, yields extension of sections from an entire divisor scheme.

M6 claims canonical nonvanishing for every smooth projective complex fourfold with pseudoeffective canonical divisor, without assumptions on numerical dimension, irregularity, or Euler characteristic. It derives nonvanishing for nef rational lc adjoints through dimension four. The result is qualitative: it does not bound the first degree.

M7 starts with a nonzero effective Cartier divisor $G\sim q(K_V+C)$ on a projective $\mathbb Q$-factorial dlt pair, supported on the coefficient-one boundary. If its line bundle is semiample on the whole reduced support of $G$, it claims positive ambient Iitaka dimension. It derives projective lc abundance after nonvanishing through dimension four, and an all-dimensional reduction assuming smooth canonical nonvanishing. Agreement across intersections is part of the support hypothesis.

M8, using the good-model input from M4, claims the Campana–Peternell inequality

$$m_0K_X-D\text{ pseudoeffective},\quad D\geq0 \quad\Longrightarrow\quad \kappa(X)\geq\kappa(X,D)$$

for a smooth projective complex variety and effective Cartier divisor $D$. For a connected-fiber fibration $f:X\to Y$, an ample Cartier $H$ on $Y$, and pseudoeffective $m_0K_X-f^*H$, it obtains $\kappa(X)=\kappa(F)+\dim Y$, plus sections of $mK_X-f^*H$ in all sufficiently large divisible degrees. The ample comparison is an essential hypothesis.

Uniform indices and uniform maps: M2, M9–M13

M10 claims a degree $m(d)$, depending only on dimension, whose complete pluricanonical system recovers the full Iitaka field of every smooth projective characteristic-zero $d$-fold with nonnegative Kodaira dimension. It also claims a uniform normal lc log-Calabi–Yau index for coefficients in a fixed rational DCC set. DCC means there is no infinite strictly decreasing sequence. Its good-model input is M4, not merely the established big-adjoint theory.

M2 extends uniform indices to connected projective equidimensional slc log-Calabi–Yau pairs in every fixed dimension at least four, with a fixed finite rational coefficient set. One integer makes the adjoint Cartier and trivial, independently of the number of components. Together with earlier dimensions at most three, this is the finite-coefficient slc claim. M2 uses the normal index theorem from M10 and good models from M4.

M13 gives uniform complete rounded log-Iitaka systems for projective complex lc fourfolds with fixed finite rational boundary coefficients and pseudoeffective adjoint. It also claims a uniform boundary-free klt fourfold canonical index. Its passage from pseudoeffectivity to a first section uses M4.

M9 gives the analogous full-field conclusion in dimensions at least five over algebraically closed characteristic-zero fields, assuming nonnegative Iitaka dimension and fixed finite rational coefficients. Its supporting result controls moduli denominators, including horizontal boundary, using M4, M10, and M12.

M11 treats log-Calabi–Yau fibrations with lc total space of dimension at most four and positive-dimensional base. For fixed finite rational coefficients, it selects an exact canonical-bundle-formula presentation with uniformly bounded trivializing degree and moduli Cartier denominator. If the base adjoint is big, uniform rounded systems recover its full function field. The denominator bound concerns this normalized presentation, not every rationally equivalent representative.

M12 supplies an arithmetic bound. For an ordinary projective lc log-Calabi–Yau pair $(X,B)$ of dimension $d$ over any characteristic-zero field $k$, with $H^0(X,\mathcal O_X)=k$, a prime boundary component $S$ of coefficient at least $t>0$ satisfies

$$[k_S:k]\leq N(d,t),$$

where $k_S$ is the relative algebraic closure of $k$ in $k(S)$. This controls the finite arithmetic constant field encountered when restricting to boundary components. It is the contraction-to-a-point form stated in that paper, not an unqualified theorem about all fibrations.

5. Why the advance matters

Importance: 5/5, provisional. Abundance is a central missing bridge in birational classification: a minimal model should reveal a canonical map, rather than merely have nonnegative curve intersections. The all-dimensional projective and conditional Kähler statements are therefore exceptionally consequential if correct.

Uniform degrees would strengthen an asymptotic description into one degree shared across an entire class. Uniform indices also matter when varieties degenerate into several components, because local trivializations must agree globally. The rating reflects the family as a whole; it does not imply equal novelty for each supporting manuscript.

6. The main idea

Move information off the boundary

The abundance arguments repeatedly restrict an adjoint to lower-dimensional boundary components, where induction supplies sections. Those sections must first be glued on the whole boundary. Finite covers and Hodge-theoretic residue calculations then control obstructions to extending them through thicker and thicker neighborhoods.

The intended geometric outcome is a moving family of subvarieties on which the adjoint is trivial. Such families constrain the directions where a nef divisor can remain positive. The papers use that constraint to obtain a fibration and ultimately generation of the actual adjoint line.

Force the first section

Nonvanishing cannot be assumed while proving it. M4 and M6 instead analyze a hypothetical variety with no canonical sections. Reductions constrain its possible fibrations and moving curves. Estimates for jets, which record successive derivatives of sections at a point, then produce incompatible bounds on vanishing orders.

One comparison uses reduction of fixed algebraic data to positive characteristic and Frobenius, the map raising coordinates to the characteristic power. The resulting rank deficit along a diagonal forces more determinant vanishing than the geometric upper bound permits. This is a proof strategy reported from the manuscripts, not a claim that the final theorem is in positive characteristic.

Bound the obstruction to a uniform degree

For uniformity, the obstacle is often a finite character: continuation around a singular fiber multiplies a volume form by a root of unity. The order of that scalar affects denominators and trivializing degrees. The papers aim to bound the relevant character even when a larger covering degree is uncontrolled.

On a reducible slc space, one can picture normalization components as vertices of a graph and their intersections as edges. Trivializations differ by scalars along edges. Products around closed walks obstruct gluing. M2 interprets these products through birational actions on smaller strata and bounds their orders uniformly. Taking one common power then makes the forms agree.

7. Scope, limitations, and open questions

The analytic and algebraic statements have different hypotheses. M1 remains an implication from logarithmic subadditivity; M3 retains its explicit program and abundance premises; M14 assumes nonvanishing. None should be shortened to an assumption-free theorem for arbitrary compact complex spaces.

Numerical equivalence, rational linear equivalence, and global generation are different. A flat twist can preserve a curvature class while changing sections. The conclusions here repeatedly require the actual line bundle. Likewise, a uniform Cartier denominator for a moduli divisor is not a uniform base-point-free multiple.

Finite rational coefficient sets in M2, M9, M11, and M13 must not be silently enlarged to arbitrary real boundaries. M10’s normal index statement has its separately stated rational DCC scope. The uniform bounds are existential; practical formulas or small numerical degrees are not supplied by the headline claims.

Effective values, broader coefficient classes, and positive-characteristic analogues remain further directions. All mathematical claims and dependencies are taken as stated; these expositions do not audit their proofs.

8. Sources and further reading

For the central logical distinctions, read M1 §1, M3 §§1–2, M4 §1, and M14 §1. M5–M7 explain metrics, first sections, and supported-boundary extension. M8 states its numerical comparison and credits the earlier zero-fiber theorem. M9–M13 specify the uniformity parameters and the equality of embedded section fields; M2 explains the extra gluing problem for slc pairs.

The linked BCHM and Jiang–Liu papers supply primary context for model existence and index bounds. The optional minimal model program overview introduces the birational operations used throughout.

Source manuscripts

• M1. OpenAI, Log abundance for compact Kähler spaces under logarithmic Iitaka subadditivity (October 6, 2026).
• M2. OpenAI, Uniform indices for semi-log-canonical log Calabi–Yau pairs (October 5, 2026).
• M3. OpenAI, Conditional good minimal models for compact Kähler fourfolds (October 6, 2026).
• M4. OpenAI, Log abundance in characteristic zero (September 24, 2026).
• M5. OpenAI, Minimal metrics and interior injectivity for nef adjoints (September 27, 2026).
• M6. OpenAI, Fourfold nonvanishing by minimal metrics and moving jets (September 27, 2026).
• M7. OpenAI, Lifting sections from the reduced support of an adjoint (September 27, 2026).
• M8. OpenAI, Schnell fiber spaces and good canonical models (September 24, 2026).
• M9. OpenAI, Uniform log Iitaka fibrations and bounded moduli denominators (October 4, 2026).
• M10. OpenAI, Uniform Pluricanonical Iitaka Fibrations (October 3, 2026).
• M11. OpenAI, Relative denominators and effective systems for log Calabi-Yau fibrations (September 27, 2026).
• M12. OpenAI, Arithmetic Stein-degree bounds for log Calabi–Yau pairs (September 25, 2026).
• M13. OpenAI, Uniform effective log Iitaka fibrations for fourfolds (September 26, 2026).
• M14. OpenAI, Abundance after nonvanishing for compact Kähler fourfolds (September 27, 2026).

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