Adding the complexity of a base and its fibers

← All digests

Family 033 · Algebraic and complex geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-12

1. The question

Suppose a geometric space is organized as a family of smaller spaces. A surface might be made of curves, one above each point of another curve. Can the complexity of the whole be estimated from the complexity of the base and a typical fiber?

Dimension itself adds: a family of one-dimensional curves over a one-dimensional base is two-dimensional. Kodaira dimension measures a subtler kind of complexity, using global differential forms. Iitaka’s conjecture asks whether this complexity is at least the sum of the contributions from base and fiber.

Family 033 contains five manuscripts. They claim ordinary, logarithmic, and orbifold lower bounds; a stronger bound that measures how much the fibers change; an equality under smoothness conditions; and a semiampleness theorem for the moduli part of a Kähler fibration. These are related questions, but the equality needs assumptions that the lower bounds do not.

2. Background and essential definitions

Counting differential forms

For a smooth complex variety $X$ of dimension $n$, the canonical line bundle $K_X$ has local sections of the form $f\,dz_1\wedge\cdots\wedge dz_n$. A section of $mK_X$ is an $m$-fold tensor power of such a top-degree form. Here additive notation for line bundles means tensor product.

The space $H^0(X,mK_X)$ consists of global holomorphic sections. Ratios of sections define a rational map to projective space. The largest image dimension among these maps is the Kodaira dimension $\kappa(X)$. If every positive-degree space is zero, its value is $-\infty$. Otherwise it lies between zero and $\dim X$. A variety is of general type when these two dimensions agree.

For smooth projective curves, genus zero gives $-\infty$, genus one gives zero, and genus at least two gives one. These examples show that Kodaira dimension measures more than ordinary dimension: all three kinds are curves, but their canonical forms behave differently.

More generally, $\kappa(X,L)$ denotes the same construction for a line bundle or rational divisor $L$. A rational divisor has rational coefficients; sufficiently divisible multiples have integer coefficients. We use $\kappa(\mathrm{point})=0$ and $-\infty+a=-\infty$.

Fibrations and boundaries

A fibration here is a proper surjective map $f:X\to Y$ with connected fibers. A general fiber $F$ avoids the exceptional parameters where behavior changes. The geometric generic fiber is the corresponding object over an algebraic closure of the function field of $Y$.

A boundary $\Delta=\sum_i\delta_iD_i$ records specified hypersurfaces with weights $0\leq\delta_i\leq1$. Simple normal crossings, abbreviated SNC, means they look locally like distinct coordinate hyperplanes. A reduced boundary has all nonzero coefficients equal to one.

If $U=X\setminus D$, with $X$ smooth projective and $D$ reduced SNC, define

$$\overline\kappa(U)=\kappa(X,K_X+D).$$

This is logarithmic Kodaira dimension. It permits controlled poles along the missing boundary and is independent of the chosen smooth compactification. For example, $\mathbb C^*$ has logarithmic Kodaira dimension zero, although its compactification $\mathbb P^1$ has ordinary Kodaira dimension $-\infty$.

A compact Kähler manifold has a positive closed real differential form of type $(1,1)$, giving a metric compatible with its complex structure. Projective manifolds are Kähler, but the converse fails. Fujiki class $\mathcal C$ consists of spaces bimeromorphic to compact Kähler spaces. Bimeromorphic maps are analytic analogues of birational maps: they become isomorphisms on dense open subsets.

Variation and positivity

The variation $\operatorname{Var}(f)$ measures how many parameters are needed to define the whole generic fiber up to birational equivalence. Zero variation means that, after an appropriate field extension, its birational type is constant. This invariant must retain the entire function field of the fiber, not merely a canonical model or one associated period map.

A line bundle is semiample if one positive tensor power is generated by global sections everywhere. Such sections define a morphism. Nefness is a weaker positivity condition: in the projective setting it says that every curve has nonnegative degree. Nefness by itself does not guarantee global sections.

3. Development history and previous partial results

Iitaka’s theory of dimensions of divisors put the addition question into a birational framework. The target inequality for a smooth projective fibration is

$$\kappa(X)\geq\kappa(F)+\kappa(Y).$$

The key difficulty is global: forms available separately on fibers do not automatically fit together into enough forms on the total space.

Kawamata established major cases involving curve bases and fibers with good minimal models. Viehweg developed weak positivity of direct images of pluricanonical bundles, and Kollár treated general-type fibers. These results draw positivity from different places. A good minimal model is a birational model whose canonical divisor is semiample, while a general-type base provides enough canonical positivity to absorb auxiliary twists. M1 §1 and M4 §1 give the original references and distinguish these hypotheses.

Birkar proved the ordinary conjecture through total dimension six. Other work treats special bases, including abelian varieties and varieties of maximal Albanese dimension, which map generically finitely to their Albanese tori. Logarithmic results of Kovács–Patakfalvi and Hashizume allow boundaries under specified positivity or abundance conditions on the fibers.

Popa’s conjectural framework separates variation lower bounds from stronger additivity questions for smooth families. Earlier results on the reverse inequality required hypotheses such as good models for fibers of the base’s own Iitaka fibration. Those are not the same as assumptions on the original fibers of $f$.

The moduli-part question has its own history. Ambro developed its birational formulation; Fujino–Gongyo proved important positivity properties. M5 credits the algebraic b-semiampleness theorem of Bakker–Filipazzi–Mauri–Tsimerman and addresses a compact Kähler extension allowing nonprojective morphisms. It should not be presented as the first projective b-semiampleness theorem.

The introductions also acknowledge earlier broad preprint claims of Tsuji and Maehara. They are not used as inputs in these manuscripts. The exposition here follows the current statements rather than making an independent priority assessment.

4. What the new manuscripts claim

M1 and M4: lower bounds

M4 asserts ordinary subadditivity for every surjective projective morphism with connected fibers between smooth connected projective varieties over an algebraically closed field of characteristic zero. There is no dimension bound or good-minimal-model assumption on the generic fiber.

M1 gives the broader rational-SNC orbifold statement in Fujiki class $\mathcal C$:

$$\kappa(X,K_X+\Delta) \geq\kappa(F,K_F+\Delta|_F)+\kappa(f\mid\Delta).$$

The last term is Campana’s birationally defined orbifold-base Kodaira dimension. Multiple fibers and source boundary coefficients contribute to it. For a base divisor $D$, its multiplicity is computed from

$$m(f,\Delta;D)=\min_{E\to D} \frac{\operatorname{ord}_E(f^*D)}{1-\Delta_E},$$

where $E$ runs over divisors dominating $D$, and division by zero means infinity. The base coefficient is $1-1/m(f,\Delta;D)$. Taking the specified birational invariant is part of the theorem; one arbitrary base model is not the whole definition.

Coefficient-one boundaries are allowed. Ordinary and logarithmic subadditivity follow as stated consequences, including the characteristic-zero projective formulations.

M2: changing fibers contribute

For a projective surjection $f:U\to V$ with connected fibers between smooth complex quasi-projective varieties, M2 assumes $\overline\kappa(V)\geq0$ and claims

$$\overline\kappa(U)\geq\kappa(F)+ \max\{\overline\kappa(V),\operatorname{Var}(f)\}.$$

Singular fibers are permitted. There is no abundance or semiampleness assumption on $F$. The variation is that of the whole geometric generic fiber. When variation exceeds the logarithmic complexity of the base, this improves the ordinary lower bound.

M3: equality under stratum smoothness

Let $f:(X,E)\to(Y,D)$ be a connected-fiber surjection of smooth projective complex reduced-SNC pairs, with $\operatorname{Supp}(f^*D)\subseteq\operatorname{Supp}(E)$. Assume the total space and every boundary stratum are smooth over $Y\setminus D$. A stratum is an irreducible component of an intersection of boundary components.

M3 claims the reverse inequality. Combining it with M1 gives

$$\kappa(X,K_X+E)=\kappa(Y,K_Y+D)+\kappa(F,K_F+E|_F).$$

Horizontal boundary components are allowed. Both $-\infty$ cases are included: if either contribution on the right is $-\infty$, all positive log-pluricanonical section spaces upstairs vanish. This is a substantive assertion, whereas a lower bound by $-\infty$ would be automatic.

M5: a moduli line defines a map

Suppose $f:Y\to X$ is a connected-fiber surjection between smooth compact connected Kähler manifolds, with effective rational SNC boundary $\Delta$ of coefficients in $[0,1]$, and

$$K_Y+\Delta\sim_{\mathbb Q}f^*L.$$

Rational linear equivalence here means equality of the corresponding actual holomorphic line bundles after a common positive multiple. The canonical bundle formula separates $L$ into a canonical part, a discriminant measuring singularities, and a moduli line.

M5 claims that on some smooth compact Kähler modification $S\to X$, the moduli line $M_S$ is semiample, and on every higher smooth modification $\nu:S_1\to S$ one has $M_{S_1}=\nu^*M_S$ as rational holomorphic lines. This is b-semiampleness. It allows coefficient-one horizontal boundary and does not assume projectivity or orbifold Iitaka subadditivity.

5. Why the advance matters

Importance: 5/5, provisional. Iitaka subadditivity is a central structural question about the birational classification of varieties. The claimed inequalities cover arbitrary dimension and connect canonical forms, family variation, and analytic positivity.

The equality statement adds information that lower bounds cannot provide, especially when the base has no logarithmic pluricanonical forms. The Kähler moduli theorem would make a numerical-looking contribution into an actual map, allowing geometric constructions that require sections rather than just positivity.

These results also serve as inputs elsewhere in the collection. Keeping their assumptions visible is essential when applying them to abundance or minimal-model claims.

6. The main idea

Replace the fiber by the part seen by its forms

A relative Iitaka map separates a fibration into a part carrying the fiber’s pluricanonical image and a remaining part of Kodaira dimension zero. The latter has essentially one pluricanonical generator in suitable degrees. Taking a cyclic root of that generator produces a differential form on a cover.

As the base varies, this form spans a Hodge line, a line inside a family of cohomology spaces. Its positivity measures some of the variation. The canonical bundle formula packages singular-fiber effects and this moduli contribution while comparing actual section spaces.

M1 and M4 develop the needed positivity on a projective period image and control boundary and ramification terms. Their addition arguments combine sections with auxiliary positive and negative twists so that the twists cancel. This exact comparison is what returns the lower bound to the original canonical divisor.

Keep all parameters of the fiber

M2 needs more than a period map which detects only part of the geometry. It first identifies a parameter field from logarithmic forms. A flat highest Hodge line is used to obtain birational constancy of the Kodaira-zero part. A separate descent argument retains the coordinates of the relative Iitaka base and finite normalization data.

Together these steps claim that the entire fiber is defined over the controlled parameter field. Bounding its transcendence degree then bounds $\operatorname{Var}(f)$. Forgetting the second step would control a simpler associated object instead of the variation appearing in the theorem.

Use a section’s zero to control the upper bound

For M3, the crucial case is a base of logarithmic Kodaira dimension zero. The paper constructs a base pluriform from a total-space pluriform, retaining a zero on a specified smooth fiber. A suitably chosen general base point avoids the zero sets of all nonzero base pluriforms. Restriction to that fiber must therefore be injective on total-space sections, giving the upper growth bound.

For M5, residue constructions handle coefficient-one boundaries. An exact order calculation identifies the extended Hodge line with the threshold-defined moduli line. Product decompositions, period compactifications, and descent supply global generation. Tracking the actual line prevents an undetected flat twist from changing its section spaces.

7. Scope, limitations, and open questions

The lower inequalities and the equality are not interchangeable. M2 requires nonnegative logarithmic Kodaira dimension of the base. M3 requires smoothness of every boundary stratum over the open base. Removing these assumptions is not part of their statements.

The discriminant used for M5’s moduli line is defined by log canonical thresholds, which measure how much a fiber divisor can be added before singularities cross the allowed limit. It is distinct from M1’s inf-multiplicity orbifold divisor.

The ordinary algebraic theorem is in characteristic zero. The analytic results concern compact Kähler or class-$\mathcal C$ spaces under the stated conditions, not arbitrary compact complex manifolds. M5 supplies some globally generating multiple, without a uniform numerical bound for that multiple across all fibrations.

Natural further questions concern effective bounds, positive-characteristic analogues, and extensions beyond the specified singularity and compactness settings. The manuscripts’ claims and dependencies are reported here without an independent mathematical verification.

8. Sources and further reading

M1 §1 specifies the orbifold invariant and the lower inequality. M2 §1 defines whole-fiber variation and identifies its two constancy steps. M3 §1 states the stratum-smooth upper bound and explains the negative-infinity cases. M4 §§1 and 6–7 give the projective Hodge-line route. M5 §§1 and 4 distinguish the threshold moduli line from numerical positivity.

The linked work of Birkar and Popa supplies accessible entry points into the earlier partial results and conjectural landscape. For terminology, see the optional overview of Kodaira dimension.

Source manuscripts

• M1. OpenAI, Orbifold and logarithmic Iitaka subadditivity (September 26, 2026).
• M2. OpenAI, Logarithmic Kodaira dimension and whole-fiber variation (September 26, 2026).
• M3. OpenAI, The reverse logarithmic Kodaira inequality and additivity (September 26, 2026).
• M4. OpenAI, Projective Hodge lines and ordinary Iitaka subadditivity (September 27, 2026).
• M5. OpenAI, B-semiampleness for compact log-smooth Kähler fibrations (September 10, 2026).

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