Base Point Free Theorem in Algebraic Geometry

• The Base Point Free Theorem is a fundamental result stating that a nef divisor or class meeting specific positivity conditions is semi-ample.
• It applies across various settings—from projective klt pairs and Kähler spaces to quasi-log schemes and foliated structures—bridging algebraic and transcendental approaches.
• The theorem translates numerical positivity into concrete geometric constructions, facilitating contraction morphisms and advancing minimal model program techniques.

Searching arXiv for recent and foundational papers on the base point free theorem and related variants. The Base Point Free Theorem is a semiampleness statement: under suitable positivity hypotheses, a nef divisor, line bundle, or adjoint class admits enough global sections to define a morphism. In the classical algebraic form, if $(X,\Delta)$ is a projective klt pair and $D$ is a nef Cartier divisor such that $D-(K_X+\Delta)$ is nef and big, then $D$ is semi-ample; equivalently, some multiple $|mD|$ is base-point free (Cascini et al., 3 Sep 2025). In contemporary literature the same paradigm appears in transcendental Kähler geometry, quasi-log geometry, foliated birational geometry, and positive or mixed characteristic, with the conclusion reformulated as global generation of a multiple, existence of a contraction, or descent of an adjoint $(1,1)$-class to a Kähler class on a target space (Höring, 2018).

1. Classical algebraic form and geometric meaning

In the projective setting, the theorem is formulated for a nef divisor $D$ relative to an adjoint datum. The standard statement recorded in the recent foliated comparison is the Kawamata–Shokurov Base-Point-Free Theorem: if $(X,\Delta)$ is a projective klt pair and $D$ is a nef Cartier divisor such that $D-(K_X+\Delta)$ is nef and big, then $D$0 is semi-ample (Cascini et al., 3 Sep 2025). Here “semi-ample” means that there exists an integer $D$1 such that $D$2 is base-point free, or equivalently that $D$3 is globally generated.

The positivity conditions used in the theorem are standard. A line bundle or Cartier divisor is nef if it pairs nonnegatively with every curve, and big if its numerical class lies in the interior of the pseudo-effective cone, or equivalently if some multiple defines a birational map in the projective setting. The conclusion produces a morphism

$D$4

so the theorem converts numerical positivity into a contraction or fibration structure.

A basic special case is the canonical bundle. For a smooth projective manifold $D$5, if $D$6 is nef and big, then there exists $D$7 such that $D$8 is base-point free for all $D$9 (Song, 2014). This is the form studied by Song, who gave an analytic proof via the Kähler–Ricci flow, degeneration theory, and $D-(K_X+\Delta)$0-methods rather than algebraic vanishing theorems.

2. Adjoint $D-(K_X+\Delta)$1-classes on Kähler threefolds

A major transcendental extension replaces divisors by Bott–Chern $D-(K_X+\Delta)$2-classes. For a normal compact complex space in the Fujiki class, one considers

$D-(K_X+\Delta)$3

together with the usual positivity notions: Kähler classes, Kähler currents, big classes, modified Kähler classes, and nef classes (Höring, 2018). If $D-(K_X+\Delta)$4 is nef and big, its Null locus is

$D-(K_X+\Delta)$5

Höring proved a transcendental base-point-free theorem for adjoint classes on Kähler threefolds. Let $D-(K_X+\Delta)$6 be a normal, $D-(K_X+\Delta)$7-factorial compact Kähler threefold with only terminal singularities, let $D-(K_X+\Delta)$8 be a Kähler class, and set

$D-(K_X+\Delta)$9

If $D$0 is nef and big, then there exist a morphism $D$1 and a Kähler class $D$2, where $D$3 is a normal compact Kähler space with isolated rational singularities, such that

$D$4

and $D$5 has connected fibres (Höring, 2018). In this setting “semi-ample” no longer means generation of a linear system; it means that the transcendental class is the pull-back of a Kähler class on a lower-dimensional Kähler space.

The proof runs through an $D$6-trivial $D$7-MMP in the Kähler category, followed by a geometric analysis of the Null locus. After passing to a model $D$8, one shows that $D$9 has pure dimension one. A contraction theorem for a one-dimensional Null locus then produces a bimeromorphic morphism $|mD|$0 contracting each connected component of $|mD|$1 to a point. Finally, one descends the class and proves that the descended class on $|mD|$2 is Kähler by a positivity criterion on spaces with isolated rational singularities (Höring, 2018).

This transcendental theorem recovers the algebraic base-point-free theorem for $|mD|$3-divisors when $|mD|$4. In the Calabi–Yau case $|mD|$5, the argument uses the Beauville–Bogomolov–Campana–Höring–Peternell decomposition and the surface result of Filip–Tosatti. The higher-dimensional Calabi–Yau and hyperkähler cases remain open, while dimension two is settled by Filip–Tosatti (Höring, 2018).

3. Quasi-log schemes and singular adjunction frameworks

A different generalization is Fujino’s quasi-log formalism. A quasi-log scheme

$|mD|$6

is built from a proper morphism from a globally embedded simple normal crossing pair $|mD|$7 satisfying $|mD|$8, together with a non-qlc locus and a collection of qlc strata (Fujino, 2014). For a proper morphism $|mD|$9, an $(1,1)$0-Cartier divisor $(1,1)$1 is nef and log big over $(1,1)$2 with respect to $(1,1)$3 if $(1,1)$4 is $(1,1)$5-nef and $(1,1)$6 is $(1,1)$7-big for every qlc stratum $(1,1)$8.

The basepoint-free theorem of Reid–Fukuda type for quasi-log schemes states that if $(1,1)$9 is a quasi-log scheme, $D$0 is projective, and $D$1 is a $D$2-nef Cartier divisor such that for some $D$3,

$D$4

and if $D$5 is $D$6-generated for every $D$7, then $D$8 is $D$9-generated for all $(X,\Delta)$0 (Fujino, 2014).

Its proof is inductive on $(X,\Delta)$1. The main tools are adjunction and vanishing for quasi-log schemes, reduction to the non-qlc locus, and a perturbation lemma replacing $(X,\Delta)$2 by $(X,\Delta)$3 for a small effective divisor $(X,\Delta)$4, thereby creating new qlc centers inside the base locus. A Noetherian induction argument using two large primes $(X,\Delta)$5 then shrinks the relative base locus by passing from $(X,\Delta)$6 to $(X,\Delta)$7 (Fujino, 2014).

This formalism recovers the usual Reid–Fukuda base-point-free theorem for log-canonical pairs: if $(X,\Delta)$8 is lc and $(X,\Delta)$9 is $D$0-nef Cartier with $D$1 nef and log big over $D$2, then $D$3 is $D$4-generated. It also applies to non-normal objects such as a nodal curve $D$5 on a smooth surface, which carries the quasi-log structure $D$6 (Fujino, 2014).

4. Foliated base-point-free theorems

In foliated birational geometry the canonical class $D$7 is replaced by the canonical class $D$8 of a foliation. For a normal variety $D$9, a rank one foliation is a saturated rank-one subsheaf

$D-(K_X+\Delta)$0

closed under the Lie bracket, and $D-(K_X+\Delta)$1 is defined by $D-(K_X+\Delta)$2. A rank-one foliated pair is $D-(K_X+\Delta)$3, where $D-(K_X+\Delta)$4 is $D-(K_X+\Delta)$5-Cartier; it is log canonical if every discrepancy satisfies $D-(K_X+\Delta)$6 (Cascini et al., 3 Sep 2025).

Cascini–Spicer prove a foliated base-point-free theorem on threefolds. Let $D-(K_X+\Delta)$7 be a $D-(K_X+\Delta)$8-factorial projective klt threefold, let $D-(K_X+\Delta)$9 be a foliation of rank one, and let $D$00 with $D$01 ample and $D$02. If $D$03 is log canonical and

$D$04

is nef, then $D$05 is semi-ample; equivalently, there exists $D$06 such that $D$07 is base-point free (Cascini et al., 3 Sep 2025).

The proof follows the same broad outline as the classical Kawamata–Shokurov theorem but requires three foliated ingredients: a cone theorem for $D$08 whose negative extremal rays are spanned by $D$09-invariant curves, a foliated MMP in dimension three, and a semialgebraic integrability statement for the non-big case. When $D$10 is nef but not big, one shows that $D$11 is algebraically integrable and that there is an $D$12-invariant covering family of curves $D$13 with $D$14, so $D$15 induces a fibration and is semi-ample (Cascini et al., 3 Sep 2025).

A related theorem for algebraically integrable foliations is due to Chaudhuri–Das. If $D$16 is a $D$17-factorial normal projective variety, $D$18 is an algebraically integrable foliation induced by a dominant rational map, $D$19 is ample, $D$20, $D$21 is F-dlt, and $D$22 is klt, then nefness of

$D$23

implies semiampleness (Chaudhuri et al., 2023). The proof uses a foliated cone theorem, an MMP with scaling, and a canonical bundle formula for generalized foliated pairs.

These results are accompanied by abundance statements. For a log canonical rank-one foliated pair of any dimension with $D$24, Cascini–Spicer prove

$D$25

(Cascini et al., 3 Sep 2025).

5. Positive characteristic and mixed characteristic

In positive characteristic, the theorem persists but the methods differ sharply from characteristic zero. Over an algebraically closed field of characteristic $D$26, if $D$27 is a projective three-dimensional klt pair and $D$28 is nef and big with $D$29 nef and big, then $D$30 is semi-ample (Xu, 2013). Xu’s proof uses Keel’s endowed-with-map theorem, reduction to a one-dimensional exceptional locus, plt extraction over curves, and abundance on a surface.

Over $D$31, Martinelli–Nakamura–Witaszek prove that if $D$32 is a three-dimensional projective log pair satisfying either lc or a normality condition on $D$33, and $D$34 is nef and big with

$D$35

then $D$36 is semiample (Martinelli et al., 2014). Their argument reduces to surfaces using Keel’s theorem, then glues sections across reducible conductor loci.

Nakamura–Witaszek subsequently removed the bigness assumption on $D$37 itself for lc threefolds over $D$38 when $D$39: if $D$40 is a nef $D$41-Cartier $D$42-divisor and $D$43 is nef and big, then $D$44 is semiample (Nakamura et al., 2016). Bernasconi refined the klt threefold theorem in large characteristic: over a perfect field of characteristic $D$45, if $D$46 is a nef Cartier divisor of numerical dimension $D$47 and $D$48 is big and nef, then $D$49 is base point free for all sufficiently large $D$50 (Bernasconi, 2019).

The mixed-characteristic analogue takes the form of Keel’s theorem. If $D$51 is projective over an excellent base $D$52 of mixed characteristic and $D$53 is nef, then $D$54 is semiample over $D$55 if and only if both $D$56 and $D$57 are semiample over $D$58, where $D$59 is the exceptional locus; the same equivalence holds for the endowed-with-a-map property (Witaszek, 2020). The proof replaces Frobenius with multiplicative perfection of the structure sheaf and constructs pushouts along universal homeomorphisms.

A nearby positive-characteristic result shows that for a strongly $D$60-regular pair $D$61, an ample $D$62-divisor $D$63, and

$D$64

strictly nef, $D$65 is ample; in particular some positive multiple $D$66 is base-point-free (Cascini et al., 2013). This shifts the discussion from semiampleness under adjoint nefness to ampleness under strict nefness.

6. Analytic proofs, effective bounds, and scope of the theorem

Song’s analytic proof of Kawamata’s theorem replaces algebraic vanishing theory by geometric analysis. Starting from a smooth projective manifold with $D$67 nef and big, the normalized Kähler–Ricci flow converges to a singular Kähler–Einstein current, almost Kähler–Einstein metrics converge in Gromov–Hausdorff sense to a compact length space homeomorphic to the canonical model, and Hörmander $D$68-estimates on the regular locus produce nonzero pluricanonical sections at every point. Compactness then yields global generation of a large multiple of $D$69 (Song, 2014).

Effective versions exist in low dimension for non-normal singularities. For a projective semi-log canonical curve $D$70, if $D$71 is Cartier and $D$72 satisfies $D$73 for every irreducible component $D$74, then $D$75 is base-point-free. For a projective semi-log canonical surface, if

$D$76

and

$D$77

then $D$78 is base-point-free (Fujino, 2015). These bounds yield very-ampleness consequences for stable surfaces and semi-log canonical Fano surfaces.

The modern literature therefore presents the base point free theorem as a family of adjoint semiampleness principles. This suggests that the phrase “base point free theorem” functions less as a single immutable statement than as a template whose precise positivity hypotheses depend on the ambient category: projective klt pairs, quasi-log schemes, compact Kähler threefolds, foliated pairs, or varieties in positive or mixed characteristic. The common endpoint is the same: nef adjoint positivity is upgraded to a morphism, a contraction, or a descended Kähler class. Open cases remain, notably the higher-dimensional Calabi–Yau and hyperkähler transcendental setting (Höring, 2018).