Seshadri Constant¶

A local positivity invariant of a nef line bundle at a point, computed by the least ratio of curve intersection to curve multiplicity, equivalently by the nef threshold on the blowup.

Version: v1 · 2026-08-30 · History
Domain-specific #: 2761
Origin domain: algebraic geometry
Subdomain: positivity
Aliases: Local Seshadri constant, Seshadri invariant at a point

Core Idea¶

The Seshadri constant \(\varepsilon(L;x)\) measures how much local positivity a nef line bundle \(L\) has at a point \(x\) of a projective variety \(X\). For an irreducible curve \(C\subset X\) through \(x\), the intersection number \(L\cdot C\) measures global degree along \(C\), while \(\operatorname{mult}_x C\) measures how heavily that curve concentrates at \(x\). Their ratio penalizes curves that consume much multiplicity with little degree:

\[ \varepsilon(L;x)=\inf_{C\ni x}\frac{L\cdot C}{\operatorname{mult}_x C}. \]

On the blowup \(\pi:\widetilde X=\operatorname{Bl}_xX\to X\) with exceptional divisor \(E\), the same value is the nef threshold

\[ \varepsilon(L;x)=\sup\{t\ge 0: \pi^*L-tE\text{ is nef}\}. \]

This curve-ratio/nef-threshold equivalence locks the identity. Ein, Küchle, and Lazarsfeld describe the constant as measuring local positivity and use it to express Seshadri's ampleness criterion and local generation behavior.^[1]

Structural Signature¶

Projective setting: a projective variety \(X\), with regularity assumptions declared for each theorem.
Positive object: a nef or ample line bundle \(L\), or its numerical divisor class.
Marked point: \(x\in X\), where positivity is tested.
Curve family: irreducible curves \(C\subset X\) through \(x\).
Global numerator: the intersection degree \(L\cdot C\).
Local denominator: \(\operatorname{mult}_x C\).
Extremal operation: an infimum over all qualifying curves.
Blowup dual: the largest exceptional coefficient \(t\) preserving nefness of \(\pi^*L-tE\).
Numerical invariance: the value depends on the numerical class of \(L\), not a chosen equation.

The constant is not established by inspecting one convenient curve unless a proof shows it computes the infimum.

What It Is Not¶

It is not the global degree, volume, ampleness, or nefness of \(L\). Those are global properties; the Seshadri constant evaluates their local resistance to imposed multiplicity. It is not a distance from \(x\), a curvature scalar, a singularity multiplicity, or a probability.

It is not merely the ratio for a selected curve. Any curve through \(x\) can upper-bound the value, but the infimum ranges over all such curves. It is not automatically attained. Nor is the multipoint Seshadri constant identical to the single-point invariant: blowing up several points gives multiple exceptional divisors and a different admissible threshold.

Scope of Application¶

Seshadri constants quantify local positivity of line bundles on projective varieties. They enter criteria for ampleness, lower bounds for separation of jets, behavior of adjoint linear systems, and comparisons between general and special points. Ein, Küchle, and Lazarsfeld prove uniform lower bounds at very general points in fixed dimension, while emphasizing that bounds valid at every point cannot ignore the particular variety and bundle.^[1]

On surfaces, the invariant is closely tied to curves of low degree and high local multiplicity; such curves can be Seshadri-exceptional and compute or bound the constant. Bauer's study of algebraic surfaces stresses both explicit bounds and the difficulty of exact computation even in low dimension.^[2]

The concept extends to multiple points, moving constants, \(\mathbf Q\)- and \(\mathbf R\)-divisors, and arithmetic analogues, but those extensions must state their modified blowup and normalization. The single-point invariant remains the catalog identity here.

Clarity¶

For \(X=\mathbf P^n\), \(L=\mathcal O_{\mathbf P^n}(1)\), and any point \(x\), a line through \(x\) gives ratio \(1/1=1\), so \(\varepsilon(L;x)\le1\). Degree/multiplicity bounds give \(\deg C\ge\operatorname{mult}_xC\), hence every curve ratio is at least \(1\). Therefore \(\varepsilon(\mathcal O(1);x)=1\).

If \(L\) is replaced by \(L^{\otimes m}\), then intersection degrees scale by \(m\) while multiplicities do not, so

\[ \varepsilon(L^{\otimes m};x)=m\varepsilon(L;x). \]

On the blowup, the threshold says the same thing geometrically: subtracting \(tE\) demands vanishing of order \(t\) at \(x\), and nefness fails exactly when some curve transform acquires negative intersection. This equivalence converts an infinite ratio search into a boundary problem in the nef cone.

Manages Complexity¶

Local positivity involves all curves through one point, each with global degree and local singularity behavior. The constant compresses this family into one extremal number without discarding the geometric obstruction: a curve approaching the infimum identifies where positivity is weakest. The blowup form reorganizes the same complexity into convex geometry of divisor classes.

The number is useful precisely because it is simultaneously local, numerical, and functorial enough for intersection-theoretic arguments. It does not replace the curve data when exact attainment, exceptional loci, or variation across points is the question.

Abstract Reasoning¶

The invariant is a constrained worst-case ratio. It balances a resource \(L\cdot C\) against a local demand \(\operatorname{mult}_xC\) and asks which curve produces the least margin. The blowup reveals a dual view: increase the local penalty \(tE\) until the global nonnegativity constraint becomes active.

This duality supports proofs by either curves or cones. A curve construction yields upper bounds. Showing \(\pi^*L-tE\) is nef yields lower bounds. Equality follows when the two meet. The method resembles optimization, but the feasible objects, intersection pairing, and nef cone are specifically algebraic-geometric.

Knowledge Transfer¶

The structural idea transfers to several marked points by blowing them up and subtracting a weighted sum of exceptional divisors. It also motivates moving Seshadri constants that discount fixed base loci. Such transfer is valid only after redefining the admissible curve family and normalization; a multipoint value cannot be silently substituted for \(\varepsilon(L;x)\).

Analogies to condition numbers or local Lipschitz constants can aid intuition because each records worst-case local behavior, but they are not aliases. The mathematical invariant requires line-bundle intersection, curve multiplicity, and a nef threshold.

Examples¶

Projective space: \(\varepsilon(\mathcal O_{\mathbf P^n}(1);x)=1\), computed by any line through \(x\).
Tensor powers: \(\varepsilon(L^{\otimes m};x)=m\varepsilon(L;x)\), so normalization of \(L\) matters.
Curve case: on a smooth projective curve, the only irreducible curve through \(x\) is the whole curve with multiplicity one, giving \(\varepsilon(L;x)=\deg L\).
Surface obstruction: a curve \(C\) with unusually small \(L\cdot C\) relative to \(\operatorname{mult}_xC\) supplies an upper bound and may compute the constant; proving it does requires ruling out all better competitors.^[2]
Very general point: in dimension \(n\), the Ein–Küchle–Lazarsfeld theorem gives a uniform lower bound \(1/n\) for an ample \(L\) outside a countable union of proper subvarieties.^[1]

Structural Tensions¶

Local quantity vs. global inputs. The point is local, but admissible curves and intersection degrees are global. Diagnostic: never infer the value from an analytic germ alone without a theorem connecting it to projective data.
Infimum vs. minimum. A curve may approximate the value without attaining it. Diagnostic: call a curve computing the constant only when equality is proved.
Positivity vs. normalization. Tensoring \(L\) scales the answer. Diagnostic: record the exact divisor or line-bundle class.
General vs. special points. Constants can jump on exceptional loci. Diagnostic: distinguish “every,” “general,” and “very general” claims.
Single-point vs. multipoint variants. Both use blowups but have different denominators and exceptional classes. Diagnostic: declare the marked-point set and weights.

Structural–Framed Character¶

The structural pattern is extremal local measurement: compare every obstruction's global capacity with its local concentration, then take the worst ratio. The frame supplies projective varieties, line bundles, intersection numbers, curve multiplicities, blowups, and nefness.

Because those objects are indispensable to recognition, the abstraction remains domain-specific. Optimization and Measurement illuminate aspects of its operation, but neither determines a Seshadri constant.

Structural Core vs. Domain Accent¶

Structural core: localized demand, family of competitors, normalized ratios, infimum, and equivalent feasibility threshold.

Domain accent: \(L\cdot C\), \(\operatorname{mult}_xC\), \(\operatorname{Bl}_xX\), exceptional divisor \(E\), and nef cone. The accent fixes both meaning and theorems.

Seshadri Constant compositionally presupposes Optimization: its definition specifies an admissible family of curves and takes an infimum, while its blowup form takes a supremal feasible coefficient. It is not a specialization of generic Optimization because the output is an invariant, not a solver or problem class. It is related to Measurement as a scalar assessment of local positivity and to Bottleneck through the curve attaining or approaching the infimum. Optimization is the minimal literal parent.

Relationships to Other Abstractions¶

Current abstraction Seshadri Constant Domain-specific

Parents (1) — more general patterns this builds on

Seshadri Constant presupposes Optimization Prime

Seshadri Constant compositionally presupposes Optimization: its definition specifies an admissible family of curves and takes an infimum, while its blowup form takes a supremal feasible coefficient.

Hierarchy path (1) — routes to 1 parentless root

Seshadri Constant → Optimization

Neighborhood in Abstraction Space¶

Seshadri Constant sits in a sparse region of the domain-specific corpus (90^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Nefness: global nonnegative intersection with every curve; the constant measures how much nefness survives a local subtraction.
Ampleness: stronger global positivity; uniform positive Seshadri bounds characterize it under the relevant criterion.
Multiplicity: the denominator of the curve ratio, not the result.
Volume of a line bundle: asymptotic global section growth, not local worst-case positivity.
Multipoint Seshadri constant: related extension with several exceptional divisors.
Seshadri criterion: an ampleness criterion expressed through constants, not the constant itself.

References¶

[1] Lawrence Ein, Oliver Küchle, and Robert Lazarsfeld, “Local Positivity of Ample Line Bundles,” Journal of Differential Geometry 42 (1995), 193–219; arXiv:alg-geom/9408003. registry ↩a ↩b ↩c

[2] Thomas Bauer, “Seshadri Constants on Algebraic Surfaces,” Mathematische Annalen 313 (1999), 547–583, DOI: 10.1007/s002080050274; arXiv:math/9903072. registry ↩a ↩b