Miyaoka–Yau Inequality Overview

• Miyaoka–Yau inequality is a set of Chern class bounds that relate curvature, topology, and stability on complex varieties with ample or nef canonical bundles.
• It extends the classical surface inequality (c₁² ≤ 3c₂) to higher dimensions, orbifolds, singular spaces, and even combinatorial settings like hyperplane arrangements.
• Equality in the inequality often signals uniformization, indicating that the variety may be realized as a ball, torus, or projective-space quotient.

The Miyaoka–Yau inequality is a family of Chern class inequalities governing complex varieties whose canonical or anti-canonical geometry is sufficiently positive, flat, or controlled by stability. In its classical surface form it is the Bogomolov–Miyaoka–Yau inequality $c_1^2 \le 3c_2$; in higher dimension it takes the form

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$

and it has been extended to orbifolds, klt spaces, log pairs, compact Kähler spaces, and more specialized settings such as hyperplane arrangements in $\mathbb{CP}^n$ (Greb et al., 2015, Borbon et al., 2024).

1. Classical formulation and differential-geometric origin

For smooth minimal surfaces of general type, the inequality is

$c_1(T_X)^2 \le 3\,c_2(T_X),$

equivalently $c_1(X)^2 \le 3c_2(X)$. In higher dimension, Yau’s Kähler–Einstein framework yields

$$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$

for compact Kähler manifolds with ample $K_X$; when $n=2$, this reduces exactly to the surface inequality (Greb et al., 2015).

A fundamental differential-geometric expression is the Chern–Weil identity

$$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$

where $\operatorname{Rm}_c(\omega)$ and $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$0 are the trace-free curvature and trace-free Ricci components. This identity clarifies why the inequality is a curvature constraint: it compares a Chern number combination to an $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$1-norm of curvature tensors (Nomura, 2018).

In the smooth negatively curved setting, equality is rigid. If equality holds for a compact Kähler manifold with $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$2 ample, then the trace-free curvature vanishes, the metric has constant negative holomorphic sectional curvature, and the universal cover is the complex unit ball $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$3 (Greb et al., 2015). This equality phenomenon is the prototype for later uniformization theorems in singular, orbifold, and logarithmic settings.

2. Singular, orbifold, and logarithmic formulations

A major development is the replacement of ordinary Chern classes by orbifold or $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$4-Chern classes of reflexive sheaves. For an $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$5-dimensional projective klt variety of general type with nef $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$6, the orbifold Miyaoka–Yau inequality is

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$7

This formulation uses the reflexive tangent sheaf, Higgs sheaves, semistability, and a $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$8-Bogomolov–Gieseker inequality on klt spaces (Greb et al., 2015).

For minimal dlt pairs $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$9 with standard coefficients and $\mathbb{CP}^n$0 nef, the inequality takes an orbifold-logarithmic form. If $\mathbb{CP}^n$1, $\mathbb{CP}^n$2, and $\mathbb{CP}^n$3, then

$\mathbb{CP}^n$4

for any ample divisor $\mathbb{CP}^n$5. When $\mathbb{CP}^n$6 is nef and big, the mixed term simplifies to $\mathbb{CP}^n$7 (Guenancia et al., 2016).

The same logic has a stack-theoretic realization. For a smooth proper Deligne–Mumford surface $\mathbb{CP}^n$8 with projective coarse moduli space and nef $\mathbb{CP}^n$9,

$c_1(T_X)^2 \le 3\,c_2(T_X),$0

In codimension one, root-stack constructions recover logarithmic inequalities of the form $c_1(T_X)^2 \le 3\,c_2(T_X),$1; in codimension two, the formula reproduces Miyaoka-type bounds for quotient singularities (Chen et al., 2011).

These singular and orbifold formulations preserve the original philosophy of the inequality while replacing smooth Chern classes by invariants adapted to quotient singularities, adapted covers, or orbifold charts. The decisive inputs are semistability of tangent or cotangent objects and Bogomolov–Gieseker-type inequalities in the appropriate category.

3. Equality and uniformization

The equality case is the rigid core of Miyaoka–Yau theory. For minimal varieties of general type with terminal singularities and nef $c_1(T_X)^2 \le 3\,c_2(T_X),$2, equality in the orbifold inequality implies that the canonical model is smooth in codimension two and admits a finite Galois quasi-étale cover by a ball quotient. Equivalently, the canonical model is a singular ball quotient, meaning it is projective, klt, smooth in codimension two, $c_1(T_X)^2 \le 3\,c_2(T_X),$3 is ample, and equality holds in the orbifold Miyaoka–Yau expression (Greb et al., 2015).

For compact Kähler klt pairs $c_1(T_X)^2 \le 3\,c_2(T_X),$4 with standard coefficients, the equality theory bifurcates according to the sign of $c_1(T_X)^2 \le 3\,c_2(T_X),$5. If $c_1(T_X)^2 \le 3\,c_2(T_X),$6 is ample and equality holds in the orbifold Miyaoka–Yau inequality, then the orbifold universal cover is $c_1(T_X)^2 \le 3\,c_2(T_X),$7. If $c_1(T_X)^2 \le 3\,c_2(T_X),$8 and equality holds in the flat version

$c_1(T_X)^2 \le 3\,c_2(T_X),$9

then the orbifold universal cover is $c_1(X)^2 \le 3c_2(X)$0 (Claudon et al., 2023).

In positive curvature, the analogous equality theory uses the canonical extension of the orbifold tangent sheaf. For a log Fano pair $c_1(X)^2 \le 3c_2(X)$1 with standard coefficients, semistability of the canonical extension together with

$c_1(X)^2 \le 3c_2(X)$2

implies that the orbifold universal cover is $c_1(X)^2 \le 3c_2(X)$3, equivalently $c_1(X)^2 \le 3c_2(X)$4 for a finite group $c_1(X)^2 \le 3c_2(X)$5 (Dailly, 10 Jan 2025).

Equality also has flat orbifold incarnations on surfaces of Kodaira dimension $c_1(X)^2 \le 3c_2(X)$6. For smooth-orbifold K3 and Enriques surfaces, equality in the orbifold Bogomolov–Miyaoka–Yau inequality is characterized by uniformization by $c_1(X)^2 \le 3c_2(X)$7. In the K3 case, this is tied to generalized Kummer surfaces $c_1(X)^2 \le 3c_2(X)$8, where equality is realized on quotient orbifolds $c_1(X)^2 \le 3c_2(X)$9 and the uniformization group is a lattice in the affine automorphism group of $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$0 (Roulleau, 2017).

Recent work also shows that these equality cases can be topologically rigid. Singular ball quotients, singular torus quotients, and projective spaces are characterized by Chern-class equalities and, in several cases, by homeomorphism type among projective klt varieties (Greb et al., 2023).

4. Kähler, nef/big, and intermediate-Kodaira-dimension extensions

The classical projective hypothesis is not essential for many modern formulations. For compact Kähler manifolds with semi-positive $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$1, one has

$$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$2

proved via the normalized Kähler–Ricci flow and an $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$3-estimate for the scalar curvature (Nomura, 2018). A shorter argument using constant scalar curvature Kähler metrics near the canonical class yields the same inequality for every compact Kähler manifold with nef canonical bundle, that is, for every smooth minimal model in the Kähler sense (Liu, 2020).

In the singular Kähler setting, minimal Kähler klt spaces satisfy a mixed Miyaoka–Yau inequality depending on the numerical dimension $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$4. If $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$5, then for any Kähler class $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$6,

$$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$7

This statement is obtained from generalized Bogomolov inequalities for Higgs sheaves on compact Kähler klt spaces and from analytic estimates for twisted Kähler–Einstein metrics on resolutions (Zhang et al., 17 Mar 2025).

When the relevant class is big but not nef, the polarization is replaced by a non-pluripolar product. For projective klt $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$8-folds with big $$\bigl(2(n+1)c_2(X)-n\,c_1(X)^2\bigr)\cdot [K_X]^{n-2}\ge 0$$9,

$K_X$0

There is an anti-canonical counterpart for klt varieties with big $K_X$1 under K-semistability, again polarized by a non-pluripolar product (Iwai et al., 11 Jul 2025).

A recent $K_X$2-sensitive extension interpolates between the numerically trivial and general-type endpoints. If $K_X$3 is a minimal projective klt $K_X$4-fold with numerical dimension $K_X$5, then for sufficiently small rational $K_X$6,

$K_X$7

for any ample Cartier divisor $K_X$8. Equality is characterized by a finite quasi-étale Galois cover $K_X$9 with $n=2$0, where $n=2$1 is an Abelian variety of dimension $n=2$2 and $n=2$3 is a smooth ball quotient (Müller, 21 Jan 2026).

There is also an anti-canonical nef version. For an $n=2$4-dimensional projective manifold with nef $n=2$5, if

$n=2$6

for some ample line bundle $n=2$7, then

$n=2$8

If $n=2$9 is nef and big, equality is equivalent to the anti-canonical model admitting a finite codimension-one étale cover (Hisamoto, 2024).

5. Arrangement-theoretic and combinatorial incarnations

A striking recent development is the translation of Miyaoka–Yau into discrete geometry for hyperplane arrangements. Let $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$0 be a hyperplane arrangement in $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$1. One defines a quadratic form $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$2 on $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$3, determined entirely by the intersection poset of $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$4, by

$$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$5

with $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$6 and $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$7. If the weighted arrangement $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$8 is stable, then $$\bigl( 2(n+1)\, c_2(X) - n\, c_1(X)^2 \bigr) \cdot [\omega]^{n-2} = \frac{1}{4\pi^2\, n(n-1)} \int_X \left( (n+1)\, |\operatorname{Rm}_c(\omega)|^2 - (n+2)\, |\operatorname{Ric}_c(\omega)|^2 \right)\, \omega^n,$$9 (Borbon et al., 2024).

The proof uses the minimal De Concini–Procesi resolution, a locally abelian parabolic structure on $\operatorname{Rm}_c(\omega)$0, and Mochizuki’s parabolic Bogomolov–Gieseker inequality. On the Calabi–Yau slice $\operatorname{Rm}_c(\omega)$1, the leading coefficient of the parabolic second Chern class is precisely the quadratic expression defining $\operatorname{Rm}_c(\omega)$2, so the arrangement inequality is a genuine Miyaoka–Yau-type statement in combinatorial disguise (Borbon et al., 2024).

In the symmetric case of equal weights, the inequality becomes a lower bound on codimension-two multiplicities. If $\operatorname{Rm}_c(\omega)$3, then

$\operatorname{Rm}_c(\omega)$4

Equality holds if and only if every hyperplane $\operatorname{Rm}_c(\omega)$5 meets the others along

$\operatorname{Rm}_c(\omega)$6

codimension-two subspaces, extending Hirzebruch’s condition from line arrangements in $\operatorname{Rm}_c(\omega)$7 to higher dimension (Borbon et al., 2024).

Earlier orbifold-logarithmic methods already produced Hirzebruch-type inequalities for plane curve arrangements. For a line arrangement with no point of multiplicity greater than $\operatorname{Rm}_c(\omega)$8, Langer’s orbifold Miyaoka–Yau inequality yields

$\operatorname{Rm}_c(\omega)$9

improving the classical Hirzebruch inequality in that range (Pokora, 2016). The arrangement version in $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$00 may therefore be viewed as a higher-dimensional parabolic generalization of an already established log-surface paradigm.

6. Related variants and broader scope

The Miyaoka–Yau mechanism extends beyond the minimal or general-type locus, but often with modified coefficients or correction terms. For every smooth projective non-uniruled variety $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$01 of dimension $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$02 and any ample divisor $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$03, one can write

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$04

with

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$05

If $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$06 is nef, then $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$07, recovering Miyaoka’s algebraic inequality $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$08 (2208.01343).

In positive characteristic, the classical surface inequality fails in its characteristic-zero form, but a substitute survives. For a minimal smooth projective surface of general type over an algebraically closed field of characteristic $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$09,

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$10

If $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$11, then the Albanese morphism induces a genus-two fibration; equality $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$12 is classified explicitly (Gu et al., 2019).

There is also a transverse Sasakian analogue. For a compact Sasakian manifold of real dimension $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$13 with nonpositive transverse holomorphic sectional curvature,

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$14

Here $$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$15 are the basic Chern classes of the Reeb foliation. Under quasi-negative transverse holomorphic sectional curvature, one further has

$$\bigl(2(n+1)c_2 - n c_1^2\bigr)\cdot [K_X]^{n-2}\ge 0,$$16

(Chen, 2021).

Across these variants, the same structural pattern recurs. The inequality is obtained from some combination of semistability, Bogomolov–Gieseker-type estimates, Chern–Weil identities, Kähler–Einstein or Ricci-flow methods, and carefully chosen polarizations. Equality, when understood, is typically uniformizing: it singles out ball quotients, torus quotients, projective-space quotients, or their orbifold analogues. This suggests that the Miyaoka–Yau inequality is best regarded not as a single formula, but as a rigidity principle linking curvature, stability, and the birational or orbifold structure of complex varieties (Greb et al., 2015, Claudon et al., 2023).