Noether inequality explained

In mathematics, the Noether inequality, named after Max Noether, is a property of compact minimal complex surfaces that restricts the topological type of the underlying topological 4-manifold. It holds more generally for minimal projective surfaces of general type over an algebraically closed field.

Formulation of the inequality

Let X be a smooth minimal projective surface of general type defined over an algebraically closed field (or a smooth minimal compact complex surface of general type) with canonical divisor K = −c1(X), and let pg = h0(K) be the dimension of the space of holomorphic two forms, then

pg\le

1
2
2
c
1(X)

+2.

For complex surfaces, an alternative formulation expresses this inequality in terms of topological invariants of the underlying real oriented four manifold. Since a surface of general type is a Kähler surface, the dimension of the maximal positive subspace in intersection form on the second cohomology is given by b+ = 1 + 2pg. Moreover, by the Hirzebruch signature theorem c12 (X) = 2e + 3σ, where e = c2(X) is the topological Euler characteristic and σ = b+ − b is the signature of the intersection form. Therefore, the Noether inequality can also be expressed as

b+\le2e+3\sigma+5

or equivalently using e = 2 – 2 b1 + b+ + b

b-+4b1\le4b++9.

Combining the Noether inequality with the Noether formula 12χ=c12+c2 gives

5

2
c
1(X)

-c2(X)+36\ge12q

where q is the irregularity of a surface, which leads toa slightly weaker inequality, which is also often called the Noether inequality:

5

2
c
1(X)

-c2(X)+36\ge0

2(X)even)
(c
1

5

2
c
1(X)

-c2(X)+30\ge0

2(X)odd).
(c
1

Surfaces where equality holds (i.e. on the Noether line) are called Horikawa surfaces.

Proof sketch

It follows from the minimal general type condition that K2 > 0. We may thus assume that pg > 1, since the inequality is otherwise automatic. In particular, we may assume there is an effective divisor D representing K. We then have an exact sequence

0\to

0(l{O}
H
X)

\toH0(K)\toH0(K|D)\to

1(l{O}
H
X)

\to

so

pg-1\le

0(K|
h
D).

Assume that D is smooth. By the adjunction formula D has a canonical linebundle

l{O}D(2K)

, therefore

K|D

is a special divisor and the Clifford inequality applies, which gives
0(K|
h
D)

-1\le

1
2

\degD(K)=

1
2

K2.

In general, essentially the same argument applies using a more general version of the Clifford inequality for local complete intersections with a dualising line bundle and 1-dimensional sections in the trivial line bundle. These conditions are satisfied for the curve D by the adjunction formula and the fact that D is numerically connected.