Lectures on Cycle Spaces J´onIng´olfur Magn´usson
1 Introduction
These lecture notes are based on a series of lectures given by the author at Ruhr Univer- sit¨atBochum in February 2004. It was supposed to be a “blitz” introduction to the theory of cycle spaces for the research group in Complex Geometry of the Mathematics Depart- ment. The first half is a rough introduction to analytic geometry and it is complemented by an appendix containing algebraic supplements. I would like to thank Professor Peter Heinzner and Professor Alan Huckelberry for having invited me to give these lectures. Special thanks are also due to Christian Miebach for his invaluable help in preparing these notes. He typed the first draft based on the lectures and wrote the original version of the appendix. Equally he read carefully the final version of these notes and made several interesting and helpful remarks. Finally I would like to thank the staff of the Mathematics Department for their kind assistance during my stay in Bochum.
2 Contents
1 Complex spaces 5 1.1 Ringed spaces ...... 5 1.2 Morphisms of ringed spaces ...... 6 1.3 Local models ...... 9 1.4 Complex spaces ...... 11 1.5 Closed complex subspaces ...... 13 n n 1.6 The bijection Hol(X, C ) → OX (X) ...... 16 1.7 Geometry of complex spaces ...... 17 1.8 Maximum principle ...... 23 1.9 Analytic sets ...... 24 1.10 Normal spaces ...... 30
2 Operations on complex spaces 31 2.1 Inverse images and fibers ...... 31 2.2 Images of complex spaces ...... 32 2.3 Fiber products ...... 33
3 Meromorphic functions and maps 35 3.1 Meromorphic functions ...... 35 3.2 Modifications ...... 37 3.3 Meromorphic maps ...... 39
4 Differential forms 42 4.1 Differential forms on complex spaces ...... 43 4.2 K¨ahler–spaces ...... 43 4.3 Integration on complex spaces ...... 46
5 Flatness 47 5.1 Flat maps ...... 47 5.2 Flat families of compact subspaces ...... 49
6 The Douady space of a complex space 50
7 The space of holomorphic maps 51 7.1 The universal holomorphic map ...... 51 7.2 The group of automorphisms ...... 52
8 Ramified coverings 53 8.1 Analytic coverings ...... 53 8.2 Ramified coverings ...... 54
9 The Barlet space of a (reduced) complex space 55 9.1 Analytic families ...... 55 9.2 The Barlet space ...... 58
3 10 The morphism Douady to Barlet 59 10.1 Fundamental cycles ...... 59 10.2 Definition of the morphism ...... 60 10.3 Some special cases ...... 61
11 Compact subsets of Barlet spaces 62 11.1 The theorem of Bishop and its consequences ...... 62 11.2 The Fujiki class ...... 64 11.3 Corresponding results for the Douady space ...... 65
A Algebraic supplements 65 A.1 Elementary theory of rings ...... 65 A.2 The full ring of fractions ...... 67 A.3 Local rings and Noetherian rings ...... 69 A.4 The ring C{z1, . . . , zn} ...... 71
4 1 Complex spaces
1.1 Ringed spaces
Definition 1.1.1 (Ringed space) A ringed space is a couple X = (|X|, SX ) where |X| is a topological space and SX is a sheaf of C–algebras having the following properties:
(i) For each x ∈ |X| the stalk SX,x is a local algebra.
(ii) For each open set U ⊂ |X| and for each section s ∈ SX (U) the map
RedU (s): U → C, x 7→ s(x) := ε(sx), is continuous, where ε denotes the evaluation map (See A.3.4).
The map RedU (s) is called the reduction of s. The sheaf SX is usually referred to as the structure sheaf of X. If no confusion is possible, the topological space |X| will also be denoted by X.
Remark. Let CX denote the sheaf of continuous functions on |X|, i.e. for every open set U of X CX (U) = {f : U → C; f is continuous}.
Then the family of maps {RedU ; U ⊂ X open}, where
RedU : S (U) → C (U), s 7→ RedU (s), is a morphism of sheaves, denoted by
Red: SX → CX .
Examples.
(1) Let X be a topological space. Then the couple (X, CX ) is a ringed space. (See A.3).
n p (2) For every open set U in R and p ∈ N0 ∪ {∞} the couple (U, CU ) is a ringed space, p where CU denotes the sheaf of complex valued p times continuously differentiable functions, i.e. for each open subset V of U
p CU (V ) := {f : V → C; f is continuously differentiable of order p}.
n (3) Let D be an open subset of C and let OD be the sheaf of holomorphic functions in D. Then the couple (D, OD) is a ringed space.
Definition 1.1.2 Let X = (|X|, SX ) be a ringed space. Let Red(SX ) ⊂ C|X| be the sheaf associated to the presheaf U → Red(SX (U)). Then
Red(X) := Xred := (|X|, Red(SX )) is a ringed space, called the reduction of X. If the morphism Red: SX → CX is injective and consequently induces an isomorphism ∼ SX = Red(SX ), then X is called a reduced ringed space and the sheaves SX and Red(SX ) will be systematically identified in that case.
5 Remarks.
(i) For a reduced ringed space X the structure sheaf SX is systematically identified with the sheaf Red(SX ) so in that case one often writes s instead of Red(s) for a local section s of SX .
(ii) The sheaf morphism Red: SX → Red(SX ) is automatically surjective and consequently induces an isomorphism ∼ Red(SX ) = SX / Ker(Red).
(iii) Let U be an open subset of X, let x ∈ U and let s ∈ SX (U). Then
s(x) = 0 is equivalent to sx ∈ mx,
where mx denotes the maximal ideal of SX,x.
Examples.
(1) The ringed spaces in examples 1, 2 and 3 are all reduced.
(2) A ringed space of the form ({x}, S ), where x is a point and S is a local C−algebra, is called a one point ringed space. It is reduced if and only if S = C.
1.2 Morphisms of ringed spaces
Let (X, SX ) be a ringed space. For every topological space Y and every continuous map f : X → Y the direct image sheaf f∗S is a sheaf of C–algebras on Y . Note however, that −1 f∗S is not necessarily a sheaf of local C–algebras because it may happen that f (V ) = ∅ for some non-empty open subset V of Y and consequently (f∗SX )y = 0 for all y ∈ V .
Definition 1.2.1 (Morphism of ringed spaces) A morphism from a ringed space X to a ringed space Y is a couple ˜ (f, f):(|X|, SX ) → (|Y |, SY ), ˜ where f : |X| → |Y | is a continuous map and f : SY → f∗SX is a morphism of sheaves of C–algebras. ˜ Definition 1.2.2 Let (f, f):(|X|, SX ) → (|Y |, SY ) and (g, g˜):(|Y |, SY ) → (|Z|, SZ ) be morphisms of ringed spaces. Then the morphism of ringed spaces ˜ (g ◦ f, g∗(f) ◦ g˜):(|X|, SX ) → (|Z|, SZ ), is called the composition of (f, f˜) and (g, g˜).
6 Notation. If there is no risk of ambiguity we usually write f instead of (f, f˜). Compo- sition of two morphisms f and g will be denoted fg or f ◦ g.
Definition 1.2.3 A morphism f : X → Y of ringed spaces is called an isomorphism if there exists a morphism g : Y → X with fg = idY and gf = idX .
Examples.
(1) Let X and Y be topological spaces and let f : X → Y be a continuous map. For each open subset V of Y the pull-back map
˜ −1 fV : CY (V ) → CY (f (V )), s 7→ s ◦ f
is a morphism of C–algebras. Hence we get a morphism of sheaves of C–algebras ˜ f : CY → f∗CX
and a morphism of ringed spaces ˜ (f, f):(X, CX ) → (Y, CY ).
n (2) Let D be an open subset of C and let i: OD → CD be the inclusion. Then
(idD, i):(D, CD) → (D, OD)
is a morphism of ringed spaces.
(3) For every ringed space X there is a canonical morphism Xred → X given by
|X| → |X|, x 7→ x, and SX → Red(SX ), s 7→ Red(s).
˜ Remark. Let (f, f):(|X|, SX ) → (|Y |, SY ) be a morphism of ringed spaces. Then, for each x ∈ X, we have an induced stalk map ˜ fx : SY,f(x) → SX,x obtained by composing the map SY,f(x) → (f∗SX )f(x) with the canonical map (f∗SX )f(x) → ˜ ˜ SX,x. The maps fx are morphisms of local C–algebras and they determine f.
The following lemma is obvious.
˜ Lemma 1.2.4 Let (f, f):(|X|, SX ) → (|Y |, SY ) be an isomorphism of ringed spaces. Then, for every x ∈ X, the map ˜ fx : SY,f(x) → SX,x is an isomorphism of local C–algebras.
7
Now we turn our attention to the way the morphisms of ringed spaces behave under re- duction.
˜ Let X = (|X|, SX ) and Y = (|Y |, SY ) be ringed spaces and let (f, f): X → Y be a ˜ morphism. Let V ⊂ Y be an open set and let s ∈ SY (V ). Then we have f(s) ∈ −1 (f∗SX )(V ) = SX (f (V )). Define ˜ ˜ f(s)x := fx(sf(x)) ∈ SX,x.
Lemma 1.2.5 In the above setting the identity s(f(x)) = f˜(s)(x) holds for all x ∈ f −1(V ).
Proof. For each x ∈ f −1(V ) we have the unique decomposition (see remark (ii) following definition A.3.4) sf(x) = s(f(x)) +s ¯f(x) ∈ C ⊕ mY,f(x) = SY,f(x). ˜ Since fx : SY,f(x) → SX,x is a morphism of local C–algebras we get ˜ ˜ ˜ ˜ f(s)x = fx sf(x) = fx s(f(x)) +s ¯f(x) = s(f(x)) + fx s¯f(x) ∈ C ⊕ mX,x = SX,x. ˜ Hence f(s)(x) = s(f(x)). Corollary 1.2.6 Let (f, f˜): X → Y be a morphism of ringed spaces. Then there exists a ˜ unique morphism (f, fred): Xred → Yred such that the following diagram commutes:
˜ (f,fred) Xred / Yred
X / Y (f,f˜) where the vertical arrows are the canonical morphisms.
˜ Proof. Define fred : Red(SY ) → Red(SX ) by ˜ ˜ fred(Red(s)) := Red(f(s)). ˜ Then the diagram above is commutative and the uniqueness of fred follows from lemma 1.2.5. Corollary 1.2.7 Let (f, f˜): X → Y be a morphism of ringed spaces with X reduced. Then f˜ is determined by f, more precisely f˜(s) = Red(s) ◦ f.
8 1.3 Local models n n Let D be a domain in C , let f1, . . . , fk ∈ OC (D). We put
J := hf1, . . . , fki := OD · f1 + ··· + OD · fk and N := N(f1, . . . , fk) := {x ∈ D; f1(x) = ··· = fk(x) = 0}.
Lemma 1.3.1 The couple
V := V (f1, . . . , fk) := (N, (OD/J )|N) is a ringed space.
Proof. For every x in N one has
(OD/J )x = OD,x/Jx and Jx 6= OD,x, so by lemma A.3.6 each stalk of the sheaf (OD/J )|N is a local C–algebra. Moreover, if mx denotes the maximal ideal of OD,x the maximal ideal of OD,x/Jx is mx/Jx. Consider the commutative diagram
ι ε C / OD,x / OD,x/mx
id π =∼
/ ( / )/(m / ) C ¯ι / OD,x Jx ¯ε / OD,x Jx x Jx
Let U ⊂ N be an open set and let s ∈ (OD/J )(U). Then the continuity of the function ¯ε: x 7→ s(x) follows from the fact that locally the section s can be lifted to a section of (OD/J )|N.
Definition 1.3.2 (Local model) The ringed space V = V (f1 . . . , fk) is called a local model with structure sheaf OV := (OD/J )|N. The elements of OV (U), where U is an open subset of N, are called holomorphic functions on U.
Warning. A holomorphic function f ∈ OV (U) as defined above need not be a function in the set theoretical sense. It has an underlying continuous function
Red f : U → C, x 7→ f(x) but different holomorphic functions in OV (U) can have the same underlying continuous function. (See example 4 below).
n Terminology. Let V = V (f1 . . . , fk) be a local model in a domain D in C and let n z = (z1, . . . , zn) be the standard coordinates in C . Then V is often referred to as the local model defined by the equations f1(z) = ··· = fn(z) = 0.
9 Examples.
n (1) Let D be an open set in C and let f denote the zero function on D. Then J := n hfi = 0 and N(f) = D. It follows that (OC /J )|D = OD. Thus, (D, OD) is a local model. Let z denote the usual coordinate on C. For every morphism of ringed spaces ˜ (φ, φ):(D, OD) → (C, OC) it follows from 1.2.7 that φ = z ◦ φ = φ˜(z)
so φ ∈ OD(D). Conversely, for every φ ∈ OD(D) one obtains a morphism of ringed spaces ˜ (φ, φ):(D, OD) → (C, OC) where φ˜ is defined by pull-back. Hence the holomorphic functions on (D, OD) correspond exactly to the morphisms (of ringed spaces) D → C. 2 (2) Consider the holomorphic function (z, w) 7→ zw on C . Then V (zw) = (N, OV ) with
2 N = N(zw) = {(z, w) ∈ C ; z = 0 or w = 0}
and ∼ OV,(a,b) = C{z − a, w − b}/hzwi. For example, at the point (0, 1) we get, since the function w is a unit in a neighborhood of (0, 1), ∼ ∼ OV,(0,1) = C{z, w − 1}/hzwi = C{z, w − 1}/hzi = C{w − 1}.
2 2 3 2 3 (3) Let U = C , let f(z, w) = w −z and consider the local model V = V (w −z ), usually 2 3 called Neil’s parabola. In order to determine the stalk OV,(0,0) = C{z, w}/hw − z i we use the map 2 3 Φ: C{z, w} → C{t}, z 7→ t , w 7→ t .
2 3 Claim. The map Φ induces an injective morphism C{z, w}/hw −z i → C{t} whose image is X k akt ; ak ∈ C, a1 = 0 . k≥0
Proof. It is clear that w2 − z3 ∈ Ker(Φ) and it is easily seen that X k akt ; ak ∈ C, a1 = 0 k≥0
10 is the image of Φ, so we only have to prove that Ker(Φ) ⊂ hw2 − z3i. Let f(z, w) ∈ Ker(Φ) and write
2 3 f(z, w) = g(z, w)(w − z ) + a0(z) + a1(z)w.
This is possible because of the identity w2 = (w2 − z3) + z3. We have to show that a0(z) + a1(z)w = 0. Suppose that this is not the case, i.e. a0(z) + a1(z)w 6= 0. Then we can write
k a0(z) + a1(z)w = z (b0(z) + b1(z)w),
where (b0(0), b1(0)) 6= (0, 0). If b0(0) 6= 0 this implies
2k Φ(a0(z) + a1(z)w) = t (b0(0) + terms of order > 0) 6= 0,
and if b1(0) 6= 0 we get
2k 2 3 Φ(a0(z) + a1(z)w) = t (b0(t ) + b1(0)t + terms of order ≥ 5) 6= 0,
which in both cases contradicts the assumption f(z, w) ∈ Ker(Φ). k (4) Let k ∈ N and consider the holomorphic function z 7→ z . Then
k k V (z ) = ({0}, C{z}/hz i)
is a one point ringed space often called a k−fold point. Obviously
k k−1 OV = C{z}/hz i = C ⊕ C ε ⊕ · · · ⊕ C ε ,
where εk = 0 so V (zk) is reduced if and only if k = 1. The underlying continuous k−1 function of a holomorphic function a0 + a1 ε + ··· + ak−1 ε is {0} → C, 0 7→ a0.
Notice that any local model in C having the point 0 as an underlying topological space is of the above form.
1.4 Complex spaces
Definition 1.4.1 (Complex space) A complex space is a ringed space X = (|X|, OX ) satisfying the following conditions:
(i) |X| is Hausdorff.
(ii) For each x ∈ |X| there exists an open neighborhood U ⊂ |X| such that the ringed space (U, OX |U) is isomorphic to a local model.
The sheaf OX is called the structure sheaf of X. Again, the elements of OX (U) with U ⊂ X open, are called holomorphic functions on U.
Remark. Let X be a complex space and let V be an open subset of |X|. Put OV := OX |V . Then (V, OV ) is a complex space. Such spaces are called open complex subspaces.
11 Remark. Let X be a complex space. For all x ∈ X the stalk OX,x is isomorphic to an algebra C{z1, . . . , zn}/a where a ⊂ C{z1, . . . , zn} is a finitely generated ideal. Such algebras are called analytic.
Definition 1.4.2 Let X and Y be complex spaces. Morphisms [resp. isomorphisms] X → Y are called holomorphic [resp. biholomorphic] maps. A biholomorphic map X → X will often be called an automorphism of X.
Notation. The set of holomorphic maps X → Y will be denoted by Hol(X,Y ) and the set of automorphisms of X will be denoted by Aut(X).
Remark. In proposition 1.6.1 we will see that the holomorphic functions on a complex space can be identified with the holomorphic maps X → C.
Examples.
(1) Every local model is a complex space.
n (2) Let D be an open set in C . Then every finitely generated ideal I ⊂ OD determines a complex space (X, OX ) with
X = Supp(OD/I ) and OX := (OD/I )|X.
(3) Let M be a complex manifold of dimension n and let OM be the sheaf of holomorphic functions on M. Then (M, OM ) is a ringed space which is locally isomorphic to a local n model of the form (D, OD) with D an open subset of C , so (M, OM ) is a complex space. Moreover, it is easily seen that holomorphic maps between complex manifolds in the ordinary sense are the same as the holomorphic maps in the sense of complex spaces. In other words the category of complex manifolds is a full subcategory of the category of complex spaces.
2 (4) Consider the complex spaces V1 := V (z) and V2 := V (z ). Both spaces have {0} as underlying topological space, and
OV1 = C{z}/hzi = C
and 2 OV2 = C/hz i = C + C ε, 2 where ε = 0. Therefore we see that V2 is not reduced since the reduction of the zero function and the reduction of ε are both the zero function, but ε is not the zero function. Let us now consider some sets of holomorphic maps. In the sequel f : {0} → {0} will denote the identity map; it is the unique (continuous) map {0} → {0}.
12 (a) Hol(V1,V1).
Since f∗OV1 = OV1 and since every C–algebra homomorphism maps 1 onto 1, we get
Hol(V1,V1) = {φ: C → C; φ is a C–algebra homomorphism} = {idC}.
In particular we get Hol(V1,V1) = Aut(V1,V1).
(b) Hol(V2,V2).
Since f∗OV2 = OV2 and since every C–algebra homomorphism φ: C⊕C ε → C⊕C ε is determined by its values on the basis 1 and ε and since φ must map 1 onto 1 and ε onto λ ε for some λ ∈ C it follows that
Hol(V2,V2) = {φλ : C ⊕ C ε → C ⊕ C ε; λ ∈ C},
where φλ(a + b ε) = a + λb ε. It is easily seen that φλ ◦ φµ = φλµ and that φλ is an automorphism if and only if λ 6= 0. Hence a group isomorphism
∼ ∗ Aut(V2) = C .
(c) Hol(V1,V2).
Since f∗OV1 = OV1 and since each C–algebra homomorphism φ: C⊕C ε → C must map ε to zero, the set Hol(V1,V2) consists only of the map φ(a + b ε) = a.
(d) Hol(V2,V1). As above we get
Hol(V2,V1) = {φ: C → C ⊕ C ε; φ is a C–algebra homomorphism}
which consists only of the natural inclusion of C into C ⊕ C ε. (e) From proposition 1.6.1 below we will see that ∼ Hol(Vj, C) = OVj (Vj) = OVj
for j = 1, 2.
1.5 Closed complex subspaces In this subsection we generalize the method of constructing a local model in a domain n D ⊂ C out of a finitely generated ideal I ⊂ OD.
Let (X, SX ) be a ringed space and let I ⊂ SX be an arbitrary ideal.
Definition 1.5.1 The set
N(I ) := {x ∈ X; Ix 6= SX,x} is called the zero set of I in X.
13 Remark. We have
N(I ) = {x ∈ X; Ix ⊂ mSX,x } = {x ∈ X;(SX /I )x 6= 0} = {x ∈ X; 1x 6∈ Ix}.
This implies N(I ) = Supp(SX /I ). Since the set X \ N(I ) = {x ∈ X; 1x ∈ Ix} is open in X, it follows that N(I ) is closed in X.
Definition 1.5.2 Let s1, . . . , sk ∈ SX (X). The zero set of the ideal I := hs1, . . . , ski is often called the zero set of the sections s1, . . . , sk and is usually denoted by N(s1, . . . , sk).
Remarks. (i) We obviously have
N(s1, . . . , sk) = {x ∈ X;(s1)x,..., (sk)x ∈ mSX,x } = N(s1) ∩ · · · ∩ N(sk).
(ii) Since sx ∈ mSX,x is equivalent to s(x) = 0, we get N(s) = {x ∈ X; s(x) = 0}.
Lemma 1.5.3 Every proper ideal I ⊂ SX defines a ringed space (Y, SY ), where Y := N(I ) and SY := (SX /I )|Y . Let ι: Y → X be the canonical inclusion and let ˜ι: SX → SX /I be the quotient map. Then the map (ι, ˜ι):(Y, SY ) → (X, SX ) is a morphism of ringed spaces.
Proof. One only has to notice that ι∗SY = SX /I .
Definition 1.5.4 The ringed space (Y, SY ) is called the subspace of (X, SX ) defined by I .
Lemma 1.5.5 Let (X, OX ) be a complex space and let I ⊂ OX be a finitely generated ideal. Then the ringed space (Y, OY ) defined by I is a complex space and the canonical inclusion (ι, ˜ι) is holomorphic.
Proof. One can assume that (X, OX ) is a local model in which case the result is obvious.
Definition 1.5.6 A complex space (Y, OY ) of the kind described in lemma 1.5.5 is called a closed complex subspace of X.
Remarks. (i) In the sequel when we refer to a complex subspace or simply a subspace of a complex space it will mean a closed complex subspace.
(ii) Let Y be a complex subspace of a complex space X. Then Y is both open and closed (as a complex subspace) if and only if Y is a union of connected components of X and OY = OX |Y .
14 n Example 1. Let D be a domain in C . Clearly, every local model V = V (f1, . . . , fk) with f1 . . . , fk ∈ O(D) is a complex subspace of D.
The following theorem is a fundamental result of analytic geometry. It is most often referred to as Oka’s coherence theorem.
n Theorem 1.5.7 (Oka) For every non-negative integer n the structure sheaf OC is a coherent sheaf of rings.
Proof. See [GrRe3].
Corollary 1.5.8 The structure sheaf of every complex space is a coherent sheaf of rings.
Proof. Let X be a complex space. Since coherence is a local property we may assume that X is a local model defined by holomorphic functions f1, . . . , fm in a domain D of n some C . The structure sheaf OX is the restriction of OD/I to X where I is the ideal generated by f1, . . . , fm. By Oka’s coherence theorem the sheaf of rings, OD, is coherent and the ideal sheaf I is OD–coherent since it is a finitely generated subsheaf of OD. Consequently OX is a coherent sheaf of rings.
Remark. For any complex space X there is a one-one correspondence between the com- plex subspaces of X and the coherent ideals of the structure sheaf OX .
Definition 1.5.9 Let X be a complex space and let Y be a subspace of X defined by a coherent ideal I . Then Y is called a hypersurface of X if Iy is a principal ideal generated by a non zero-divisor in OX,y for all y in Y .
Definition 1.5.10 Let (X, SX ) be a ringed space. The sheaf of nilpotent sections of SX is called the nilradical of SX and denoted by NS.
Remark. Since CX is without nilpotent sections, NS is contained in the kernel of the morphism Red: SX → CX .
For complex spaces we have much stronger results. These are very deep and fundamental results in analytic geometry.
Theorem 1.5.11 Let (X, OX ) be a complex space and let N be the nilradical of OX . Then
(i) (Hilbert’s Nullstellensatz) Ker(Red) = N ,
(ii) (Cartan’s Coherence Theorem) N is coherent.
Proof. See [GrRe3].
15 Remark. For a ringed space (X, SX ) the kernel of Red: SX → CX is in general bigger than the nilradical NS. If for example ({x}, S ) is the one point ringed space with S the algebra of convergent power series in one variable, then Ker(Red) is the maximal ideal of S but the nilradical of S is zero.
The following result is an obvious consequence of theorem 1.5.11.
Corollary 1.5.12 The reduction of a complex space is a complex space.
Example 2. Let X be a complex space. Then Xred is a complex subspace of X but Xred is not an open subspace of X unless X is reduced.
Definition 1.5.13 Let X and Y be complex spaces, let Z be a complex subspace of X and let ι: Z → X denote the canonical inclusion. For every holomorphic map F : X → Y one defines a holomorphic map by
F |Z := F ◦ ι: Z → Y.
It is called the restriction of the map F to the subspace Z.
n n 1.6 The bijection Hol(X, C ) → OX (X) n ˜ n Let X be a complex space, let (f, f):(X, OX ) → (C , OC ) be a holomorphic map and n n let z1, . . . , zn ∈ OC (C ) denote the coordinate functions. Then we get for every j: ˜ n −1 n f(zj) ∈ (f∗OX )(C ) = OX (f (C )) = OX (X). Hence the mapping
n n ˜ ˜ ˜ Hol(X, C ) → (OX (X)) , (f, f) 7→ (f(z1),..., f(zn)).
n n Proposition 1.6.1 The mapping Hol(X, C ) → (OX (X)) is a bijection. Moreover, if X is reduced then (f˜(z1),..., f˜(zn)) = (z1 ◦ f, . . . , zn ◦ f).
Proof. We first show that the mapping is injective. ˜ n ˜ Let (f, f), (g, g˜) ∈ Hol(X, C ) and suppose that f(zj) =g ˜(zj) for all 1 ≤ j ≤ n. Then lemma 1.2.5 implies
zj(f(x)) = f˜(zj)(x) =g ˜(zj)(x) = zj(g(x)) for all x ∈ X. and consequently f = g. n In order to prove f˜ =g ˜ we fix a point x ∈ X and set a := f(x) = g(x) ∈ C . Since the n n ˜ analytic algebras OC ,a and OC ,0 are isomorphic we can assume a = 0. Then fx andg ˜x n are C–algebra morphisms OC ,0 → OX,x. Now, OX,x is Noetherian, z1, . . . , zn generate n ˜ the maximal ideal of OC ,0 and fx andg ˜x coincide on these generators so corollary A.3.11 tells us that f˜ =g ˜. Hence the injectivity of the mapping.
16 In order to prove the surjectivity of the mapping choose f1, . . . , fn ∈ OX (X). Suppose m first that X is a local model in a domain D in C and that f1, . . . , fn are induced by F1,...,Fn ∈ OD(D) = O(D), more precisely
fj := Fj ◦ ι, for all 1 ≤ j ≤ n, where ι: X → D is the inclusion. Consider the holomorphic map
n F : D → C , y 7→ (F1(y),...,Fn(y))
n and let f : D → C be its restriction to X, i. e. f := F ◦ ι.
For every 1 ≤ j ≤ n we have F˜(zj) = zj ◦ F = Fj and consequently
f˜(zj) = ˜ι ◦ F˜(zj) = ˜ι(Fj) = Fj ◦ ι = fj.
Now consider an arbitrary complex space X. First we choose an open covering {Uα}α∈A with each Uα biholomorphic to a local model as above and then we apply the construction to f1|Uα, . . . , fn|Uα for every α ∈ A. Since we have already shown that the map
n Hol(Uα ∩ Uβ, C ) → OX (Uα ∩ Uβ)
n n is injective, two maps fα : Uα → C and fβ : Uβ → C , which are obtained by this con- ˜ struction, coincide on Uα ∩ Uβ. Hence, the mappings(fα, fα) glue together to define a map n (f, f˜) ∈ Hol(X, C ), such that f˜(zj) = fj for all 1 ≤ j ≤ n.
1.7 Geometry of complex spaces A germ of a complex space is a pair (X, x) where X is a complex space and x is a point in X. Germs of complex spaces are often referred to as space germs. Let (X, x) and (Y, y) be two space germs. The germs of holomorphic maps at x from open neighborhoods of x into Y which map x to y are called morphisms of space germs. Let U be an open neighborhood of x in X and f : U → Y be a holomorphic map with f(x) = y. Then its germ at x will be denoted by fx and one writes fx :(X, x) → (Y, y). It is easy to see that fx is an isomorphism if and only if there exists an open neighborhood V of x and an open neighborhood W of y such that f induces a biholomorphic map between V and W . With each space germ (X, x) there is associated the analytic algebra OX,x and with each morphism fx :(X, x) → (Y, y) there is associated the algebra morphism ˜ fx : OY,y → OX,x.
Hence a contravariant functor from the category of space germs to the category of analytic algebras.
Theorem 1.7.1 The above contravariant functor is an antiequivalence.
17 Proof. The proof consists of verifying the three following statements.
• For every analytic algebra A there exists a space germ (X, x) such that OX,x is isomorphic to A.
• For every algebra morphism φ: OY,y → OX,x there exists a morphism fx :(X, x) → (Y, y) such that φ = f˜x.
• If fx and gx are two morphisms from (X, x) to (Y, y) such that f˜x =g ˜x then fx = gx.
For the details see [Gro1]. Compare also [Fis] and [Re].
The above theorem enables us to translate local geometric properties into algebraic prop- erties and vice versa.
Definition 1.7.2 Let X = (|X|, OX ) be a complex space. A point x ∈ X is called
(i) smooth or non-singular, if OX,x is regular, i. e. if it is isomorphic to C{z1, . . . , zn} for some n ∈ N. Otherwise x is called singular.
(ii) irreducible, if OX,x is an integral domain.
(iii) reduced, if OX,x is reduced.
(iv) normal, if OX,x is reduced and integrally closed in its full ring of fractions.
Remark. A complex space X is reduced if and only if every point of X is reduced.
Examples. (1) Let M be a complex manifold of dimension n and let x ∈ M. Then by choosing a
∼ n chart centered at x we get OM,x = OC ,0. This shows that every point in M is smooth. (2) Let V = V (zw). Clearly the analytic algebra
OV,(0,0) = C{z, w}/hzwi is not an integral domain so the point (0, 0) is not irreducible. For the same reason the stalk OV,(0,0) cannot be isomorphic to C{z1, . . . , zn} for any n ∈ N0. Hence, the point (0, 0) is singular.
(3) Let V = V (w2 − z3). Every point in V is a smooth point of V except (0, 0) and from example 3 in subsection 1.3 we see that (0, 0) is an irreducible point of V .
n (4) The complex space (D, OD) where D denotes a domain in C is everywhere reduced. Thus, a complex manifold is everywhere reduced, too.
(5) In the space V (zk) the point 0 is not reduced if k ≥ 1.
Remark. Every non–irreducible point is singular.
Definition 1.7.3 Sing(X) := {x ∈ X; x singular} is called the singular locus of X.
18 Example. For X = V (zw) we have Sing(X) = {(0, 0)}.
Lemma 1.7.4 Let X be a complex space. Then X \ Sing(X) is an open subspace of X. Moreover, X \ Sing(X) is a complex manifold.
Proof. Let x ∈ X be a smooth point. Then there exists an isomorphism φx : OX,x → n OC ,0 for some n ∈ N0 and by 1.7.1 there exists an open neighborhood of x in X which n is biholomorphic to an open neighborhood of 0 in C . Consequently every point in that neighborhood of x is non-singular.
Definition 1.7.5 The Zariski–tangent space or simply the tangent space of X at x is the vector space 2 ∗ TxX := (mx/mx) . The vector space ∗ 2 Tx X := mx/mx is called the Zariski–cotangent space or the cotangent space of X at x.
Examples.
n (1) Let D be a domain in C with 0 ∈ D. Then we have OD,0 = C{z1, . . . , zn}. Moreover, the maximal ideal of C{z1, . . . , zn} is given by
m = hz1, . . . , zni.
It follows that 2 m/m = C ε1 ⊕ · · · ⊕ C εn, 2 where εj is the image of zj in m/m for 1 ≤ j ≤ n. It then follows that the image of an element f of m is equal to ∂f ∂f (0) ε1 + ··· + (0) εn . ∂z1 ∂zn
2 n If we identify m/m with C via the linear isomorphism determined by
εj 7→ (0,..., 1,..., 0)
n for 1 ≤ j ≤ n, the projection map m → C is given by
∂f (0) ∂z1 . f 7→ . . ∂f (0) ∂zn
Now suppose that f1, . . . , fm generate the maximal ideal m. Then from the above considerations it follows that ∂f rk i (0) = n. ∂zj 1≤i≤m 1≤j≤n
19 k (2) Let X = V (z ) and let m denote the maximal ideal of OX,0. Then we have
k k−1 OX,0 = C{z}/hz i = C ⊕ C ε ⊕ · · · ⊕ C ε with εk = 0 and k k−1 m = hzi/hz i = C ε ⊕ · · · ⊕ C ε . k 2 2 k−1 Moreover, we see from ε = 0 that m = C ε ⊕ · · · ⊕ C ε . (This is understood to be zero if k = 1.) so 0 , if k = 1 m/m2 = . C ε , if k ≥ 2 (3) Let M be a complex manifold of dimension n, let x ∈ M and let m denote the
∼ n maximal ideal of OM,x. By choosing a chart centered at x we get OM,x = OC ,0 and consequently 2 dimC(m/m ) = dimC M. n (4) Let D ⊂ C be a domain and let V = V (f1, . . . , fk) be the local model given by f1, . . . , fk ∈ O(D). For every x = (x1, . . . , xn) in N(f1, . . . , fk) we have OV,x = OD,x/Jx where J is the ideal generated by the functions f1, . . . , fk and
mx = h(z1 − x1),..., (zn − xn)i/Jx
is the maximal ideal of OV,x. Consequently the images of the elements z1 −x1, . . . , zn − 2 xn in the cotangent space mx/mx generate it but they need not be linearly independent. Hence 2 mx/mx = C ε1 + ··· + C εn, 2 where εj denotes the image of zj −xj in mx/mx and in particular we get the inequality 2 dimC(mx/mx) ≤ n. (∗)
Definition 1.7.6 Let X and Y be complex spaces. (i) A holomorphic map f : X → Y is called an embedding, if there exist a complex subspace Z of Y and a biholomorphic map g : X → Z, such that the diagram
f X / Y AA O AA ι g AA A Z is commutative, where ι denotes the canonical inclusion
(ii) We say that X can be embedded into Y at x ∈ X, if there exist an open neighborhood U of x in X, an open subset V of Y and an embedding f : U → V . As an abbreviation we say that f :(X, x) → (Y, f(x)) is a local embedding. In this case we also say that the space germ (X, x) can be embedded into the space germ (Y, f(x)).
(iii) The embedding dimension of X at x is the minimal integer k such that X can be k embedded into C at x. It will be denoted by embx X.
20 Proposition 1.7.7 A holomorphic map (f, f˜): X → Y is an embedding if and only if the following two properties are satisfied (i) f(X) is a closed subset of Y and the induced map X → f(X) is a homeomorphism. ˜ ˜ (ii) The sheaf morphism f : OY → f∗OX is surjective and the ideal Ker f ⊂ OY is finitely generated.
Proof. A proof can be found in [GrRe3].
n Proposition 1.7.8 Let X = V (f1 . . . , fk) be a local model in a domain D ⊂ C and let x ∈ X. Then we have ∂fi embx V + rk (x) = n. ∂zj
Proof. Let x ∈ X and put ∂fi e := embx X and r := rk (x) . ∂zj First we prove e ≤ n − r. Without loss of generality we may assume that ∂f det i (x) 6= 0. ∂zj 1≤i,j≤r From the implicit function theorem it follows that there exist an open neighborhood U = n−r r V × W ⊂ D, with V ⊂ C and W ⊂ C , and functions g1, . . . , gr ∈ O(V ) such that the set N := {y ∈ U; f1(y) = ··· = fr(y) = 0} is the graph of the holomorphic map
r V → C , (y1, . . . , yn−r) 7→ (g1(y1, . . . , yn−r), . . . , gr(y1, . . . , yn−r)). If we equip N with the reduced complex structure, it follows that N is biholomorphically n−r equivalent to V ⊂ C . Since X ∩ (V × W ) is a complex subspace of N, this implies e ≤ n − r.
e Now we show r ≥ n − e. By definition of e there exists an embedding (X, x) → (C , 0) and the corresponding algebra morphism C{z1, . . . , ze} → OX,x is surjective. Thus for 0 0 any choice of generators (g1)x,..., (ge)x ∈ OX,x for the maximal ideal mx of OX,x we can choose holomorphic functions g1, . . . , ge in an open neighborhood of x in D such that
0 (gj|X)x = (gj)x for all 1 ≤ j ≤ e. Then we get
0 0 n OC ,x/h(f1)x,..., (fk)x, (g1)x,..., (ge)xi = OX,x/h(g1)x,..., (ge)xi = C and it follows that the Jacobi matrix of f1, . . . , fk, g1, . . . , ge has rank n. Therefore, the Jacobian of f1, . . . , fk has rank at least n − e and that proves the assertion.
Proposition 1.7.9 A holomorphic map f :(X, x) → (Y, f(x)) is a local embedding if and ˜ only if fx : OY,f(x) → OX,x is surjective.
21 Proof. For a proof see [Fis].
˜ Remark. Note that fx : OY,f(x) → OX,x is surjective if and only if the induced linear map TxX → Tf(x)Y is injective.
Example. Consider the complex space V = V (z2 − w). From proposition 1.7.8 we see that the embedding dimension of V at the origin is 1. Since the maximal ideal m0 of OV,0 is generated by z + hz2 − wi and w + hz2 − wi, we get the following two generators of 2 m0/m0: 2 2 2 2 (z + hz − wi) + m0 and (w + hz − wi) + m0. Because of 2 2 2 2 w + hz − wi = z + hz − wi ∈ m0 2 2 we have (w + hz − wi) + m0 = 0. Notice that the map V → C,(z, w) 7→ w, is not injective near 0 while the map (z, w) 7→ z defines even a global embedding V → C. At every point a = (a1, a2) ∈ V \{(0, 0)} the maps (z, w) 7→ w and (z, w) 7→ z define local embeddings. This is due to the fact that both 2 2 2 2 ((z − a1) + hz − wi) + ma and ((w − a2) + hz − wi) + ma 2 are non-zero elements in ma/ma.
The following proposition will make this correspondence clearer.
Proposition 1.7.10 Let X be a complex space, let x be a point in X and let m denote 2 the maximal ideal of OX,x. Then embx X = dimC(m/m ).
n Proof. Put n := embx X. Then there exists an embedding ι:(X, x) → (C , 0) so we n can assume that X is a local model in C given by equations f1 = ··· = fk = 0 in a n neighborhood of the origin. Let I be the ideal generated f1, . . . , fk in OC ,0, let n be the n n maximal ideal of OC ,0 and let m be the maximal ideal of OX,0 = OC ,0/I. The algebra morphism
˜ n fx : OC ,0 → OX,0 is surjective and it is easy to see that it induces a surjective morphism from n onto m and hence a surjective C−linear mapping
n/n2 → m/m2.
2 The kernel of this C−linear mapping is obviously the image of I in n/n .
From proposition 1.7.8 we see that the Jacobian matrix of (f1, . . . , fk) is zero at the 2 origin which means that the ideal I is contained in n . Consequently the the above C– linear mapping is an isomorphism and in particular the two vector spaces are of the same dimension. Since we have already shown that