Startseite Real structures on primary Hopf surfaces
Artikel Open Access

Real structures on primary Hopf surfaces

  • Zahraa Khaled EMAIL logo
Veröffentlicht/Copyright: 6. Dezember 2024

Abstract

The first goal of this article is to give a complete classification (up to Real biholomorphisms) of Real primary Hopf surfaces ( H , s ) , and, for any such pair, to describe in detail the following naturally associated objects : the group Aut h ( H , s ) of Real automorphisms, the Real Picard group ( Pic ( H ) , s ˆ * ) , and the Picard group of Real holomorphic line bundles Pic R ( H ) . Our second goal is the classification of Real primary Hopf surfaces up to equivariant diffeomorphisms, which will allow us to describe explicitly in each case the real locus H ( R ) = H s and the quotient H s .

MSC 2010: 32J15; 32M18

1 Introduction

Let X be a complex manifold, and let J be the (integrable) almost holomorphic structure on its underlying differentiable manifold X defining its complex structure. We will denote by X ¯ the complex manifold defined by J . Note that the data of an anti-holomorphic isomorphism X X is equivalent to the data of a biholomorphism X X ¯ .

A Real structure (in the sense of Atiyah) on X is an anti-holomorphic involution of X [1,4,8]. A Real complex manifold is a pair ( X , s ) consisting of a complex manifold and a Real structure on it.

The theory of Real complex manifolds originates from algebraic geometry, see [1, section 1]: a smooth projective variety X P C n defined by a system of homogeneous polynomial equations with real coefficients has a natural Real structure induced by the conjugation P C n P C n . An ample literature is dedicated to this theory in algebraic geometric framework [4,8]. On the other hand, not much is known on the classification of Real structures on non-algebraic manifolds. An important contribution in this direction is Paola Frediani’s article [3], which is dedicated to the (holomorphic and topological) classification of Real Kodaira surfaces. In this article, we treat similar problems but for primary Hopf surfaces.

The real locus of a Real complex manifold ( X , s ) is just the fixed-point locus X s (also denoted X ( R ) if s has been fixed) of its Real structure.

Let ( X , s ) , ( Y , σ ) be Real complex manifolds. A biholomorphism f : X Y is called Real (or compatible with the Real structures) if σ f = f s . The fundamental problem of the theory is the classification of Real complex manifolds up to Real biholomorphisms.

The group of real biholomorphisms of a Real complex manifold ( X , s ) is the subgroup

Aut ( X , s ) { f : X X f biholomorphism , f s = s f }

of the biholomorphism group Aut h ( X ) .

Let ( M , s ) be a differentiable manifold endowed with an involution s and E be a complex vector bundle on M . We recall [1, Section 1] that

Definition 1.1

A Real structure on E is a fibrewise anti-linear s -covering isomorphic involution φ : E E . A Real bundle on ( M , s ) is pair ( E , ϕ ) consisting of a complex bundle E on M and a Real structure ϕ on E .

Let ( X , s ) be a Real complex manifold.

Definition 1.2

A Real holomorphic bundle on X is a pair ( E , ϕ ) , where E is a holomorphic bundle on X and ϕ an anti-holomorphic Real structure on E .

Let E be a holomorphic bundle of rank r on X . The pull-back s * ( E ¯ ) has a natural structure of a holomorphic bundle on X [7, section 1.2]: it is just the pull-back of E ¯ , regarded as a holomorphic bundle on X ¯ , via the holomorphic map s : X X ¯ . The map [ E ] [ s * ( E ¯ ) ] defines a natural involution on the set of isomorphism classes of holomorphic bundles on X . In particular, for r = 1 , we obtain an involution

s ¯ * : Π c ( X ) Π c ( X ) , s ¯ * ( [ L ] ) [ s * ( L ¯ ) ]

on the Picard group of X ; this involution is an anti-holomorphic group isomorphism, so ( Π c ( X ) , s ¯ * ) becomes a Real complex Lie group.

The aforementioned definitions allow us to associate to a compact, connected Real complex manifold two natural invariants constructed using holomorphic line bundles:

  • The group Π c R ( X ) of isomorphism classes of Real holomorphic line bundles on X .

  • The Real complex Lie group ( Π c ( X ) , s ¯ * ) .

Note that one has an obvious comparison real Lie group morphism

Π c R ( X ) Π c ( X ) ( R ) , [ L , ϕ ] [ L ] ,

which is always injective and is an isomorphism when X ( R ) .

The goals of this article are as follows:

  1. To give complete classification of Real primary Hopf surfaces (up to Real biholomorphisms) with an explicit description of the set of isomorphism classes.

  2. To describe explicitly, for any Real primary Hopf surface ( H , s ) , the following naturally associated objects:

    1. its automorphism group Aut ( H , s ) Aut h ( H ) .

    2. its Real Picard group ( Π c ( H ) , s ¯ * ) of holomorphic line bundles, and its Picard group Π c R ( H ) of Real holomorphic line bundles.

  3. To classify differential-topologically the Real primary Hopf surfaces, and, for any Real primary Hopf surface ( H , s ) , to describe explicitly the fixed-point (real) locus H s and the quotient H s .

For (G1), recall first that [2]:

Definition 1.3

A primary Hopf surface is a compact complex surface H whose universal covering is biholomorphic to W C 2 \ { 0 } , and whose fundamental group is isomorphic to Z .

From this definition, it follows that any primary Hopf surface is biholomorphic to a quotient of the form

H f = W f ,

where f is the cyclic group generated by a biholomorphism f Aut h ( W ) . By a fundamental theorem of Kodaira [5, Theorem 30, p. 696], it follows that any primary Hopf surface is biholomorphic to W f , where f is a biholomorphism of the form

f ( z , w ) = ( α z + λ w n , β w ) ,

where

0 < α β < 1 , n N * , λ ( α β n ) = 0 .

If the coefficients of f are real, the standard conjugation c : W W will obviously descend to a Real structure on H f . We will see that there exists interesting classes of Real primary Hopf surfaces which are not of this type. Moreover, there exists Real primary Hopf surfaces defined by holomorphic contractions f whose coefficients are not real.

Note first that Kodaira’s theorem does not give a precise classification of primary Hopf surfaces, because it is not clear under which conditions the surfaces associated with two 4-tuples ( α , β , λ , n ) , ( α , β , λ , n ) as above are biholomorphic. Following [11], we introduce five classes of holomorphic contractions:

(1) I V f : W W f z w = α z α w 0 < α < 1 , I I I f : W W f z w = δ r z δ w r N 2 , 0 < δ < 1 , I I a f : W W f z w = δ r z + w r δ w r N 2 , 0 < δ < 1 , I I b f : W W f z w = α z + w α w 0 < α < 1 , I I c f : W W f z w = α z δ w 0 < α < 1 0 < δ < 1 , α δ r r Z .

The map

(2) I V I I I I I a I I b I I c f [ H f ] ,

which assigns to a holomorphic contraction f the biholomorphism class of the corresponding Hopf surface H f is surjective, but not injective. Indeed, the contractions f , f I I c associated with the pairs ( α , δ ) , ( α , δ ) = ( δ , α ) are biholomorphic. Note that this exception to injectivity is not mentioned in [11]. In fact in [11], the class I I c is defined by imposing the additional condition α < δ . Unfortunately with this restrictive definition of I I c , one loses the surjectivity of the map (2), because biholomorphism types of Hopf surfaces associated with pairs ( α , δ ) satisfying

(3) 0 < α = δ < 1 , α δ r r Z

will not belong to its image. This remark is important for our purposes because precisely in the subclass of I I c defined by (3) – the subclass which is omitted in [11] – we will find contractions f for which H f admits Real structures although the coefficients of f are not real.

Our first step in the classification of Real structures on primary Hopf surfaces is to divide them into two classes: a Real structure ϕ on H f will be called even (odd) if it admits a lift ϕ ˆ : W W with ϕ 2 = id W (respectively ϕ 2 = f ). The even (odd) Real structures are classified by Theorem 3.6 (respectively Theorem 3.15).

Concerning (G2), we will give explicit descriptions of the automorphism group Aut ( X , s ) of all Real primary Hopf surface. For instance, when f I V with negative coefficient α , we obtain Aut ( H f , s f ) Spin c ( 3 ) , where s f denotes the canonical odd Real structure on H f (see Corollary 4.4).

Our results concerning the Real complex group ( Π c ( X ) , s ¯ * ) and the group Π c R ( X ) of a Real primary Hopf surface ( X , s ) are as follows (see Proposition 4.8):

  1. ( Π c ( X ) , s ¯ * ) is always isomorphic to ( C * , ¯ ) .

  2. The canonical monomorphism Π c R ( X ) Π c ( X ) ( R ) = R * is an isomorphism if ( X , s ) is even and identifies Π c R ( X ) with R > 0 if ( X , s ) odd.

Our results for the goal (G3) give a complete differential topological classification of Real primary Hopf surfaces (see Theorems 5.1 and 5.11 and Remark 5.12). The final result is:
  • Any even Real primary Hopf surface is equivariantly diffeomorphic to either

    ( S 1 × S 3 , ( ζ , ( u , v ) ) ( ζ , ( u ¯ , v ¯ ) ) )

    or

    ( S 1 × S 3 , ( ζ , ( u , v ) ) ( ζ , ( u ¯ , ζ v ¯ ) ) ) .

  • Any odd Real primary Hopf surface is equivariantly diffeomorphic to

    ( S 1 × S 3 , ( ζ , Z ) ( ζ , Z ) ) .

Taking into account the results of Section 5.3.1, this shows that the equivariant differential topological type of a Real primary Hopf surface is determined by the type (even or odd) and the orientability of the real locus.

The main idea in the proof of this classification result is: for a contraction f I V I I I I I a I I b I I c with Real coefficients and positive diagonal coefficients, we construct one-parameter group of diffeomorphisms ( f t ) t R of W acting freely on W such that f = f 1 . Moreover, we also construct a compact three-dimensional submanifold Σ W which is transversal to the orbits of this group and can be identified to S 3 via a diffeomeorphism which commutes with the conjugation and the involutions ( z , w ) ( ± z , ± w ) .

Finally will show that:

  1. The real locus X s of an even Real primary Hopf surface ( X , s ) is either a torus, or a Klein bottle, whereas the real locus of an odd Real primary Hopf surface is always empty.

  2. The quotient X s associated with a Real primary Hopf surface ( X , s ) is always homeomorphic to S 1 × S 3 , and we describe the position of the fixed-point locus X s in this quotient.

Note that, by the equivariant slice theorem, for any Real complex surface ( X , s ) , the quotient X s is a topological 4-manifold.

2 Holomorphic and anti-holomorphic automorphisms

A fundamental role in this article will be played by the results of Wehler on the classification of primary Hopf surfaces and their automorphism group. In this section, we review these results and we continue with the classification of the anti-holomorphic automorphisms of primary Hopf surfaces.

2.1 Wehler’s classification of primary Hopf surfaces

A precise classification of primary Hopf surfaces – with explicit descriptions of the automorphism groups – has been given by Wehler [11]. His result can be formulated as follows:

Theorem 2.1

Consider the sets I V , I I I , I I a , I I b , I I c Aut h ( W ) defined in formula (1).

  1. For every primary Hopf surface H, there exists f I V I I I I I a I I b I I c such that H H f .

  2. For f , f I V I I I I I a I I b I I c , we have H f H f if and only if either f = f , or f and f belong to I I c , and the corresponding coefficients α , δ , α , δ satisfy α = δ , δ = α .

  3. For any f, the group Aut h ( W ) f of holomorphic automorphisms of W commuting with f is given by the table below:

    The class of f Aut h ( W ) f
    I V GL ( 2 , C )
    I I I z w a z + b w r d w a C * , d C * , b C
    I I a z w a r z + b w r a w a C * , b C
    I I b z w a z + b w a w a C * , b C
    I I c z w a z d w a C * , d C *

  4. In each case, the automorphism group Aut h ( H f ) is identified with the quotient group Aut h ( W ) f f .

Therefore, the name of the class gives the dimension of the automorphism group of the corresponding surface.

Proof

We explain below only the part of Theorem 2.1 which is different from Wehler’s original statement.

(1) follows from Kodaira’s [5, Theorem 30, p. 696] quoted in the introduction taking into account that in the case λ 0 , using a linear rescaling of the first coordinate, we can assume λ = 1 . Note that the argument does not apply if one imposes the strict inequality α < δ in the definition of the class I I c .

For the statement (2) in the case omitted in Wehler’s [11, Theorem 1], i.e. when f , f belong to the class I I c as defined in formula (1):

If f , f define isomorphic Hopf surfaces, then, as explained in [11, p. 24], there exists a biholomorphism φ : C 2 C 2 with φ ( 0 ) = ( 0 ) such that f = φ f φ 1 . Taking the Jacobian matrices at 0 of the two sides, we see that the matrices defining f , f must be similar, in particular their spectra coincide. Therefore, we have either α = α , δ = δ , or α = δ , δ = α .

Conversely, if the spectra of the two diagonal matrices coincide, they are obviously conjugate by a linear automorphism of C 2 .

Concerning (3): Note that for the computation of the automorphism groups Wehler quotes Namba’s article [6]. In this article, the class of linear primary Hopf surfaces (i.e. the class of primary Hopf surfaces defined by a linear contraction) is parameterised using the set [6, p. 133]

M α t 0 β α , β , t C , 0 < α < 1 , 0 < β < 1 .

No inequality between α , β is imposed (even in the special case t = 0 ). This agrees with the convention we adopted in the definition of the class I I c .□

Remark 2.2

In [11], the case α = δ is omitted in the definition of I I c , so the exception (2) to the injectivity of the map

I V I I I I I a I I b I I c f [ H f ]

is not mentioned either.

2.2 Anti-holomorphic automorphisms of primary Hopf surfaces

We start with a general remark about topological automorphisms of Hopf surfaces:

Proposition 2.3

Let H = H f = W f be a primary Hopf surface and let π : W H be the canonical map. Let σ : H H be a homeomorphism. Then

  1. There exists a homeomorphism σ ˆ : W W such that π σ ˆ = σ π .

  2. For any such homeomorphism σ ˆ we have σ ˆ f σ ˆ 1 { f , f 1 } .

Proof

(4)

  1. The composition σ π remains a covering, and, since W is simply connected, the uniqueness theorem of the universal covering guarantees the existence of a homeomorphism σ ˆ : W W verifying the equality

    π σ ˆ = σ π .

  2. The group

    Aut H ( W ) = { g : W W g homeomorphism , π g = π }

    of topological automorphisms of the universal covering π (of deck transformations) coincides with the cyclic group f . On the other hand, the map

    g σ ˆ g σ ˆ 1

    is a group automorphism of Aut H ( W ) , so it coincides either with id Aut H ( W ) or with the automorphism g g 1 . Replacing g by f , we obtain σ ˆ f σ ˆ 1 { f , f 1 } as claimed.□

In the case when σ : H H is holomorphic or anti-holomorphic, we have a more precise result:

Proposition 2.4

Let σ : H H be a holomorphic (anti-holomorphic) automorphism of H = H f . Then

  1. There exists a a holomorphic (anti-holomorphic) automorphism σ ˆ of W such that π σ ˆ = σ π .

  2. For any such automorphism σ ˆ , we have σ ˆ f σ ˆ 1 = f .

Proof

Since π is locally biholomorphic, it follows that σ ˆ is holomorphic (respectively anti-holomorphic) if σ is holomorphic (respectively anti-holomorphic). By Hartogs theorem (applied to σ ˆ or to its composition with the conjugation automorphism), it follows that σ ˆ extends to a holomorphic (anti-holomorphic) automorphism σ ˜ of C 2 with σ ˜ ( 0 ) = 0 .

We can suppose that f I V I I I I I a I I b I I c . Any such f is a holomorphic contraction. It follows that for any w 0 W , we have

lim n f n ( w 0 ) = 0 , lim n f n ( w 0 ) =

in the end compactification W { 0 , } of W . Since σ ˆ extends to a homeomorphism σ ˜ of C 2 with σ ˜ ( 0 ) = 0 , it follows that the permutation end ( σ ˆ ) induced by σ ˆ on the set of ends { 0 , } is id { 0 , } . Therefore,

lim n ( σ ˆ f σ ˆ 1 ) n ( w 0 ) = lim n ( σ ˆ f n σ ˆ 1 ) ( w 0 ) = end ( σ ˆ ) ( lim n f n ( σ ˆ 1 ( w 0 ) ) ) = end ( σ ˆ ) ( 0 ) = 0 ,

whereas lim n f 1 ( w 0 ) = . Therefore, the case σ ˆ f σ ˆ 1 = f 1 is ruled out.□

We introduce the subclass I I c of I I c defined by

I I c f : W W f z w = α z α ¯ w 0 < α < 1 , α R .

Proposition 2.5

Let f I V I I I I I a I I b I I c . The following conditions are equivalent:

  1. The primary Hopf surface H f admits anti-holomorphic automorphisms.

  2. Either the coefficients of f are real or f I I c .

Proof

The H f admits an anti-holomorphic automorphism if and only if H f is biholomorphic to H ¯ f . On the other hand, the conjugation automorphism c : W W induces an anti-holomorphic isomorphism s : H f H f , where f c f c 1 . Recall now that, in general, the data of an anti-holomorphic isomorphism X Y are equivalent to the data of a holomorphic isomorphism X ¯ Y .

Therefore, H ¯ f H f , so H f admits an anti-holomorphic automorphism if and only if H f H f . Now note that f is obtained from f by conjugating the coefficients of the polynomial expression which defines f . On the other hand, Wehler’s classes are conjugation invariant, in particular f also belongs to I V I I I I I a I I b I I c . By the classification Theorem 2.1, it follows that H f H f if and only if either the coefficients of f and f coincide (in other words the coefficients of f are real), or f and f belong to I I c and the coefficients α ¯ , δ ¯ of f are obtained from the coefficients α , δ of f by changing the order. The latter condition is equivalent to f I I c .□

Remark 2.6

A direct proof of Proposition 2.5 can be obtained using Proposition 2.4 and the Taylor expansion of the anti-holomorphic automorphism σ ˜ of C 2 obtained by applying Hartogs theorem to the lift σ ˆ of an anti-holomorphic automorphism σ .

For a primary Hopf surface H , we denote by Ah ( H ) the set of anti-holomorphic automorphisms. If H = H f with f I V I I I I I a I I b I I c , this set can be obtained explicitly using the idea of Remark 2.6. An anti-holomorphic automorphism σ Ah ( H ) has a lift σ ˆ Ah ( W ) , which extends to an anti-holomorphic automorphism σ ˜ Ah ( C 2 ) with σ ˜ ( 0 ) = 0 . Denoting by f ˜ Aut h ( C 2 ) the extension of f , we see that the condition σ ˆ f = f σ ˆ is equivalent to σ ˜ f ˜ = f ˜ σ ˜ , which can be interpreted in terms of the Taylor expansion

σ ˜ ( z , w ) p , q N a p q z ¯ p w ¯ q , p , q N b p q z ¯ p w ¯ q

of σ ˜ at 0. By using this method, we obtain easily

Proposition 2.7

Let f I V I I I I I a I I b I I c with real coefficients. The set

Ah ( W ) f { u Ah ( W ) u f = f u }

is given by the table:

In each case, the cyclic group f acts freely on the set Ah ( W ) f , and Ah ( H f ) is identified with the quotient set Ah ( W ) f f .

The class of f Ah ( W ) f
I V z w A z ¯ w ¯ A GL ( 2 , C )
I I I z w a z ¯ + b w ¯ r d w ¯ a C * , d C * , b C
I I a z w a r z ¯ + b w ¯ r a w ¯ a C * , b C
I I b z w a z ¯ + b w ¯ a w ¯ a C * , b C
I I c \ I I c z w a z ¯ d w ¯ a C * , d C *
I I c z w a w ¯ d z ¯ a C * , d C *

3 The classification of Real primary Hopf surfaces

3.1 Even Real structures, odd Real structures on primary Hopf surfaces

We start by the following simple result:

Proposition 3.1

Let H = H f be a primary Hopf surface. Let σ : H H be an anti-holomorphic involution of H and σ ˆ : W W be a lift of σ (see Proposition 2.4). Then

  1. There exists n Z such that σ ˆ 2 = f n .

  2. The parity of n in the previous formula is well defined (depends only on σ , not on the lift σ ˆ ).

Proof

Indeed, since σ 2 = id H f it follows that σ ˆ 2 Aut H ( W ) = f . This proves the first claim.

Let now σ ˆ , σ ˆ two lifts of σ and n , n Z be the associated integers. We have σ ˆ σ ˆ 1 Aut H ( W ) , so there exists k N such that σ ˆ = σ ˆ f k . Since σ ˆ commutes with f , we obtain σ ˆ 2 = σ ˆ 2 f 2 k , so n = n + 2 k .□

Taking into account this proposition, it is natural to define:

Definition 3.2

A Real structure σ : H H on H is said to be:

  1. even, if one of the following equivalent conditions is verified:

    1. For any lift σ ˆ : W W of σ , σ ˆ 2 coincides with an even power of f .

    2. There exists a lift σ ˆ : W W of σ such that σ ˆ 2 = id W .

  2. odd, if one of the following equivalent conditions is verified:

    1. For any lift σ ˆ : W W of σ , σ ˆ 2 coincides with an odd power of f .

    2. There exists a lift σ ˆ : W W of σ such that σ ˆ 2 = f .

Example 3.1

Let f I V I I I I I a I I b I I c be with real coefficients. The standard complex conjugation c : W W induces a Real structure s f on H f , which is obviously even. The Real structure will be called the standard Real structure of H f .

Let now f I I c be given by f z w = α z α ¯ w , where 0 < α < 1 and α R . The anti-holomorphic automorphism of c : W W defined by

c z w = w ¯ z ¯

commutes with f and is involutive, so it defines an even Real structure s f on H f , which will also be called the standard Real structure on H f .

3.2 The classification of even Real structures on primary Hopf surfaces

Let E be a complex vector space of dimension n . Recall that a real structure on E is an anti-linear involution a : E E . Recall that any two real structures a , b : E E on E are equivalent, i.e. there exists l GL ( E ) such that a = l b l 1 . Indeed, putting

E ± a ker ( a id E ) , E ± b ker ( b id E ) ,

we have real direct sum decompositions

E = E + a E a , E = E + b E b

with E a = i E + a , E b = i E + b . Choose an R -linear isomorphism h : E + b E + a and note that l : E E defined by

l ( x + i y ) h ( x ) + i h ( y ) for any x , y E + b

is a C -linear isomorphism satisfying a l = l b . In particular, in the special case, when E = C n and b is the standard conjugation c : C n C n , we obtain

Remark 3.3

Let a : C n C n be an anti-linear involution on C n . Then there exists a C -linear automorphism l : C n C n such that a = l c l 1 .

In other words, any anti-linear Real structure on C n is GL ( n , C ) -conjugate to the standard conjugation.

The classification of even Real structures on primary Hopf surfaces will follow from the following proposition:

Proposition 3.4

Let f I V I I I I I a I I b I I c be with real coefficients and let ϕ be a Real structure on W such that ϕ f = f ϕ . Then there exists ψ Aut h ( W ) f such that ϕ = ψ c ψ 1 with c is the standard conjugation on W.

Proof

  1. f I V . In this case, using Proposition 2.7, we see that ϕ has the form

    ϕ z w = a b c d z ¯ w ¯

    with a b c d GL ( 2 , C ) . The obvious extension ϕ ˜ : C 2 C 2 of ϕ to C 2 is an anti-linear Real structure on C 2 , so Remark 3.3 applies and gives a C -linear isomorphism l : C 2 C 2 such that ϕ ˜ = l c l 1 . Denoting by ψ : W W the automorphism induced by l (which obviously commutes with f ), we obtain ϕ = ψ c ψ 1 as claimed.

  2. f I I I . In this case, Proposition 2.7 shows that ϕ is given by

    ϕ z w = a z ¯ + b w ¯ r d w ¯ ,

    where a , d C * , b C . We have

    ϕ 2 z w = a 2 z + ( a b ¯ + b d ¯ r ) w r d 2 w .

    The condition ϕ 2 = id becomes

    (5) a 2 = 1 a b ¯ + b d ¯ r = 0 d 2 = 1 .

    Let now ψ Aut h ( W ) f . By Wehler’s Theorem 2.1, we have

    ψ z w = A z + B w r D w

    with A , D C * , B C . The condition

    ϕ = ψ c ψ 1

    is equivalent to the system

    (6) a = A A ¯ 1 b = B D ¯ r A A ¯ 1 B ¯ D ¯ r d = D D ¯ 1 .

    Since a = d = 1 we can write a = e i θ , d = e i δ with θ , δ R . Put A e i θ 2 , D e i δ 2 . With this choice, we have A A ¯ 1 = A 2 = a , D D ¯ 1 = D 2 = d , so the first and the third equations in (6) are satisfied. We are seeking B C such that the second equation is also satisfied.

    Consider the R -linear map λ : C C defined by

    λ ( z ) = u z v z ¯ ,

    where u D ¯ r , v A A ¯ 1 D ¯ r = A 2 D ¯ r . This map is not surjective. By using u = v = 1 , it follows easily that its image is the real line

    (7) im ( λ ) = { ζ C u 1 ζ + v ζ ¯ = 0 } C .

    Now note that b im ( λ ) , because

    u 1 b + v b ¯ = D ¯ r b + A 2 D ¯ r b ¯ = D ¯ r ( D ¯ 2 r b + A 2 b ¯ ) = D ¯ r ( d ¯ r b + a b ¯ ) = 0

    by the second equation in (5). Therefore, there exists B C such that

    b = λ ( B ) = D ¯ r B A A ¯ 1 D ¯ r B ¯ ,

    which proves that, with this choice, the second equation in (6) is satisfied, too.

  3. f I I a . In this case, Proposition 2.7 shows that ϕ has the form

    ϕ z w = a r z ¯ + b w ¯ r a w ¯

    with a C * and b C . The condition ϕ 2 = id W is equivalent to the system

    (8) a 2 = 1 a r b ¯ + b a ¯ r = 0 .

    By Wehler’s theorem, an automorphism ψ Aut h ( W ) f has the form

    ψ z w = A r z + B w r A w

    with A C * , B C . The condition

    ϕ = ψ c ψ 1

    is equivalent to the system

    (9) a = A A ¯ 1 b = B A ¯ r A r B ¯ A ¯ 2 r .

    Since a 2 = 1 , we can write a = e i θ with θ R . Putting A = e i θ 2 , we have A A ¯ 1 = A 2 = a . As in the previous case, consider the R -linear map λ : C C

    λ ( z ) = u z v z ¯ ,

    where this time we choose u A r and v A 3 r . We have again u = v = 1 , so the image of λ is again given by (7). Note that b im ( λ ) because

    u 1 b + v b ¯ = A r b + A 3 r b ¯ = A r ( A 2 r b ¯ + A 2 r b ) = A r ( a r b ¯ + b a ¯ r ) = 0

    by the second equation in (8). Therefore, there exists B C such that

    b = λ ( B ) = A r B A 3 r B ¯ = B A ¯ r A r B ¯ A ¯ 2 r ,

    which shows that ( A , B ) C * × C is a solution of the system (9).

  4. f I I b . In this case, Proposition 2.7 shows that ϕ has the form

    ϕ z w = a z ¯ + b w ¯ a w ¯ ,

    where a C * , b C . The condition ϕ 2 = id W is equivalent to

    (10) a 2 = 1 a b ¯ + b a ¯ = 0 .

    By Wehler’s theorem, an automorphism ψ Aut h ( W ) f has the form

    ψ z w = A z + B w A w ,

    where A C * , B C . We use the same arguments as in the case I I a but taking r = 1 in all formulae used in this case.

  5. f I I c . In this case, Proposition 2.7 shows that ϕ has the form

    ϕ z w = a z ¯ b w ¯ ,

    where a , b C * . The condition ϕ 2 = id W is equivalent to

    (11) a 2 = 1 b 2 = 1 .

    By Wehler’s theorem, in this case, an automorphism ψ Aut h ( W ) f has the form

    ψ z w = A z B w ,

    where A , B C * , and the condition

    ϕ = ψ c ψ 1

    is equivalent to the system

    (12) a = A A ¯ 1 b = B B ¯ 1 .

    It suffices to put A e i θ 2 , B e i β 2 , where a = e i θ , b = e i β .□

By using the notation c introduced in Example 3.1, we can state the following analogue of Proposition 3.4 for the the subclass I I c :

Proposition 3.5

Let f I I c be given by f z w = α z α ¯ w , where 0 < α < 1 and α R . Let ϕ be a Real structure on W such that ϕ f = f ϕ . There exists ψ Aut h ( W ) f such that ϕ = ψ c ψ 1 .

Proof

By using Proposition 2.7, we see that ϕ has the form ϕ z w = a w ¯ d z ¯ , and the condition ϕ 2 = id W becomes a ¯ = d 1 . It suffices to note that

z w a z w

defines an element ψ Aut h ( W ) f and that ϕ = ψ c ψ 1 .□

By using Propositions 3.4 and 3.5, we obtain the following classification theorem for even Real structures on primary Hopf surfaces:

Theorem 3.6

Let f I V I I I I I a I I b I I c be either with real coefficients or an element of the subclass I I c . Let σ : H f H f be an even Real structure on H f . There exists a holomorphic automorphism g Aut h ( H f ) such that σ = g s f g 1 , where s f is the standard Real structure on H f .

In other words, any even Real structure on H f is equivalent to its standard Real structure s f .

Proof

Since σ is even, there exists a lift σ ˆ : W W which is an involution, and hence, a Real structure on W . By Propositions 3.4 and 3.5, there exists ψ Aut h ( W ) f such that σ ˆ = ψ c ψ 1 , respectively, σ ˆ = ψ c ψ 1 . Denoting by g Aut h ( H f ) the automorphism induced by ψ , we obtain σ = g s f g 1 .□

Note that Proposition 3.1 and Definition 3.2 generalize in a natural way to primary Hopf n -folds for any n 2 . Moreover, for a primary Hopf n -fold H f defined by a holomorphic contraction f given by a polynomial formula with real coefficients, we can define the standard Real structure s f of H f as in Example 3.1.

By using Remark 3.3, we obtain in the same way the classification of even Real structures on a primary Hopf n -fold

H f α C n \ { 0 } f α ,

where 0 < α < 1 and f α ( z ) = α z .

Remark 3.7

The primary Hopf n -fold H f α admits an even Real structure if and only if α R . If this is the case, any even Real structure on H f α is equivalent to its standard Real structure s f α .

3.3 The classification of odd Real structures on primary Hopf surfaces

The classification of odd Real structures is more difficult. We consider first the case when the diagonal coefficients of f (denoted by α , δ , δ r in Theorem 2.1) are positive. We will start with the following remark, which can be proved easily by direct computations:

Remark 3.8

Let f I V I I I I I a I I b I I c be with real coefficients and positive diagonal coefficients, and let k N * . The automorphism f 1 k Aut h ( W ) defined in the following table has the properties:

  1. f 1 k is a polynomial holomorphic contraction with real coefficients.

  2. f 1 k is a root of order k of f .

  3. Aut h ( W ) f 1 k = Aut h ( W ) f , Ah ( W ) f 1 k = Ah ( W ) f .

The class of f f z w f 1 k z w
I V α z α w α 1 k z α 1 k w
I I I δ r z δ w δ r k z δ 1 k w
I I a δ r z + w r δ w δ r k z + 1 k δ r ( 1 k k ) w r δ 1 k w
I I b α z + w α w α 1 k z + 1 k α 1 k k w α 1 k w
I I c α z δ w α 1 k z δ 1 k w

Proposition 3.9

Let f I V I I I I I a I I b I I c be with real coefficients and positive diagonal coefficients, and let f 1 2 be the square root of f given by Remark 3.8. Let c : W W be the standard conjugation. Then:

  1. The composition c f 1 2 belongs to Ah ( W ) f and satisfies ( c f 1 2 ) 2 = f .

  2. For any ϕ Ah ( W ) f with ϕ 2 = f , there exists ψ Aut h ( W ) f such that

    ϕ = ψ ( c f 1 2 ) ψ 1 .

In other words, under the assumption of the proposition, any odd Real structure on H f is equivalent to the odd real structure σ f induced by c f 1 2 .

Proof

  1. This follows taking into account that, since f 1 2 has real coefficients, it commutes with c .

  2. Put ϕ = f 1 2 ϕ , and note that ϕ belongs to Ah ( W ) f and (since ϕ commutes with f 1 2 by Remark 3.8 (3)) satisfies ϕ 2 = id W . Therefore, ϕ is a Real structure on W which commutes with f . By Proposition 3.4, there exists ψ Aut h ( W ) f such that ϕ = ψ c ψ 1 . Therefore,

    f 1 2 ϕ = ψ c ψ 1 .

    By Remark 3.8 (3) again ψ commutes with f 1 2 , so we obtain

    ϕ = ψ ( c f 1 2 ) ψ 1 ,

    as claimed.□

The proposition below shows that if H f admits an odd Real structure and

f I I I I I a I I b I I c

has real coefficients, then its diagonal coefficients are always positive, so Proposition 3.9 classifies odd Real structure on all primary Hopf surfaces except those of class I I c and those of class I V with negative parameter α .

Proposition 3.10

Let f I I I I I a I I b I I c be with real coefficients. The following conditions are equivalent:

  1. There exists ϕ Ah ( W ) f such that ϕ 2 = f .

  2. The diagonal coefficients of f are positive.

Proof

( 1 ) ( 2 ) : Suppose f I I I , so f z w = δ r z δ w with 0 < δ < 1 . An element ϕ Ah ( W ) f has the form

ϕ z w = a z ¯ + b w ¯ r d w ¯ .

The condition ϕ 2 = f becomes

a 2 = δ r a b ¯ + b d ¯ r = 0 d 2 = δ ,

which obviously implies δ > 0 . The other cases are treated in a similar way.

( 2 ) ( 1 ) : This follows from Proposition 3.9 (1).□

The analogue of Proposition 3.9 for f I I c follows easily by direct computation:

Proposition 3.11

Let f I I c be given by f z w = α z α ¯ w , where 0 < α < 1 and α R . Let α 1 2 be a square root of α , α ¯ 1 2 be its conjugate, and let f 1 2 be the square root of f defined by f 1 2 z w = α 1 2 z α ¯ 1 2 w . Then

  1. c commutes with f 1 2 .

  2. c f 1 2 Ah ( W ) f satisfies ( c f 1 2 ) 2 = f .

  3. For any ϕ Ah ( W ) f with ϕ 2 = f , there exists ψ Aut h ( W ) f such that ϕ = ψ ( c f 1 2 ) ψ 1 .

In other words, under the assumption of the proposition, for f I I c , any odd Real structure on H f is equivalent to the odd real structure σ f induced by c f 1 2 .

For the classification of odd Real structures on class IV primary Hopf surfaces with negative parameter α , we will need a simple remark concerning the classification of quaternionic structures on a finite dimensional real vector space.

Let F be a real vector space of dimension 4 k . Recall that a left H -vector space structure on F is equivalent to the data of a pair ( I , J ) End ( F ) × End ( F ) such that I 2 = J 2 = id F and I J = J I . In the presence of such a pair ( I , J ) , we put K I J , and we define a left quaternonic vector space structure on F (extending its real space structure) by mapping the quaternonic units i , j , k H to I , J , K , respectively.

Since two left H -vector spaces of the same dimension are isomorphic, it follows that any two left H -vector space structures ( I , J ) , ( I , J ) on F are always equivalent, i.e. there exists an automorphism l GL ( F ) such that l I = I l , l J = J l . In the special case I = I , we obtain l I = I l , i.e. l is linear with respect to the complex structure defined by I , and J = l J l 1 . This shows the following:

Remark 3.12

Let E be a complex vector space of dimension 2 k and let J , J be anti-linear isomorphisms such that J 2 = J 2 = id E . Then there exists an automorphism l GL ( E ) such that J = l J l 1 .

Consider the anti-linear map J : C 2 k C 2 k defined by

J z w = w ¯ z ¯ = 0 id C k id C k 0 z ¯ w ¯

for pairs ( z , w ) C k × C k = C 2 k . Note that J 2 = id C 2 k . By Remark 3.12, we obtain the following analogue of Remark 3.3:

Remark 3.13

Let a : C 2 k C 2 k be an anti-linear isomorphism satisfying a 2 = id C 2 . Then there exists l GL ( 2 k , C ) such that a = l J l 1 , i.e. a is GL ( 2 k , C ) -conjugate to J .

With this preparation we can prove:

Proposition 3.14

Let f : W W I V be given by f z w = α z α w , where α ( 1 , 0 ) . Let α 1 2 be a square root of α , and let f 1 2 be the square root of f defined by f 1 2 z w = α 1 2 z α 1 2 w . Then

  1. J anti-commutes with f 1 2 .

  2. J f 1 2 Ah ( W ) f and satisfies J f 1 2 2 = f .

  3. For any ϕ Ah ( W ) f with ϕ 2 = f , there exists ψ Aut h ( W ) such that ϕ = ψ J f 1 2 ψ 1 .

Therefore, under the assumptions of the proposition, any odd Real structure σ on H f is equivalent to s f , where s f is the odd Real structure induced by J f 1 2 .

Proof

  1. It suffices to note that f 1 2 = α 1 2 id W with α 1 2 pure imaginary and to recall that J is anti-linear.

  2. Follows from (1) taking into account that J 2 = id C 2 .

  3. Let ϕ Ah ( W ) f with ϕ 2 = f . The extension ϕ ˜ of ϕ to C 2 gives an anti-linear isomorphism ϕ ˜ : C 2 C 2 . Since ϕ ˜ anti-commutes with f ± 1 2 , the condition ϕ 2 = f is equivalent to ( ϕ ˜ f 1 2 ) 2 = id C 2 . By Remark 3.13, there exists l GL ( 2 , C ) such that

    ϕ ˜ f 1 2 = l J l 1 .

    Denoting by ψ the automorphism of W induced by l , which obviously commutes with f 1 2 , we obtain ϕ = ψ J f 1 2 ψ 1 , as claimed.□

By using Propositions 3.9 and 3.14, we obtain the following classification theorem for odd Real structures on primary Hopf surfaces:

Theorem 3.15

  1. Let f I I I I I a I I b I I c .

    1. The following conditions are equivalent:

      1. H f admits an odd Real structure.

      2. f either has real coefficients and positive diagonal coefficients, or belongs to I I c .

    2. If one of these equivalent conditions is satisfied, any odd Real structure on H f is equivalent to the Real structure σ f defined earlier.

  2. Let f I V be given by f ( z , w ) = ( α z , α w ) where 0 < α < 1 .

    1. The following conditions are equivalent:

      1. H f admits an odd Real structure.

      2. α R .

    2. If α ( 0 , 1 ) , any odd Real structure on H f is equivalent to σ f . If α ( 1 , 0 ) , any odd Real structure on H f is equivalent to s f .

Consider again the primary Hopf n -fold H f α C n \ { 0 } f α . By using Remark 3.13 and defining the odd Real structures σ f α (for α ( 0 , 1 ) ), s f α (for α ( 1 , 0 ) and n even) as earlier, we obtain in a similar way the classification of odd Real structures on H f α :

Remark 3.16

Let H f α be the primary Hopf n -fold associated with the holomorphic contraction f α , where 0 < α < 1 .

  1. Suppose n is odd. H f α admits an odd Real structure if and only if α ( 0 , 1 ) . If this is the case, any odd Real structure on H f α is equivalent to σ f α .

  2. Suppose n is even. H f α admits an odd Real structure if and only if α R . If α ( 0 , 1 ) , any odd Real structure on H f α is equivalent to σ f α . If α ( 1 , 0 ) , any odd Real structure on H f α is equivalent to s f α .

4 The automorphism group and the Real Picard group

4.1 The automorphism group

For the automorphism group of an even Real primary Hopf surface ( H f , s f ) , we have the following result which follows easily from Theorem 2.1:

Theorem 4.1

Let f I V I I I I I a I I b I I c .

  1. Suppose f has real coefficients. The group Aut h ( W ) f , c of holomorphic automorphisms of W commuting with f and the standard conjugation c is given by the following table.

The class of f Aut h ( W ) f , c
I V GL ( 2 , R )
I I I z w a z + b w r d w a R * , d R * , b R
I I a z w a r z + b w r a w a R * , b R
I I b z w a z + b w a w a R * , b R
I I c z w a z d w a R * , d R *

  1. Suppose f I I c . Then

    Aut h ( W ) f , c = z w a z a ¯ w a C * .

  2. In each case, the cyclic group f is a central subgroup of Aut h ( W ) f , c , respectively Aut h ( W ) f , c , and the automorphism group Aut h ( H f , s f ) is identified with the quotient Aut h ( W ) f , c f , respectively Aut h ( W ) f , c f .

Proof

The claims (1) and (2) follow directly from Theorem 2.1. For (3), it suffices to prove that a holomorphic automorphism φ Aut h ( H f ) induced by φ ˆ Aut h ( W ) f commutes with s f if and only if φ ˆ commutes with c , respectively c . In other words, we have to prove that c φ ˆ = φ ˆ c f k , respectively c φ ˆ = φ ˆ c f k (with k Z ), then k = 0 . This follows by elementary computations.□

By using Theorem 4.1, we can describe the automorphism groups of even Real primary Hopf surfaces in terms of (semi-direct products of) classical groups. For instance, for f I V , we obtain Aut ( H f , s f ) = GL ( 2 , R ) α I 2 .

For f I I I , the group Aut h ( W ) f , c coincides with the group

{ g a , d , b ( a , d , b ) R * × R * × R } , where g a , d , b z w = a z + a b w r d w .

This group can be identified with the semi-direct product ( R * × R * ) ρ r R associated with the morphism ρ r : ( R * × R * ) GL ( 1 , R ) given by

ρ r ( a , d ) ( b ) = a d r b .

Since ρ r ( δ r , δ ) = id R , it follows that ρ r descends to a morphism

ρ ˆ r : R * × R * ( δ r , δ ) GL ( 1 , R ) .

With this remark, we obtain:

Corollary 4.2

Let f I I I with real coefficients δ r , δ . The automorphism group Aut ( H f , s f ) can be identified with the semi-direct product

[ R * × R * ( δ r , δ ) ] ρ ˆ r R .

Similar descriptions are obtained for f I I a and f I I b . For f I I c with real coefficients, and we have obviously Aut ( H f , s f ) = ( R * × R * ) ( α , δ ) , and for f I I c we obtain Aut ( H f , s f ) = C * α , which is a one-dimensional complex torus.

The automorphism group of the odd Real Hopf surfaces is given by the following:

Theorem 4.3

Let f I V I I I I I a I I b I I c .

  1. Suppose f has real coefficients and positive diagonal coefficients. Then

    Aut ( H f , σ f ) = Aut ( H f , s f ) .

  2. Suppose f I I c . Then again

    Aut ( H f , σ f ) = Aut ( H f , s f ) .

  3. Suppose that f I V with negative diagonal coefficient α . Then

    Aut ( H f , s f ) = a b ¯ b a ¯ ( a , b ) C 2 \ { 0 } α I 2 .

Proof

Use similar arguments, based on elementary computations, as in the proof of Theorem 4.1.□

The group a b ¯ b a ¯ ( a , b ) C 2 \ { 0 } can be identified with the subgroup R + * SU ( 2 ) of GL ( 2 , C ) . This subgroup is isomorphic with H * via the map:

z + j w z w ¯ w z ¯ .

Therefore, in case (3), we have

(13) Aut ( H f , s f ) R + * SU ( 2 ) α I 2 .

The right-hand quotient in (13) can be written as the quotient of R + * SU ( 2 ) α 2 I 2 by the order 2 group α I 2 α 2 I 2 . Via the Lie group isomorphism

Φ : R + * SU ( 2 ) R + * × SU ( 2 ) , Φ ( A ) = det ( A ) 1 2 , det ( A ) 1 2 A ,

the matrices α I 2 , α 2 I 2 are mapped to ( α , I 2 ) , ( α 2 , I 2 ) , respectively. Therefore, Φ induces an isomorphism

ϕ : R + * SU ( 2 ) α 2 I 2 R + * × SU ( 2 ) ( α 2 , I 2 ) = R + * α 2 × SU ( 2 ) ,

and the image of α I 2 in the right-hand group is ( [ α ] , I 2 ) . By identifying R + * α 2 with S 1 via the isomorphisms [ ρ ] e π i ln ( ρ ) ln α , and noting that the image of [ α ] in S 1 via this identification is 1 , we obtain an isomorphism

R + * SU ( 2 ) α I 2 S 1 × SU ( 2 ) ( 1 , I 2 ) .

Therefore:

Corollary 4.4

Let α ( 1 , 0 ) and f α I 2 . The automorphism group of the odd Real Hopf surface ( H f , s f ) is naturally isomorphic to the group

Spin c ( 3 ) = S 1 × Z 2 Spin ( 3 ) = S 1 × Z 2 SU ( 2 ) .

4.2 The Real Picard group of a Real primary Hopf surface

For a class VII surface X , the canonical Lie group morphism

(14) Hom ( π 1 ( X , x 0 ) , C * ) = Hom ( H 1 ( X , Z ) , C * ) Π c ( X )

is injective and its image is the subgroup Π c T ( X ) of isomorphism classes of holomorphic line bundles with torsion Chern class (see [9, Remark 3.2.3]). For a primary Hopf surface X , we have H 2 ( X , Z ) = { 0 } , so Π c 0 ( X ) = Π c T ( X ) = Π c ( X ) , so (14) is an isomorphism. Since π 1 ( X , x 0 ) Z , we obtain an isomorphism

λ : C * = Hom ( π 1 ( X , x 0 ) , C * ) Π c ( X ) ,

which can be obtained explicitly as follows. Suppose X = H f for a holomorphic contraction f Aut h ( W ) (Section 2.1). For ζ C * , put

L ζ W × C f ζ ,

where f ζ : W × C W × C is the fibrewise linear automorphism

f ζ ( x , z ) ( f ( x ) , ζ z ) .

Endowed with the obvious surjective submersion L ζ H f given by

[ ( x , z ) ] f ζ [ x ] f ,

L ζ is naturally a holomorphic line bundle on H f . Recall (see for instance [9, Section 2.2]) that

Remark 4.5

The map λ : C * Π c ( H f ) defined by λ ( ζ ) = [ L ζ ] is a Lie group isomorphism.

The following proposition shows that, for any f I V I I I I I a I I b I I c and any Real structure s on H f , the anti-holomorphic involutive isomorphism

s ¯ * : Π c ( H f ) Π c ( H f )

induced by s (see Section 1) is given by the same formula.

Proposition 4.6

Let H be a primary Hopf surface, and let s Ah ( H ) (not necessarily involutive). Then for any ζ C * , we have

s ¯ * ( [ L ζ ] ) [ L ζ ¯ ] .

Proof

By Proposition 2.4, we know that s is induced by an anti-holomorphic automorphism s ˆ Ah ( W ) f . The commutative diagram

shows that the map ( s ˆ , ) descends to a well-defined, fibrewise anti-linear, anti-holomorphic, s -lifting isomorphism L ζ ¯ L ζ . The same map can be regarded as a fibrewise linear, holomorphic, s -lifting isomorphism L ζ ¯ L ¯ ζ , where this time s has been regarded as a holomorphic map H f H ¯ f . Therefore, L ζ ¯ s * ( L ¯ ζ ) = s ¯ * ( L ζ ) .□

Proposition 4.6 shows in particular that, for any Real primary Hopf surface ( H , s ) , the associated Real structure s ¯ * on Π c ( H ) is always given, via the isomorphism λ , by the standard conjugation C * C * .

If follows that if a line bundle L ζ on a Real Hopf surface ( H , s ) admits an anti-holomorphic Real structure ϕ , then ζ R * . Let ζ R * . The proof of Proposition 4.6 shows that the map ( s ˆ , ¯ ) descends to a fibrewise anti-linear, anti-holomorphic, s -lifting isomorphism ϕ 0 : L ζ L ζ . Any fibrewise anti-linear, anti-holomorphic, s -lifting isomorphism ϕ : L ζ L ζ has the form ϕ = ν ϕ 0 for a constant ν C * . Indeed, the composition ϕ ϕ 0 1 is a holomorphic id H -covering automorphism of L , so ϕ ϕ 0 1 = ν id L ζ with ν C * .

We have

(15) ϕ 2 = ( ν ϕ 0 ) ( ν ϕ 0 ) = ν 2 ϕ 0 2 .

On the other hand, ϕ 0 2 is induced by the map W × C W × C given by

(16) W × C ( x , z ) ( s ˆ 2 ( x ) , z ) .

  • If s is even, we can choose s ˆ such that s ˆ 2 = id W ; in this case, ϕ 0 is already an anti-holomorphic Real structure on L ζ and formula (15) shows that the set of all anti-holomorphic Real structures on L ζ is S 1 ϕ 0 .

  • If s is odd, we can choose s ˆ such that s ˆ 2 = f ; in this case, formula (16) shows that

    ϕ 0 2 ( [ x , z ] ) = [ f ( x ) , z ] = [ x , ζ 1 z ] ,

    so ϕ 0 2 = ζ 1 id L ζ . Taking into account (15), it follows that ϕ = ν ϕ 0 is involutive if and only if ζ = ν 2 . Therefore, in this case, L ζ admits anti-holomorphic Real structures if and only if ζ > 0 , and if this is the case, the set of anti-holomorphic Real structures on L ζ is S 1 ζ ϕ 0 .

We have proved:

Proposition 4.7

Let ( H f , s ) be a Real primary Hopf surface, let s ˆ Ah ( W ) f be a lift of s with s ˆ 2 { id W , f } , and let ζ R * .

  1. If s ˆ 2 = id W (i.e. if s is even), then the set of anti-holomorphic Real structures on L ζ is S 1 ϕ 0 .

  2. If s ˆ 2 = f (i.e. if s is odd), then L ζ admits anti-holomorphic Real structures if and only if ζ > 0 , and, if this the case, the set of anti-holomorphic Real structures on L ζ is S 1 ζ ϕ 0 .

In all cases, the group C * id L ζ of holomorphic automorphisms acts on the set of anti-holomorphic Real structures on L ζ by conjugation, and the explicit formula for this action is

( z id ) ϕ ( z id ) 1 = z z ¯ 1 ϕ = ( z z 1 ) 2 ϕ .

Since the map C * S 1 given by z ( z z 1 ) 2 is obviously surjective, it follows that any ψ S 1 ϕ is isomorphic (equivalent) to ϕ . This shows that, under the assumptions of Proposition 4.7, all anti-holomorphic Real structures on L ζ are isomorphic to either ϕ 0 (if s ˆ 2 = id W ), or to ζ ϕ 0 (if s ˆ 2 = f and ζ > 0 ).

In conclusion, we obtain:

Proposition 4.8

Let ( H , s ) be a Real primary Hopf surface.

  1. Via the isomorphism C * λ Π c ( H ) , the Real structure Π c ( H ) s * ¯ Π c ( H ) induced by s coincides with the standard conjugation.

  2. The map [ L ζ , ϕ ] ζ defines

    1. An isomorphism Π c R ( H ) R * if s is an even Real structure.

    2. An isomorphism Π c R ( H ) R > 0 if s is an odd Real structure.

The second statement shows that, for odd Real primary Hopf surfaces, the obvious group morphism Π c R ( H ) Π c ( H ) ( R ) is not surjective; the classes [ L ζ ] with ζ R < 0 do not correspond to Real holomorphic line bundles in the sense of Definition 1.2, although they are fixed-points of s ¯ * .

5 The differential-topological classification

5.1 The differential-topological classification of even Real Hopf surfaces

On the product S 1 × S 3 , we consider the following involutions

(17) τ ( ζ , ( u , v ) ) ( ζ , ( u ¯ , v ¯ ) ) τ ( ζ , ( u , v ) ) ( ζ , ( u ¯ , ζ v ¯ ) ) .

Our goal is the following classification result:

Theorem 5.1

Any even Real primary Hopf surface ( H , σ ) is equivariantly diffeomorphic, to either ( S 1 × S 3 , τ ) or ( S 1 × S 3 , τ ) .

We will also need the following involutions on S 1 × S 3 :

(18) j ( ζ , ( u , v ) ) ( ζ , ( u , v ) ) , j ( ζ , ( u , v ) ) ( ζ , ( u , v ) ) .

The order 2 groups j , j act freely and properly discontinuously on S 1 × S 3 , so we obtain double covers

q : S 1 × S 3 Q S 1 × S 3 j , q : S 1 × S 3 Q S 1 × S 3 j .

Note that τ commutes with j and j , so we obtain induced involutions

θ : Q Q , θ : Q Q

induced by τ . We will need the following notation:

Definition 5.2

For ζ S 1 we denote by R ζ SO ( 2 ) GL ( 2 , C ) the 2 × 2 matrix which corresponds to ζ via the standard isomorphism S 1 SO ( 2 ) .

Lemma 5.3

Consider the maps a : S 1 × S 3 S 1 × S 3 , a : S 1 × S 3 S 1 × S 3 defined by

a ( ζ , ( u , v ) ) ( ζ 2 , ( u , ζ v ) ) , a ( ζ , ( u , v ) ) ( ζ 2 , R ζ ( u , v ) ) .

Then

  1. a is j -invariant and induces a diffeomorphism of Z 2 -spaces

    a ˆ : ( Q , θ ) ( S 1 × S 3 , τ ) .

  2. a is j -invariant and induces a diffeomorphism of Z 2 -spaces

    a ˆ : ( Q , θ ) ( S 1 × S 3 , τ ) .

Proof

It is easy to see that a ( a ) is j -invariant (respectively j -invariant) and that the induced maps a ˆ : Q S 1 × S 3 , a ˆ : Q S 1 × S 3 are bijective and verify τ a ˆ = a ˆ θ , τ a ˆ = a ˆ θ . On the other hand, a ˆ , a ˆ are local diffeomorphisms because a , a have this property. Therefore, a ˆ , a ˆ are diffeomorphisms.□

The idea of the proof of Theorem 5.1 is the following: using our classification Theorem 3.6, we may suppose that ( H , σ ) = ( H f , s f ) , where f I V I I I I I a I I b I I c is either with real coefficients or f I I c , and s f is the canonical Real structure on H f . We will show (Propositions 5.4 and 5.10) that ( H f , s f ) is equivariantly diffeomorphic to ( S 1 × S 3 , τ ) , to ( Q , θ ) or to ( Q , θ ) . Theorem 5.1 will then follow by Lemma 5.3.

Proposition 5.4

Let f I V I I I I I a I I b I I c be with real coefficients.

  1. If the diagonal coefficients of f are positive, then ( H f , s f ) is equivariantly diffeomorphic to ( S 1 × S 3 , τ ) .

  2. If a single diagonal coefficients of f is negative, then ( H f , s f ) is equivariantly diffeomorphic to ( Q , θ ) .

  3. If both diagonal coefficients of f are negative, then ( H f , s f ) is equivariantly diffeomorphic to ( Q , θ ) .

The proof of Proposition 5.4 requires a preparation. Let f I I a ,

f z w = δ r z + w r δ w .

Suppose δ R . The second power f 2 of f is given by

f 2 z w = δ 2 r z + 2 δ r w r δ 2 w

has always positive diagonal coefficients, but, unfortunately f 2 I I a . Similar remark for f I I b . Therefore, we will need the slightly larger classes I I ˜ a I I a , I I ˜ b I I b defined by

I I ˜ a f : W W f z w = δ r z + c w r δ w , 0 < δ < 1 , c C I I ˜ b f : W W f z w = α z + c w α w , 0 < δ < 1 , c C .

We begin with the following remark, which shows that any contraction f I V I I I I I a I I b I I c with real coefficients and positive diagonal coefficients can be identified with the term f 1 of an explicit smooth 1-parameter group ( f t ) t R of holomorphic automorphisms of W . More precisely:

Remark 5.5

Let f I V I I I I I ˜ a I I ˜ b I I c with real coefficients and positive diagonal coefficients. For t R , we define f t Aut h ( W ) by the formula specified in the third column of the following table:

The class of f f z w f t z w
I V α z α w α t z α t w
I I I δ r z δ w δ r t z δ t w
I I ˜ a δ r z + c w r δ w δ r t z + c t δ r ( t 1 ) w r δ t w
I I ˜ b α z + c w α w α t z + c t α t 1 w α t w
I I c α z δ w α t z δ t w

Then

  1. The family ( f t ) t R a one-parameter group of automorphisms of W , i.e. the map R t f t Aut h ( W ) is a group morphism.

  2. f 1 = f .

For a map η : W R and Z = ( z , w ) W , we define η Z : R R by

(19) η Z ( t ) η ( f t ( Z ) ) .

Since ( f t ) t R is a one-parameter group of diffeomorphisms, we have the identity

(20) η Z ( t ) = η f t ( Z ) ( 0 ) .

Remark 5.6

Let B ( 0 , ) and q N * . Let η B q : W ( 0 , ) be the map defined by

η B q ( z , w ) = z 2 + B w 2 q .

  1. η B q is a submersion, in particular, the fibre Σ B q ( η B q ) 1 ( 1 ) is a smooth hypersurface of W .

  2. The restriction n B q : Σ B q S 3 of the normalization map N : W S 3 , N ( Z ) 1 Z Z to Σ B q is a diffeomorphism which commmutes with the involutions ( z , w ) ( z , w ) , ( z , w ) ( z , w ) , ( z , w ) ( z ¯ , w ¯ ) .

Proof

The first claim follows by elementary computation. For the second, note first that for any Z W , the half-line R + * Z intersects Σ B q in a unique point, which will be denoted N B q ( Z ) . Using the implicit function theorem, it follows easily that the obtained map N B q : W Σ B q is smooth, so the restriction n B q N B q S 3 : S 3 Σ B q will also be smooth. It suffices to note that

n B q n B q = id Σ B q , n B q n B q = id S 3 .

Lemma 5.7

Let f I V I I I I I ˜ a I I ˜ b I I c with real coefficients and positive diagonal coefficients. Let η : W ( 0 , + ) and C < 0 be given by the following table:

The class of f f z w η C
I V α z α w η 1 1 2 ln ( α )
I I I δ r z δ w η 1 1 2 ln ( δ )
I I ˜ a δ r z + c w r δ w η B r   with  B c 2 1 r 2 δ 2 r ln ( δ ) 2 r ln ( δ )
I I ˜ b α z + c w α w η B 1   with  B c 2 1 α 2 ln ( α ) 2 ln ( α )
I I c α z δ w η 1 1 2 max ( ln ( α ) , ln ( δ ) )

  1. In each case and for any Z W , the map η Z satisfies the differential inequality

    (21) η Z ( t ) C η Z ( t ) .

    In particular, η Z is strictly decreasing and

    (22) lim t η Z ( t ) = 0 , lim t η Z ( t ) = + .

  2. Put Σ η 1 ( 1 ) . The map F : R × Σ W defined by F ( t , Z ) = f t ( Z ) is a diffeomorphism.

  3. Endowing W with the conjugation c and R × Σ with the involution id × c Σ (where c Σ denotes the involution induced by c on Σ ), F is equivariant.

Proof

(1) The proof of (21) is based on formula (20). For f I V I I I I I c , we have η ( z , w ) = z 2 + w 2 and the computation of η Z ( 0 ) is very easy. For f I I ˜ a , one obtains for any ε > 0 :

(23) η Z ( 0 ) = 2 r ln ( δ ) z 2 + δ r r ln ( δ ) c ( z w ¯ r ) + B w 2 r 2 r ln ( δ ) 1 + δ r ε 2 r ln ( δ ) z 2 + B + δ r c 2 2 r ln ( δ ) ε w r 2 .

We first choose ε = δ r r ln ( δ ) , and we note that, for B c 2 r 2 δ 2 r ln ( δ ) 2 , we have

B + δ r c 2 2 r ln ( δ ) ε 1 2 B .

The case f I I b is similar. The formulae (22) follow from (21) by integrating the inequality η η C .

(2) For the injectivity of F , let ( t , Z ) , ( t , Z ) R × Σ such that F ( t , Z ) = F ( t , Z ) . This implies f t t ( Z ) = Z . By applying η on both sides, we obtain η Z ( t t ) = 1 = η Z ( 0 ) . Since η Z is strictly decreasing, this implies t = t , so we also have Z = Z .

For the surjectivity, its suffices to note that for any Z W , there exists t R such that f t ( Z ) Σ , which is equivalent to η ( f t ( Z ) ) = 1 , i.e. η Z ( t ) = 1 . But (22) shows that η Z ( R ) = ( 0 , + ) .

F is obviously differentiable. To see that it is a diffeomorphism it suffices to prove that F is a local diffeomorphism, i.e. that the differential d ( t , Z ) F is invertible for any ( t , Z ) R × Σ .

For t R denote by τ t : R R the translation by t , i.e. τ t ( s ) = s + t . Note that

F ( τ t , id Σ ) = f t F .

This implies

F * ( s + t , Z ) = F * ( τ t , id Σ ) ( s , Z ) ( τ t , id Σ ) * ( s , Z ) = f * F ( s , Z ) t F * ( s , Z ) .

For s = 0 , we obtain

F * ( t , Z ) = f * Z t F * ( 0 , Z ) .

Since f t is a diffeomorphism, f * Z t is a linear isomorphism, so it suffices to prove that F * ( 0 , Z ) is a linear isomorphism. Taking into account the dimensions, it suffices to prove that ker ( F * ( 0 , Z ) ) = 0 . But

(24) F * ( 0 , Z ) ( h , v ) = F t ( 0 , Z ) h + v , ( h , v ) R × T Z ( Σ ) .

Let ( h , v ) R × T Z ( Σ ) such that F * ( 0 , Z ) ( h , v ) = 0 . It follows

0 = d η ( F * ( 0 , Z ) ( h , v ) ) = h d η F t ( 0 , Z ) + d η ( v ) .

Since η Σ 1 and v T Z ( Σ ) , we have d η ( v ) = 0 . On the other hand,

d η F t ( 0 , Z ) = d d t t = 0 η ( F ( t , Z ) ) = d d t t = 0 η Z ( t ) ,

which is negative by (21). Therefore, F * ( 0 , Z ) ( h , v ) = 0 implies h = 0 . Coming back to (24), we obtain v = 0 .

(3) Since f t has real coefficients, for any t , we have

F ( t , Z ¯ ) = f t ( Z ¯ ) = f t ( Z ) ¯ = F ( t , Z ) ¯ .

Proof

(of Proposition 5.4)

(1) By Lemma 5.7, the map F : R × Σ W defined by F ( t , Z ) = f t ( Z ) is a diffeomorphism. F induces a diffeomorphism

F ˜ : ( R Z ) × Σ W f = H f

given by F ˜ ( [ t ] Z , Z ) = [ f t ( Z ) ] f .

Let e : R Z S 1 be the standard diffeomorphism and n : Σ S 3 be the given by Remark 5.6. We obtain the diffeomorphisms

S 1 × S 3 e 1 × n 1 ( R Z ) × Σ F ˜ H f ,

which (by Remark 5.6 (2) and Lemma 5.7 (3)) are equivariant with respect to the involutions τ on S 1 × S 3 , id × c Σ on ( R Z ) × Σ and s f on H f .

(2) Note first that, under our assumption, we have either f I I I I I a with δ < 0 and r even (in which cases the second diagonal coefficient of f is negative), or f I I c with α δ < 0 . In the latter case, we may suppose δ < 0 (see Theorem 2.1(2)). Therefore, we may always assume that the second diagonal coefficient of f is negative.

We apply Lemma 5.7 to g = f 2 which has real coefficients and positive diagonal coefficients. Note that for f I I a with diagonal coefficients δ r , δ we have f 2 I I ˜ a with diagonal coefficients ( δ 2 ) r , δ 2 and non-diagonal coefficient c = 2 δ r .

The diffeomorphism G : R × Σ W , G ( t , Z ) = g t ( Z ) given by Lemma 5.7 applied to g induces a diffeomorphism G ˜ : ( R Z ) × Σ H g = H f 2 (as mentioned earlier) which is equivariant with respect to the involutions id × c Σ and s g .

Our primary Hopf surface H f = W f is identified with the quotient of H g by the involution f ˆ induced by f on H g , which is given explicitly by

f ˆ ( [ Z ] g ) = [ f ( Z ) ] g

and whose fixed-point locus is empty. Let J : R × Σ R × Σ be the diffeomorphism

J ( t , ( u , v ) ) t + 1 2 , ( u , v ) ,

and let j : ( R Z ) × Σ ( R Z ) × Σ be the involution induced by J . Direct computations give

(25) G J = f G ,

which obviously implies

(26) G ˜ j = f ˆ G ˜ .

Therefore, G ˜ induces a diffeomorphism

G ˆ : Q R Z × Σ j H g f ˆ = H f

(between the indicated free quotients) which is equivariant with respect to the following involutions: s f on H f and the involution t induced by id × c Σ on Q .

It suffices to note that the diffeomorphism e × n : R Z × Σ S 1 × S 3 induces a diffeomorphism Q Q , which, by Remark 5.6 (2), is equivariant with respect to the involutions t , θ .

(3) We use similar arguments noting that in this case

(27) G J = f G ,

where J : R × Σ R × Σ is given by J ( t , ( u , v ) ) ( t + 1 2 , ( u , v ) ) . Denoting by Q the quotient of R Z × Σ by the involution j induced by J , we obtain a diffeomorphism Q Q induced again by e × n which is equivariant with respect to the involutions t (defined similarly) and θ .□

Let f I I c be of the form

f z w = α z α ¯ w

with α < 1 , α C \ R . Consider the Real structure c on W given in Example 3.1:

c z w = w ¯ z ¯ .

Lemma 5.8

Let τ : ( 0 , + ) R be a C map. The map

Ψ τ : W W , Ψ τ ( Z ) R e i τ ( Z ) Z

is a diffeomorphism.

Proof

It suffices to note that Ψ τ is obviously differentiable and that, for τ , θ C ( ( 0 , + ) , R ) , we have Ψ τ Ψ θ = Ψ τ + θ . It follows that Ψ τ Ψ τ = Ψ τ Ψ τ = id W , in particular Ψ τ is bijective and its inverse is Ψ τ , which is also differentiable.□

Lemma 5.9

Let θ R be such that by 1 α α = e i θ and τ C ( ( 0 , + ) , R ) given by τ ( t ) = θ ln α ln ( t ) . Let L 1 i 1 i GL ( 2 , C ) , l : W W the associated diffeomorphism, and l l Ψ τ . Then

  1. We have

    l 1 c l = c , l 1 f l = α R 1 α α .

  2. We have

    l 1 f l = f α , l 1 c l = c .

    Therefore, l induces an equivariant diffeomorphism ( H f α , s f α ) l ˆ ( H f , s f ) .

Proof

Direct computations□

Taking into account that f α belongs to the class I V and has positive diagonal coefficients, we obtain by Proposition 5.4(1):

Proposition 5.10

Any even Real Hopf surface ( H f , s f ) with f I I c is equivariantly diffeomorphic to ( S 1 × S 3 , τ ) .

5.2 The differential-topological classification of odd Real Hopf surfaces

The goal of this section is the following classification theorem:

Theorem 5.11

Every odd Real primary Hopf surface is equivariantly diffeomorphic to ( S 1 × S 3 , μ ) , where μ is the involution

μ ( ζ , Z ) ( ζ , Z ¯ ) .

Proof

By the classification Theorem 3.15, we know that any odd Real primary Hopf surface is (equivariantly biholomorphically) isomorphic to one of the following:

  1. ( H f , σ f ) , where f I V I I I I I a I I b I I c has real coefficients and positive diagonal coefficients and σ f is induced by c f 1 2 , where f 1 2 is defined in Remark 3.8.

  2. ( H f , σ f ) , where f I I c and σ f is induced by c f 1 2 , where f 1 2 is defined in Proposition 3.11.

  3. ( H f , s f ) , where f I V has negative diagonal coefficient and s is induced by J f 1 2 , where f 1 2 is defined in Proposition 3.14.

(1) In the first case, note that the square root we use the diffeomorphism

F ˜ : ( R Z ) × Σ W f = H f

as in the proof of Proposition 5.4, and we note that the involution σ f on H f corresponds via F ˜ to the involution

( [ t ] , Z ) t + 1 2 , Z ¯

on ( R Z ) × Σ , which corresponds to the involution μ on S 1 × S 3 via e × n .

(2) Lemma 5.9 gives an equivariant diffeomorphism l ˆ : H f α H f induced by l : W W . A direct computation gives

l c f α 1 2 l 1 = c f 1 2 ,

which implies l ˆ σ f α l ˆ 1 = σ f . Therefore, ( H f , σ f ) is equivariantly diffeomorphic to ( H f α , σ f α ) , which belongs to the class considered earlier.

(3) Let f = f α with α ( 1 , 0 ) . As in the even case, put g f 2 = f α 2 , note that by Lemma 5.7, the hypersurface Σ associated with g coincides with S 3 , and consider the diffeomorphism G ˜ : ( R Z ) × S 3 H g induced by G : R × S 3 W .

Let m : R × S 3 R × S 3 be the diffeomorphism defined by

m ( t , Z ) t + 1 4 , J i Z .

The induced map m ˜ : ( R Z ) × S 3 ( R Z ) × S 3 is a diffeomorphism of order 4 of ( R Z ) × S 3 whose second power m ˜ 2 is given by the formula:

m ˜ 2 ( [ t ] , Z ) = t + 1 2 , Z .

Direct computations give

(28) G m = J f 1 2 G .

Recall that our primary Hopf surface H f = W f is identified with the (free) quotient of H g by the involution f ˆ induced by f on H g , and, via G ˜ , f ˆ corresponds to the involution

j : ( R Z ) × S 3 ( R Z ) × S 3 , j ( [ t ] , Z ) = t + 1 2 , Z .

Therefore, G ˜ induces a diffeomorphism

G ˆ : Q ( R Z ) × S 3 j H f .

Formula (28) shows that, via G ˆ , the involution s f on H f corresponds to the involution m ˆ induced by m ˜ on Q . Identifying R Z with S 1 and Q with Q in the usual way, we see that the involution m ˆ on Q which corresponds to m ˆ is given by

m ˆ ( [ ζ , Z ] ) = [ e i π 2 ζ , J i Z ] .

Via the diffeomorphism a ˆ given by Lemma 5.3 the involution on S 1 × S 3 which corresponds to m ˆ is

μ : S 1 × S 3 S 1 × S 3 , μ ( ζ , Z ) = ( ζ , i Z ¯ ) .

It suffices to note that ( ζ , Z ) ( ζ , e i π 4 Z ) defines an equivariant diffeomorphism ( S 1 × S 3 , μ ) ( S 1 × S 3 , μ ) .□

Remark 5.12

Let μ 0 : S 1 × S 3 S 1 × S 3 be the involution

μ 0 ( ζ , Z ) = ( ζ , Z ) .

Let Φ be the R -linear orientation preserving isometry

( z , w ) ( ( z ) + i ( w ) , ( z ) + i ( w ) )

and ϕ : S 3 S 3 the induced diffeomorphism of S 3 . Let ψ : S 1 × S 3 S 1 × S 3 be the diffeomorphism ( ζ , ( m , n ) ) ( ζ , ( m , ζ n ) ) . The composition ψ ( id × ϕ ) is an equivariant diffeomorphism

( S 1 × S 3 , μ ) ( S 1 × S 3 , μ 0 ) .

Therefore, in the classification Theorem 5.11, we may replace μ by μ 0 .

Proof

It suffices to note that, putting μ ( ζ , ( z , w ) ) ( ζ , ( z , w ) ) , we have

( id × ϕ ) μ = μ ( id × ϕ )

and that μ 0 ψ = ψ μ .□

5.3 The real locus H s and the quotient H s

5.3.1 The real locus

Note first that the fixed-point locus M σ of any involution σ of a differentiable manifold M is a submanifold of M . This follows by the equivariant slice theorem ([10, Theorem 5.6]), which shows that an point x X σ has a σ -invariant open neighborhood which is equivariantly diffeomorphic to ( T x X , σ * , x ) .

We have proved that any even Real structure on a primary Hopf surface H f with f I V I I I I I a I I b I I c is equivalent to the standard Real structure s f (which is induced by the standard conjugation c when f is with real coefficients, and by c when f I I c ). Therefore, for describing the real locus of an arbitrary even Real primary Hopf surface, it suffices to consider only this standard Real structure.

The fixed-point locus W c (respectively W c ) is

W c = R 2 \ { 0 } , W c = { ( z , z ¯ ) z C * } .

Let f c , respectively, f c be the contraction induced by f on W c (respectively W c ). The canonical maps

W c f c [ W f ] s f = H f s f , W c f c [ W f ] s f = H f s f

are obviously diffeomorphisms.

Proposition 5.13

The fixed-point locus H f s f of a standard even Real primary Hopf surface is diffeomorphic either to the torus T 2 or to a Klein bottle according to the sign of the determinant of the real part of f. In particular, if f I I c , then H f s f is a torus.

Proof

By the classification Theorem 5.1, we have only two equivariant diffeomorphism classes of even Real primary Hopf surface. The proof of this theorem shows any even Real primary Hopf surface is equivariantly diffeomorphic to either ( H g , s g ) or ( H h , s h ) , where g , h : W W are the contractions defined by

g ( z , w ) = 1 2 z , 1 2 w , h ( z , w ) = 1 2 z , 1 2 w .

Moreover, this proof also shows that if f I I c , then ( H f , s f ) is equivariantly diffeomorphic to ( H g , s g ) . It is easy to see that the quotient W c g c is a torus, whereas the quotient W c h c is a Klein bottle.□

Note that, for f I I c , the projections ( z , z ¯ ) z , ( z , z ¯ ) z ¯ induce identifications

W c f c C * α , W c f c C * α ¯ ,

where α , α ¯ are the coefficients of f . This shows that, in this case, the Real locus H f s f comes with a canonical (non-oriented) conformal structure, which is conformally isomorphic to the elliptic curve E α C * α . In other words:

Remark 5.14

When f belongs to the subclass I I c , the real locus H f s f is a 2-torus which comes with a natural (non-oriented) conformal structure.

For the odd Real structures, we have:

Remark 5.15

The Real locus of any odd Real primary Hopf surface is empty.

Proof

Let ( H f , σ ) be an odd Real Hopf surface and σ ˆ Aut h ( W ) be a lift of σ to W such that σ ˆ 2 = f . Suppose that x = [ ( z , w ) ] H f is a fixed-point of σ . Therefore, there exists k Z such that σ ˆ ( z , w ) = f k ( z , w ) . Therefore, σ ˆ 2 k 1 ( z , w ) = ( z , w ) , which implies σ ˆ 2 ( 2 k 1 ) ( z , w ) = σ ˆ 2 k 1 ( z , w ) = ( z , w ) . We obtain f 2 k 1 ( z , w ) = ( z , w ) . Since f acts freely on W , it follows 2 k 1 = 0 (contradiction).□

5.3.2 The quotient of a Real Hopf surface by its involution

Note first that the quotient X σ associated with any Real complex surface ( X , σ ) is a topological 4 manifold. This follows using

  1. The equivariant slice theorem quoted above.

  2. The classification of anti-linear involutions on a complex vector space (see for instance Remark 3.3 in this article).

  3. The homeomorphism R 2 id R 2 R 2 induced (for instance) by the id R 2 -invariant map

    β : R 2 R 2 , β ( ρ cos ( θ ) , ρ sin ( θ ) ) = ( ρ 2 cos ( 2 θ ) , ρ 2 sin ( 2 θ ) ) .

    Using a complex coordinate ζ on R 2 , this map is given by ζ ζ 2 .

Taking into account this remark we will describe the quotient associated with a Real primary Hopf surface as a topological 4-manifold.

We have seen that any even Real primary Hopf surface is isomorphic to ( H f , s f ) , where either f I V I I I I I a I I b I I c with real coefficients or f I I c . The involution s f is induced by the anti-linear involution c , respectively, c on W .

The quotient

Q f H f s f

can be identified with the quotient W W c , respectively, W W c by the contraction f induced by f on W , respectively W .

We obtain a decomposition C 2 = V + V of C 2 as direct sum of c -invariant (respectively c -invariant) 2-dimensional real linear subspaces V ± such that c V ± = ± id V ± (respectively c V ± = ± id V ± ).

Therefore, the quotient C 2 c ( C 2 c ) can be identified with the quotient

V + × V id V R 4 .

Here, we have used linear isomorphisms V ± R 2 and the homeomorphism

R 2 id R 2 R 2

induced by β . The image of W c ( W c ) via this homeomorphism is R 4 \ { 0 } and the image of the fixed-point locus W c ( W c ) in W ( W ) is ( R 2 \ { 0 } ) × { 0 } . Using the classification Theorem 5.1, we see that in all cases the contraction induced by f on W c ( W c ) is orientation preserving. Therefore, we obtain the following result, which describes the quotient Q f associated with an even Real primary Hopf surface, as well as the image of the fixed-point locus H f s f in this quotient:

Proposition 5.16

With the aforementioned notations, we have

  1. The quotient Q f H f s f can be identified topologically with the quotient

    Q f = R 4 \ { 0 }

    of R 4 \ { 0 } by the cyclic group generated by an orientation preserving contraction . Therefore, in all cases, Q f is homeomorphic to S 1 × S 3 .

  2. leaves invariant the punctured plane ( R 2 \ { 0 } ) × { 0 } and, via the above identification, the fixed point locus H f s f corresponds to the quotient of R 2 \ { 0 } by the contraction 0 induced by .

  3. 0 is orientation preserving if and only if the diagonal coefficients of f have the same sign. If this is the case, R 2 \ { 0 } 0 is a 2-torus. If the diagonal coefficients of f have opposite signs, R 2 \ { 0 } 0 is a Klein bottle.

Taking into account the classification Theorem 5.11 and Remark 5.12, we obtain the following simple description of the quotient H σ of any odd Real primary Hopf surface ( H , σ ) :

Proposition 5.17

Let ( H , σ ) be an odd Real primary Hopf surface. The quotient H σ can be identified with S 1 × S 3 and the canonical projection

H H σ

is a double cover whose (non-trivial) deck transformation is an anti-holomorphic involution.

Acknowledgements

I am indebted to my PhD advisor Andrei Teleman for suggesting me the problems treated in this article and for many helpful ideas. I also thank Karl Oeljeklaus and Anne Pichon for their advices and their interest in my results. I am grateful to the referee for the careful reading of the paper and the valuable suggestions.

  1. Funding information: The author states no funding involved.

  2. Author contributions: The author confirms the sole responsibility for the conception of the study, presented results and manuscript preparation.

  3. Conflict of interest: The author states no conflict of interest.

References

[1] M. Atiyah, K-theory and Reality, Quart. J. Math. 17 (1966), no. 1, 367–386. 10.1093/qmath/17.1.367Suche in Google Scholar

[2] W. Barth, K. Hulek, Ch. Peters, and A. Van de Ven. Compact Complex Surfaces, Springer-Verlag, Berlin, Heidelberg, New York, 2004. 10.1007/978-3-642-57739-0Suche in Google Scholar

[3] P. Frediani, Real Kodaira surfaces, Collectanea Mathematica 55 (2004), 61–96. Suche in Google Scholar

[4] B. Gross and J. Harris, Real algebraic curves, Ann. scient. Éc. Norm. Sup., 4e série, Tome 14 (1981), 157–182. 10.24033/asens.1401Suche in Google Scholar

[5] K. Kodaira, On the structure of compact complex analytic surfaces, II, Amer. J. Math., The Johns Hopkins University Press, 88 (1966), no. 3, 682–721. 10.2307/2373150Suche in Google Scholar

[6] M. Namba, Automorphism groups of Hopf surfaces, Toohoku Math. J. 26 (1974), 133–157. 10.2748/tmj/1178241239Suche in Google Scholar

[7] C. Okonek, and A. Teleman, Abelian Yang-Mills theory on real Tori and Theta divisors of Klein surfaces, Commun. Math. Phys. 323 (2013), 813–858. 10.1007/s00220-013-1793-zSuche in Google Scholar

[8] R. Silhol, Real algebraic surfaces, Lecture Notes in Mathematics, vol. 1392, Springer-Verlag, Berlin, 1989. 10.1007/BFb0088815Suche in Google Scholar

[9] A. Teleman, Non-Kählerian compact complex surfaces, In: D. Angella, L. Arosio, E. DiNezza, (Eds), Complex Non-Kähler Geometry. Lecture Notes in Mathematics, vol. 2246, Springer, Cham, 2019. 10.1007/978-3-030-25883-2_3Suche in Google Scholar

[10] Tammo tom Dieck, Transformation Groups, De Gruyter, Berlin-New York, 1987. 10.1515/9783110858372Suche in Google Scholar

[11] J. Wehler, Versal deformation of Hopf surfaces, J. für die reine und angewandte Mathematik 328 (1981), 22–32. 10.1515/crll.1981.328.22Suche in Google Scholar

Received: 2023-12-04
Revised: 2024-08-17
Accepted: 2024-10-18
Published Online: 2024-12-06

© 2024 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Heruntergeladen am 17.10.2025 von https://www.degruyterbrill.com/document/doi/10.1515/coma-2024-0007/html?lang=de
Button zum nach oben scrollen