What is an OT-manifold? March 20, 2014
Transcription
What is an OT-manifold? March 20, 2014
What is an OT-manifold? March 20, 2014 Non-K¨ahler geometry is a mysterious field and new examples were built recently. The class of OT-manifolds form such an example. Their construction was introduced by Oeljeklaus and Toma in [5]. Our goal in this series of talks is to describe this construction and give recent developments about their study. Contents 1 Number-theoretic preliminaries 1.1 Field extensions . . . . . . . . . . . . 1.2 Algebraic and transcendental numbers . 1.3 Number fields . . . . . . . . . . . . . 1.4 Galois group of a field extension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2 3 4 5 2 Construction of OT-manifolds 7 2.1 A special class of Cousin groups . . . . . . . . . . . . . . . . . . 8 2.2 OT-manifolds of simple type . . . . . . . . . . . . . . . . . . . . 10 3 Geometrical properties of OT-manifolds 11 3.1 Non-K¨ahler structure . . . . . . . . . . . . . . . . . . . . . . . . 11 3.2 Locally conformally K¨ahler structure . . . . . . . . . . . . . . . . 13 1 1 Number-theoretic preliminaries We start by recalling some facts from number theory, mainly without proof. A classical references is [2]. In the following, unless specified otherwise, K always denotes a field and K[X] is the ring of polynomials with coefficients in K. 1.1 Field extensions Definition 1.1. A field extension of a field K is a pair (L, i) where L is a field and i : K → L is a morphism. It is denoted by K ⊂ L. Notice that i is injective because a field only has trivial ideals. Definition 1.2. When one sees L as a K-vector space, its dimension dimK (L) is called the degree of L over K and is denoted by [L : K]. An extension is said to be finite if its degree [L : K] is finite. Proposition 1.3. Let K ⊂ L ⊂ M be two finite field extensions, then one has [M : K] = [M : L] × [L : K]. Proof : Choose an L-basis (bj )j∈J of M and a K-basis (ai )i∈I of L. Then, (ai bj )i∈I,j∈J is a K-basis for M . Let K ⊂ L be a field extension and A be a subset of K (not necessarily finite). One notes K(A) (resp. K[A]) the smallest subfield (resp. ring) of L containing A. It is clear that K[A] ⊂ K(A) hence K(A) is the quotient field of K[A]. Definition 1.4. Let K ⊂ L be a field extension and let x ∈ L. The extension K ⊂ K(x) is called simple and x is said to be a primitive element for K(x). Proposition 1.5. Let K ⊂ L be a field extension and A, B be two subsets of L. Then one has K(A ∪ B) = K(A)(B) = K(B)(A). Proof : K(A) ⊂ K(A ∪ B) and B ⊂ K(A ∪ B) imply K(A)(B) ⊂ K(A ∪ B) while K ⊂ K(A)(B) and A ∪ B ⊂ K(A)(B) imply K(A ∪ B) ⊂ K(A)(B). 2 1.2 Algebraic and transcendental numbers Let K ⊂ L be a field extension and x ∈ L. Define the following map θx : K[X] −→ K[x] P 7−→ P (x). The map θx may be injective or not. We study these two cases. Case 1 : θx is not injective. Since K[X] is principal, one has ker θx = (P ) for some polynomial P . Up to choosing P irreducible and monic, it is unique and called the minimal polynomial of x, denoted µx . Moreover, x is said to be algebraic over K of degree degK (x) := deg µx . Because µx is irreducible, the ideal (µx ) is a prime ideal of K[X] hence maximal1 so θx is an isomorphism between the fields K[X]/(µx ) and K[x] = K(x). Case 2 : θx is injective: for all P ∈ K[X], P (X) 6= 0 and x is called transcendental over K. √ Example 1.6. The real number 2 is algebraic over Q while π is transcendental. Proposition 1.7. Let K ⊂ L be a field extension and x ∈ L be algebraic over K. Then, [K(x) : K] = degK (x). Proof : One shows that 1, x, x2 , ..., xdegK (x)−1 is a K-basis of K(x). Definition 1.8. Let P ∈ K[X] be a polynomial. A splitting field of P is an extension L of K in which splits, i.e. decomposes as a product of linear factors, and such that P doesn’t split on any proper subfield of L. Proposition 1.9. Every polynomial has a splitting field. Definition 1.10. A polynomial P ∈ K[X] is separable if it only has simple roots in its splitting field. A field extension K ⊂ L is separable if every x ∈ L is separable over K, i.e. transcendental over K or algebraic over K with its minimal polynomial µx ∈ K[X] being separable. Theorem 1.11 (Primitive Element Theorem). Let K ⊂ L be a finite and separable field extension. Then, there is a primitive element x ∈ L for L over K, i.e. L = K(x). 1 In a principal ring, every non-trivial prime ideal is maximal. This is not true in general, while the implication maximal⇒prime is always true. 3 1.3 Number fields Definition 1.12. A number field is a finite extension Q ⊂ K over the rationals. Proposition 1.13. A number field is always a separable extension. Proof : A polynomial is separable if and only if it is prime with its formal derivative. Because Q has characteristic 0, for every irreducible polynomial P ∈ Q[X] of degree > 1, one has P 0 6= 0 hence P has simple roots. By the primitive element theorem, a number field K is always of the form Q(α) for α ∈ K an algebraic number over Q. √ √ √ √ Example 1.14. Q( 2, 3) = Q( 2 + 3). Definition 1.15. An element x ∈ C is called an algebraic number if it is algebraic over Q. Moreover, it is an algebraic integer if its minimal polynomial µx only has integer coefficients, i.e. µx ∈ Z[X]. The set of algebraic integers of a number field K is denoted OK . Proposition 1.16. One has OK ∼ = Zn . It is a ring and its quotient field is K. Definition 1.17. An algebraic integer is called an algebraic unit if it is a divisor of 1, i.e. if its inverse is also an algebraic integer. If K is a number field, the ∗ . group of units of the ring OK is denoted by OK Theorem 1.18. Let K be a number field of degree n. Then there are exactly n distinct injective homorphisms from K to C. Proof : Call α a primitive element for K and let µα be its minimal polynomial, which is of degree n. Name α1 = α, α2 , ..., αn the n distinct roots of µα . Every element of K = Q(α) can be written uniquely in the form a0 +a1 α+...+an−1 αn−1 where the ai ’s are rational numbers. For k ∈ {1, ..., n}, define σk : K → C by σk (a0 + a1 α + ... + an−1 αn−1 ) = a0 + a1 αk + ... + an−1 αkn−1 . The fact that the σk is a field homomorphism is left as an exercise. Now, if σ : K → C is a field homomorphism, then µα (σ(α)) = σ(µα (α)) = 0 so σ(α) is a root of µα , say αk0 for k0 ∈ {1, ..., n}. We have σ(α) = αk0 so σ(a0 + a1 α + ... + an−1 αn−1 ) = σk0 (a0 + a1 α + ... + an−1 αn−1 ) for all a0 , ..., an−1 ∈ Q, which shows that σ = σk0 . 4 Definition 1.19. The αi ’s are called the conjugates of α over Q and the fields Q(αi ) are the conjugate fields of K. Remark 1.20. Let K be a number field and call σ1 , ..., σs its real embeddings and σs+1 , ..., σs+2t the complex ones. The norm of an element k ∈ K is the product σ1 (k)×...×σs+2t (k). In particular, the norm of a unit is ±1. Conversely, if the norm of an element of OK is ±1 then it is a unit. Definition 1.21. Let K be a number field. If σ : K → C is an injective homomorphism which satisfies σ(K) ⊂ R, it is called a real embedding of K, otherwise it is called a complex embedding. √ √ 3 embedding sending α to 2 Example 1.22. Let K = Q(α) with α = 3 2. The √ √ 2iπ 4iπ is real, while the two embeddings sending α to e 3 3 2 and e 3 3 2 are complex ones. Let σ : K → C be a complex embedding, σ : K −→ C is another x 7−→ σ(x) complex embedding of K, called conjugate to σ. One has σ 6= σ and σ = σ, hence the set of complex embeddings of K contains an even number of elements, which we denote 2t. On the other hand, note s the number of real embeddings of K. One has [K : Q] = s + 2t. The following theorem gives the structure of the group of units of a number field. One can find a proof in chapter 2, section 4 of [2]. Theorem 1.23 (Dirichlet’s Unit Theorem). Let K be a number field of degree s + 2t where s is its number of real embeddings and 2t the number of complex ∗ ∼ embeddings. One has OK = C × Zs+t−1 where C is a finite group of roots of unity of K. ∗ ∼ Remark 1.24. If s > 0, then OK = {±1} × Zs+t−1 . Indeed, if ε ∈ K is a k root of unity, one has σ1 (ε ) = 1 for some integer k > 0. Because σ1 is a homomorphism, this implies that σ1 (ε)k = 1, i.e. σ1 (ε) = ±1. The injectivity of σ1 (and the fact that σ1 (1) = 1) forces ε = ±1. 1.4 Galois group of a field extension Definition 1.25. Let K ⊂ L be a field extension. An automorphism σ : L → L is called a K-automorphism if σ|K = idK . 5 Let P ∈ K[X] be a polynomial, L its splitting field over K and σ be a K-automorphism of L. If α ∈ L is a root of P then σ(α) is also a root of P . Definition 1.26. Let K ⊂ L be a field extension. The Galois group of this extension, denoted AutK (L), is the group of K-automorphisms of L. If L is the splitting field of a polynomial P ∈ K[X], AutK (L) is called the Galois group of P . The following should be clear: Theorem 1.27. Let P [X] be a polynomial of degree m and A = {α1 , ..., αn } ⊂ L the set of the n 6 m distinct roots of P in its splitting field L. Then AutK (L) is isomorphic to a subgroup of the symmetric group Sn . Proof : This isomorphism is given by AutK (L) −→ Sn . σ 7−→ σ|A Theorem 1.28. If P ∈ K[X] is a separable polynomial and L is its splitting field, one has # AutK (L) = [L : K]. Definition 1.29. Let K ⊂ L be an algebraic field extension. It is called normal or quasi-Galois if every irreducible polynomial P ∈ K[X] having a root in L has all its other roots in L. Example 1.30. The extension K ⊂ K is normal. If K ⊂ Ω is an algebraic closure of K, it is also normal. Definition 1.31. Let K ⊂ L be a finite field extension. It is called a Galois extension if it is normal and separable. Proposition 1.32. For a finite field extension K ⊂ L, the following properties are equivalent: • K ⊂ L is a Galois extension; • # AutK (L) = [L : K]; • Every irreducible polynomial P ∈ K[X] which admits a root in L is separable and splits in L; • L is the splitting field of a separable polynomial P ∈ K[X]. 6 2 Construction of OT-manifolds Let K be an algebraic number field and n := [K : Q] be its degree over Q. The field K admits exactly s real embeddings and 2t complex ones (with n = s + 2t). We call the real embeddings σ1 , ..., σs and σs+1 , ..., σs+2t the complex ones, ordered in such a way that σs+i = σ s+i+t for 1 6 i 6 t. For the construction we assume that s > 0 and t > 0. Call m := s + t and consider the map σK : K −→ Rs × Ct ⊂ Cm . a 7−→ (σ1 (a), ..., σm (a)) The set σK (OK ) is a lattice of real rank n in Cm , so OK acts properly discontinuously on Cm . Moreover, there is an action of the multiplicative group K ∗ = K \ {0} on Cm given by az := (σ1 (a)z1 , ..., σm (a)zm ) for a ∈ K and z = (z1 , ..., zm ) ∈ Cm . In particular, σK (OK ) is stable under this action. ∗ : Define the following subgroup of OK ∗,+ ∗ OK := {u ∈ OK | σi (u) > 0 for all 1 6 i 6 s}. Let H be the complex upper half-plane. One has a free action of the semi∗,+ direct product OK n OK on Hs × Ct . This action is defined by (u, a).z := (σ1 (u)z1 + σ1 (a), ..., σm (u)zm + σm (a)) ∗,+ for (u, a) ∈ OK n OK and z = (z1 , ..., zs , zs+1 , ..., zm ) ∈ Hs × Ct . ∗,+ Define ` : OK → Rm by `(a) = (log |σ1 (a)|, ..., log |σs (a)|, 2 log |σs+1 (a)|, ..., 2 log |σm (a)|). ∗,+ By Dirichlet’s units theorem, `(OK ) is a lattice of rank m − 1 in the real Preal m m (m−1)-dimensional vector space L := {x ∈ R | i=1 xi = 0}. The projection pr : L → Rs given by the s first coordinates is surjective so there are subgroups ∗,+ U of rank s of OK such that pr(`(U )) is a lattice of rank s of Rs . Such an U is called admissible. Lemma 2.1. The set (Hs ×Ct )/σK (OK ) is diffeomorphic to (R>0 )s ×(S1 )n and ∗,+ the action of a subgroup U of OK on it is properly discontinuous and cocompact if and only if U is admissible. 7 Proof : the map are real, Hs × C t Write zj = xj + iyj for 1 6 j 6 s. One has Hs ∼ = (R>0 )s × Rs via (z1 , ..., zs ) 7→ (y1 , ..., ys , x1 , ..., xs ). Since the s first embeddings of K an element a ∈ OK sends a point (y1 , ..., ys , x1 , ..., xs , zs+1 , ..., zm ) ∈ to the point (y1 , ..., ys , x1 + σ1 (a), ..., xs + σs (a), zs+1 + σs+1 (a), ..., zm + σm (a)) ∈ Hs × Ct . This shows that (Hs ×Ct )/σK (OK ) ∼ = (R>0 )s ×((Rs ×Ct )/σK (OK )), which s 1 n is in turn diffeomorphic to (R>0 ) × (S ) . ∗,+ Since the quotient of Hs × Ct by OK n OK is the same as the quotient of ∗,+ s t (H × C )/σK (OK ) by OK and since the projection map (R>0 )s × (S1 )n → ∗,+ (R>0 )s is proper, the action of a subgroup U of OK on (Hs × Ct )/σK (OK ) is proper if and only if it is proper on (R>0 )s . ∗,+ Indeed, the action of OK on (Hs ×Ct )/σK (OK ) send an element (y1 , ..., ys ) ∈ ∗,+ s (R>0 ) to (σ1 (u)y1 , ..., σs (u)ys ), for u ∈ OK . This and the cocompactness of the action is now equivalent to the definition of an admissible subgroup. ∗,+ The previous lemma shows that the action of OK n OK on Hs × Ct is not properly discontinuous when t > 1, while it is always the case for every admissible ∗,+ subgroup of OK . ∗,+ Definition 2.2. Choose an admissible subgroup U of OK , the quotient X := s t X(K, U ) = (H × C )/(U n OK ) is a compact complex manifold 2 of dimension m. This manifold is called an Oeljeklaus-Toma manifold, or an OT-manifold for short. 2.1 A special class of Cousin groups Lemma 2.3 ([5], lemma 2.4). Every holomorphic function on Hs × Ct /σK (OK ) is constant. Proof : The proof is divided into two steps: Step 1: For every v ∈ Hs , one sees that ({v} × Ct )/σK (OK ) is dense in (v + Rs × Ct )/σK (OK ). 2 When one has a fixed-point free and properly discontinuous action of a group G on a complex manifold W , then the quotient W/G has a complex structure induced from W . See [4], Theorem 2.2 for a proof. 8 0 Let V be the connected component of the closure of σK (OK ) in Rs containing 0 0. Let M := σK−1 (V ) ⊂ OK . Assume that V 6= Rs . This implies that the rank of M is strictly smaller than n. Now let α ∈ OK be a primitive element for K. There is a multiplicative action on OK of α, and the induced linear action of α on Rs leaves V invariant. This implies that M is also invariant under the action of α, which implies that the characteristic polynomial of α has a factor of same degree as the rank of M , a contradiction. Step 2: For every holomorphic function f on Hs × Ct /σK (OK ) and v ∈ Hs , the map f is bounded on (v + Rs × Ct )/σK (OK ) ∼ = (S1 )n . In particular the lift fe of f to Hs × Ct is bounded on (v + Rs ) × Ct , hence constant on {v} × Ct thanks to Liouville’s theorem. The observation in Step 1 implies that the function fe is constant on (v+Rs )×Ct , and hence constant on Hs ×Ct by the identity principle. Remark 2.4. This implies that the complex Lie group Cm /σK (OK ) is a Cousin group, i.e. a connected complex Lie group having no non-constant holomorphic functions. In [7], Vogt exhibits a special class of Cousin groups by showing 3 : Theorem 2.5. Let C = Cn /Λ be a Cousin group. Then the following conditions are equivalent: 1. The space H 1 (C, OC ) is finite-dimensional. 2. Let P = (In S) be a period basis of Λ. Then there exist constants C > 0 and a > 0 such that kt σS + t τ k > C exp(−a|σ|) for all σ ∈ Zn \ {0} and all τ ∈ Zm , where n + m is the rank of Λ. 3. Every topologically trivial line bundle over C comes from a representation of Λ. In fact, thanks to the following generalization of Liouville’s theorem on approximation of algebraic numbers, one can show that the Cousin groups Cm /σK (OK ) arising in the construction of OT-manifolds belong to this special class. 3 The theorem given in [7], p. 208 has 8 equivalent assertions, here we only recall three of them. 9 Theorem 2.6. Let α1 , ..., αm be algebraic numbers, of respective degrees nk , with deg Q(α1 , ..., αm ) = n, and let P (z1 , ..., zm ) = N1 X k1 =0 ··· Nm X km ∈ Z[z1 , ..., zm ]. ak1 ,...,km z1k1 · · · zm km =0 If P (α1 , ..., αm ) is non-zero, then one has the inequality 1−δn |P (α1 , ..., αm )| > L(P ) m Y L(αk )−δNk n/nk , k=1 with δ = 1 if all the αi are real, δ = 1/2 otherwise and where L(P ) is the sum of the absolute values of the coefficients of P (and L(α) is the quantity L(µ), with µ being the minimal polynomial of α). Theorem 2.7. Let Λ ⊂ Cn be a lattice such that C = Cn /Λ is a Cousin group with a period basis whose coefficients are all algebraic numbers, then C satisfies the equivalent conditions of theorem 2.5. In particular, Cousin groups arising in the construction of OT-manifolds belong to this family. 2.2 OT-manifolds of simple type We recall two lemmas and a definition (see [5]) for further use: ∗,+ Lemma 2.8. Let U be a subgroup of OK which is not contained in Z. Then the two following conditions are equivalent: 1. The action of U on OK admits a (non-trivial) proper invariant sub-module of lower rank. 2. There exists a proper intermediate field extension Q ⊂ K 0 ⊂ K with ∗,+ U ⊂ OK 0 . Definition 2.9. We say that X(K, U ) is of simple type if U does not satisfy one of the equivalent conditions of the previous lemma. ∗,+ Lemma 2.10. Let Q ⊂ K 0 ⊂ K an intermediate field extension with U ⊂ OK 0 0 0 an admissible subgroup for K. Let s , 2t be the number of real and complex embeddings of K 0 respectively. Then s = s0 , t0 > 0 and U is admissible for K 0 . For more details, see [5]. 10 3 3.1 Geometrical properties of OT-manifolds Non-K¨ ahler structure Definition 3.1. Let M be a complex manifold with complex structure J. A Riemannian metric g = h · , · i is compatible with J if hJX, JY i = hX, Y i for all vector fields X, Y ∈ T X. A complex manifold with a compatible Riemannian metric is called a Hermitian manifold. The 2-form ω(X, Y ) := g(JX, Y ) is the associated fundamental form. One recovers g by g(X, Y ) = ω(X, JY ). Finally, g is a K¨ ahler metric if ω is closed i.e. dω = 0 and ω is called the K¨ ahler form. A complex manifold is called a K¨ ahler manifold if it admits a K¨ahler metric and a non-K¨ ahler manifold otherwise. Example 3.2. Cn , Pn (C) (ω = i∂∂ log |z|2 ), complex tori, Riemann surfaces are K¨ahler manifolds. Proposition 3.3. Let (X, g) be a compact K¨ahler manifold. Then there exists a decomposition M H p,q (X). H k (X, C) = p+q=k Moreover, one has H p,q (X) = H q,p (X). As a consequence, the first Betti number b1 (X) = dimC H 1 (X, C) of a compact K¨ahler manifold X is equal to 2h1,0 = 2 dim H 1 (X, OX ). Example 3.4. Hopf surfaces. Lemma 3.5. Let u1 , . . . , us be some s ≥ 1 multiplicatively independent units in UK . Then either Q(u1 , . . . , us ) is a proper subfield of K or UK contains s multiplicatively independent units v1 , . . . , vs ∈ S(u1 , . . . , us ) := {ak11 · · · aks s | ki ∈ N} of degree d each. Definition 3.6. A reciprocal unit is an algebraic unit x ∈ C such that the inverse 1/x is one of its conjugate over Q. Proposition 3.7. Thefirst two Betti numbers of an OT-manifold X = X(K, U ) s are b1 = s and b2 = . For b2 we need to assume that U contains a non2 reciprocal unit. 11 Proof : The manifold X is an Einlenberg-MacLane space of type K(G, 1), that is, it has a fundamental group isomorphic to G and its universal covering is contractible. By theorem I in [3] (pp. 482-483), the cohomology groups of X with coefficients in Q are the same as those of its fundamental group. So we are left with the computation of H 1 (U n OK , Q) and H 2 (U n OK , Q). As in [5], we use the Lyndon-Hochschild-Serre spectral sequence associated to the short exact sequence 0 → OK → U n OK → U → 0. One has E2p,q = H p (U, H q (OK , Q)) ⇒ H p+q (U n OK , Q) and the following exact sequence: E20,1 }| { z 1 U 1 OK 1 0 → H (U, Q ) → H (U n OK , Q) → H (OK , Q) → H 2 (U, QOK ) → H 2 (U n OK , Q)1 → H 1 (U, H 1 (OK , Q)) | {z } 1,1 E2 where H 2 (U n OK , Q)1 is defined by the exact sequence 0 → H 2 (U n OK , Q)1 → H 2 (U n OK , Q) → H 2 (OK , Q)U . {z } | 0,2 E2 See for instance [6]. If one proves that E20,1 = E20,2 = E21,1 = 0, the result follows. For the fact that E20,1 = E21,1 = 0, the proofs are in [5], proposition 2.3, so we will not repeat it here (the existence of a non-reciprocal unit in U is not needed). ∼ As for the group E20,2 = H 2 (OK , Q)U P = Alt2 (OK , Q)U , recall that an element of Alt2 (OK , Q)U is of the form γ = i<j ai,j σi ∧ σj with ai,j ∈ C. Moreover, the U -invariance of γ means that for every pair (i, j) such that ai,j 6= 0, one has σi (η)σj (η) = 1 for all η ∈ U . Now when U contains a non-reciprocal unit η0 and therefore the relation σi (η0 )σj (η0 ) = 1 can never hold for any choice of i < j. Hence γ is trivial and so is the group E20,2 . Proposition 3.8. An OT -manifold X = X(K, U ) is non-K¨ahler. 12 Proof : One has dimC H 1 (X, C) = s and dim H 1,0 = dim H 0,1 > s. Let ρ : U → C be a group homomorphism. We will associate to this homomorphism a principal C-bundle above X and show that if this bundle is trivial then ρ is also trivial, which will lead to the desired inequality. Consider the action of U on the product ((Hs × Ct )/σK (OK )) × C given by u.(x, z) := (u.x, z + ρ(u)) for all u ∈ U , x ∈ (Hs × Ct )/σK (OK ) and z ∈ C. The quotient F of ((Hs × Ct )/σK (OK )) × C under this action of U is a principal C-bundle above X. Indeed, the action of ξ ∈ C on an element [x, z] := U.(x, z) ∈ F is given by ξ.[x, z] := [x, z + ξ]. If F is trivial, it must have a global section, i.e. there is a holomorphic function f : (Hs × Ct )/σK (OK ) → C satisfying the equality f (u(x)) = f (x) − ρ(u) for all u ∈ U and u ∈ U . By lemma 2.3, the function f is constant and this implies ρ ≡ 0. This proves that h1,0 > s. If an OT manifold was K¨ahler we would have s > 2s, a contradiction. In theoretical physics, string theory, non-K¨ahler examples are needed to work with. 3.2 Locally conformally K¨ ahler structure f its universal covering. Definition 3.9. Let M be a complex manifold and M Then M is called a locally conformally K¨ ahler manifold ( LcK manifold, for short) if there exists a representation ρ : π1 (M ) → R>0 and a positive closed f such that g ∗ ω = ρ(g)ω for all g ∈ π1 (M ). (1, 1)-form ω on M Example 3.10. A K¨ahler manifold is of course LcK. A Hopf manifold is also LcK. P Proof : Let ω := i dzj ∧ dzj be the K¨ahler metric on Cn \ {0} and ρ : Z −→ R>0 . λ 7−→ |λ|2 Example 3.11. When t = 1, an OT-manifold is LcK. 13 Theorem 3.12. An OT-manifold X(K, U ) admits an LcK metric if and only if the following holds: for all u ∈ A, |σs+1 (u)| = ... = |σs+t (u)|. (1) Proof : Assume that ω is a K¨ahler metric on Hs × Ct upon which U n OK acts by homotheties. Then ω can be written as X ω= hij dzi ∧ dz j . i,j=1,s+t Since ω descends to (Hs × Ct )/σK (OK ) ∼ = (R>0 )s × (S1 )n , by averaging ω on the (S1 )n part we can assume that all the coefficients of ω depend only on z1 , . . . , zs . Now we show that all hii , i = s + 1, . . . , s + t are constant. Assume this is not the case, i.e.: ∂hii 6= 0 for some 1 ≤ k ≤ s. ∂zk From the assumption that ω is a K¨ahler form, hence closed, we get ∂hki 6= 0, ∂zi a contradiction with the fact that hki does not depend on zi for i > s + 1. Since ω is an LcK metric on X, there is a map ρ such that g ∗ ω = ρ(g) for all g ∈ U n OK . Because there are constant coefficients in ω, ρ is equal to ρ(u, a) = |σs+1 (u)|2 = ... = |σs+t (u)|2 for all (u, a) ∈ U n OK . We are left with checking that this condition is sufficient. Let X(K, U ) be an OT-manifold with U satisfying condition (1) and define the following real function on Hs × Ct : ϕ(z) := s Y i z − zj j=1 j ! 1t + t X |zs+k |2 . k=1 This definition of ϕ is very natural: when t = 1, this function is the same as the function F defined in [5], example p. 169, to show that when t = 1 an OT-manifold admits an LcK metric. It is enough to prove that it is a K¨ahler potential on Hs × Ct . For this, we will see that the matrix (∂zp ∂zq ϕ1 ) is positive definite, where we set ϕ1 (z) = 14 Q s i j=1 zj −zj 1t . For all q ∈ {1, ..., s}, one has: ∂zq ϕ1 (z) = 1 1 ϕ1 , t zq − zq and for all p ∈ {1, ..., s}, one has: ∂zp ∂zq ϕ1 (z) = 1 −1 ϕ1 if p 6= q 2 t (zp − zp )(zq − zq ) 1 −1 (1 + t) ϕ1 if p = q. 2 t (zp − zp )2 1 ϕ1 B where the matrix B is t2 (1+t) 1 1 · · · 2 4y1 y2 4y1 ys 4y 11 (1+t) 4y2 y1 4y2 · · · 4y12 ys , 2 B= .. . (1+t) 1 1 · · · 4y2 4ys y1 4ys y2 Hence (∂zp ∂zq ϕ1 ) = s and zj = xj + iyj for all j ∈ {1, ..., s + t}. Notice that B is the sum between a diagonal positive definite matrix and a positive semidefinite one, hence B is positive definite. Now, let ω0 := i∂∂ϕ and for all g = (u, a) ∈ A n OK set ρ(g) := |σs+1 (u)|2 . First, notice that because u is a unit we have (σ1 (u)...σs (u))(|σs+1 (u)|2 ...|σs+t (u)|2 ) = 1. Then, write ∂∂(ϕ ◦ g)(z) = 1 (σ1 (u)...σs (u)) 1 t ∂∂ϕ1 (z) + ∂∂ t X k=1 = ρ(g)∂∂ϕ1 (z) + ρ(g)∂∂ t X k=1 = ρ(g)∂∂ϕ(z) 15 |zs+k |2 |σs+k (u)zs+k + σs+k (a)|2 We now obtain the following equalities: g ∗ ω0 = g ∗ (i∂∂ϕ) = i∂∂(ϕ ◦ g) = iρ(g)∂∂(ϕ) = ρ(g)ω0 . This concludes the proof. Theorem 3.13. If a number field K of degree d = s + 2t over Q with s real and 2t complex embeddings, where t ≥ 2, contains a nonreciprocal unit u ∈ UK of degree d whose 2t algebraic conjugates satisfy (1) then for some integers m ≥ 0 and q ≥ 2 we have s = (2t + 2m)q − 2t. (2) On the other hand, if s and t ≥ 2 satisfy (2) then there is a number field K with s real and 2t complex embeddings that contains a nonreciprocal unit u ∈ UK of degree d satisfying (1). Theorem 3.14. Let K be a number field of degree d = s + 2t over Q with s real and 2t complex embeddings, where t ≥ 2 and s is not of the form (2). Then the rank of the subgroup U of UK of units satisfying (1) is smaller than s and, therefore, for any choice of an admissible subgroup U for K the Oeljeklaus-Toma manifold X(K, U ) has no LcK metric. Proof : Recall that no OT-manifold admits a K¨ahler structure (this is Proposition 2.5, loc. cit.). Assume that K is a number field of degree d = s + 2t with t > 2 and with s not being of the form (2). We now suppose that the rank of the subgroup U of UK of units satisfying equation (1) is at least (therefore, equal to) s and we want to show that it leads to a contradiction. First, notice that l(U ) has a trivial intersection with the kernel of the projecting map P : S → Rs where l and P are defined in the introduction of the main article. Thus, U is an admissible subgroup of UK . Now, consider the OTmanifold X(K, U ); it admits an LcK metric by Theorem 3.12. Call u1 , ..., us a family multiplicatively independent units generating U and K 0 := Q(u1 , ..., us ). The same argument as the main article shows that K 0 can not be a proper subfield of K. As a consequence of Lemma 3.5 and Theorem 3.13, every element u of U satisfies |σs+j (u)| = 1, for all j ∈ {1, ..., t}. Let ω be a K¨ahler form on Hs × Ct giving rise to an LcK metric on X(K, U ). For all g = (u, a) ∈ U n OK , one has g ∗ ω = |σs+1 (u)|2 ω (see the paragraph before Theorem 3.12), which simplifies as g ∗ ω = ω. The form ω being invariant 16 under the action of A n OK , it descends to a K¨ahler form on X(K, U ). This implies that X(K, U ) is a K¨ahler manifold, which is the desired contradiction. 17 References [1] L. Battisti and K. Oeljeklaus, Holomorphic line bundles over domains in Cousin groups and the algebraic dimension of OT-manifolds, preprint at arXiv:1306.3047v1, 2013, 13 p. [2] Borevich, A. I. and Shafarevich, I. R.,Number theory, Translated from the Russian by Newcomb Greenleaf. Pure and Applied Mathematics, Vol. 20, Academic Press, 1966, [3] S. Eilenberg, and S. MacLane,Relations between homology and homotopy groups of spaces,Ann. of Math. (2) 46,(1945), 480–509. [4] K. Kodaira, Complex manifolds and deformation of complex structures, Translated from the 1981 Japanese original by K. Akao, Springer-Verlag, 2005, x+465 [5] K. Oeljeklaus and M. Toma, Non-K¨ahler compact complex manifolds associated to number fields, Ann. Inst. Fourier 55 (2005), 161–171. [6] Chih Han Sah, Cohomology of split group extensions, J. Algebra 29 (1974), 255–302. [7] C. Vogt, Line bundles on toroidal groups, J. Reine Angew. Math. 335 (1982), 197–215. 18