Complexified Symplectomorphisms and Geometric Quantization
Miguel Barbosa Pereira
Thesis to obtain the Master of Science Degree in Master in Mathematics and Applications
Supervisor(s): Prof. João Luís Pimentel Nunes
Examination Committee Chairperson: Prof. Miguel Tribolet de Abreu Supervisor: Prof. João Pimentel Nunes Member of the Committee: Prof. José Manuel Vergueiro Monteiro Cidade Mourão Member of the Committee: Prof. Sílvia Nogueira da Rocha Ravasco dos Anjos
November 2017 ii This thesis is dedicated to my parents and to my brother.
iii iv Acknowledgments
I would like to thank my advisor, Prof. Joao˜ Nunes, for all the time he took helping me, from my first semester as a mathematics student in Riemannian geometry, up to now in my thesis. He was always available for help and was always extremely patient. I also want to express my gratitude to all the teachers of the courses I took during these last two years, for all the knowledge they gave me in class, and for helping me when I had questions about the material outside of the class. This thesis started as a project for the Centre for Mathematical Analysis, Geometry, and Dynamical Systems, that awarded me a scholarship one year and a half ago. So I would like to thank CAMGSD for this opportunity to learn and do math in a way that is closer to the professional level. I also want to thank the members of the Examination Committee, Prof. Miguel Abreu, Prof. Jose´ Mourao,˜ and Prof. S´ılvia Anjos for their time reading and evaluating my thesis, and for their helpful corrections. During the time I was in Lisbon, I had the opportunity to meet new people and spend time with many excellent friends, who enriched my life. They would be too many to list, but I want to thank my friends from aerospace engineering, especially Tiago Silva, Manuel Barjona and Telmo Pires who encouraged me to become a better student, my neighbors from B1 for making it the best floor in RDP, as I do want to thank my friends from my class Z in Braga for the past 9 years. A special mention goes to Andre´ Pereira, for helping me in many occasions. Coming to study to a different city was not an easy experience for me, and I couldn’t have done it without the continuous support of my parents. When I decided to change from engineering to maths, they gave me complete freedom to do so. They have always been there for me. A big thank you, Mom and Dad. I want to thank my brother Pedro for all the memes and for putting up with my explanations of the math I was learning. Pedro, every aspiring physicist should know what a cotangent bundle is. Thank you to all the other members of my family, uncles, and cousins for caring for me. A special thank you to my grandmother Maria, for all the meals she cooked for me, but also for being a caring, strong and inspiring person. Finally, if you are reading this, I want to thank you, and I hope that the text is clear, organized and I hope that it is helpful to your studies.
v vi Resumo
Nesta tese estudamos geometria complexa, simplectica´ e Kahler.¨ O cap´ıtulo1 e´ introdutorio´ e apresentamos os preliminares teoricos´ relevantes para o resto da tese. No cap´ıtulo2 apresentamos um resumo de [14]. Partindo de uma variedade K ahler,¨ atuamos na sua estrutura complexa via pullback pelo fluxo de um campo vetorial Hamiltoniano. Provamos que as metricas´ Kahler¨ resultantes formam uma curva numa variedade Riemanniana (de dimensao˜ infinita) cujos pontos sao˜ potenciais Kahler,¨ e que esta curva e´ na realidade uma geodesica´ relativamente a` metrica´ (que e´ a metrica´ de Mabuchi). No cap´ıtulo3 estudamos estruturas K ahler¨ no fibrado cotangente de um grupo de Lie G. Explicamos as estruturas Kahler¨ conhecidas de Hall e Kirwin e de Kirwin, Mourao˜ e Nunes, e relacionamos as duas. Em [12] os autores provam que e´ poss´ıvel definir estruturas Kahler¨ em T ∗G usando func¸oes˜ complexificadoras Ad-invariantes. Provamos que a condic¸ao˜ de Ad-invarianciaˆ pode ser substitu´ıda por um conjunto de equac¸oes˜ mais geral. No cap´ıtulo5, damos um exemplo de quantizac¸ ao˜ geometrica´ de T ∗G. Introduzimos um fibrado de ∗ linha L em T G e uma polarizac¸ao˜ Pτ definida por fazer um pullback da polarizac¸ao˜ vertical pelo fluxo complexificado de um campo vetorial Hamiltoniano. A partir destas estruturas, definimos um espac¸o de
Hilbert cujos elementos sao˜ as secc¸oes˜ integraveis´ e Pτ -polarizadas de L.
Palavras-chave: geometria Kahler,¨ quantizac¸ao˜ geometrica,´ metrica´ de Mabuchi, grupo de Lie, fibrado cotangente, difeomorfismos complexificados
vii viii Abstract
In this thesis, we study complex, symplectic and Kahler¨ geometry. Chapter1 is introductory and presents the relevant theoretical background for the rest of the thesis. In chapter2 we present an overview of [14]. Starting with a K ahler¨ manifold, we act on its complex structure by pulling back along the complexified flow of an Hamiltonian vector field. We prove that the resulting Kahler¨ metrics form a curve in an (infinite dimensional) Riemannian manifold whose points are Kahler¨ potentials and that this curve is actually a geodesic with respect to the metric (which is the Mabuchi metric). In chapter3 we study K ahler¨ structures on the cotangent bundle of a Lie group G. We explain the known Kahler¨ structures of Hall and Kirwin and of Kirwin, Mourao,˜ and Nunes and relate the two. In [12], the authors prove that one can define Kahler¨ structures on T ∗G using Ad-invariant complexifier functions. We prove that the condition of Ad-invariance can be replaced by a more general set of equations. In chapter5, we give an example of geometric quantization of T ∗G. We introduce a line bundle L on ∗ T G and a polarization Pτ defined by pulling back the vertical polarization along the complexified flow of an Hamiltonian vector field. From these structures, we define a Hilbert space whose elements are integrable Pτ -polarized sections of L.
Keywords: Kahler¨ geometry, geometric quantization, Mabuchi metric, Lie group, cotangent bundle, complexified diffeomorphisms
ix x Contents
Acknowledgments...... v Resumo...... vii Abstract...... ix List of Tables...... xiii Nomenclature...... xv Glossary...... 1
1 Theoretical Background1 1.1 Symplectic Manifolds...... 1 1.2 Moser’s Theorem...... 3 1.3 Complex Geometry...... 3 1.3.1 Almost complex manifolds. Complex manifolds...... 3 1.3.2 Splittings and forms on complex manifolds...... 5 1.3.3 When is an almost complex manifold a complex manifold?...... 6 1.3.4 Compatible structures. Kahler¨ manifolds...... 7 1.3.5 Cohomology results on Kahler¨ manifolds...... 8 1.4 Lie groups and Lie algebras...... 11 1.4.1 Representations...... 11 1.4.2 Adjoint representations...... 12 1.5 Structure functions and structure constants...... 13
2 Geodesics on the Space of Kahler¨ Metrics of a Manifold 15 2.1 Lie Series and complexified flows...... 15 2.2 Action of a complexified analytic flow on a complex structure...... 18 2.3 Restriction to Hamiltonian flows...... 19 2.4 Action on Kahler¨ structures...... 19 2.5 Geodesics on the Space of Kahler¨ metrics...... 24 2.6 Symplectic picture and complex picture...... 25 2.7 The path of Kahler¨ metrics (generated by a complex flow of an Hamiltonian vector field) is a geodesic...... 26
xi 3 Kahler¨ structures on the Cotangent Bundle of a Lie Group 31 3.1 T ∗G is diffeomorphic to G × g ...... 31 3.2 Kahler¨ structures on T ∗G ...... 32 3.2.1 Standard Kahler¨ structure on T ∗G ...... 32 3.2.2 The Kahler¨ structure of Hall and Kirwin...... 33 3.2.3 The Kahler¨ structure of Kirwin, Mourao˜ and Nunes...... 34 3.2.4 Problems to study...... 35 3.3 Useful coordinates based on the diffeomorphism T ∗G =∼ G × g ...... 35 3.4 Relation between theorems 3.2.2 and 3.2.4...... 41 3.5 The theorem of Hall and Kirwin in the case of a Lie group...... 43 3.5.1 Attempt at proving conjecture 3.5.1...... 43
3.5.2 ωβ as a pullback of ωST by a diffeomorphism...... 45 3.6 The theorem of Kirwin, Mourao˜ and Nunes in the case of non Ad-invariant h ...... 47 τ 3.6.1 ϕh is a diffeomorphism. The complex structure J ...... 47 ∗ 3.6.2 A basis for T1,0(T G) ...... 48 3.6.3 Equations for T ∗G being Kahler¨ in terms of h ...... 49
4 A Short Overview of Geometric Quantization 53 4.1 Introduction and motivation...... 53 4.2 General set up...... 54
5 An Example of Quantization of the Cotangent Bundle of a Lie Group 55 5.1 General setup...... 55 5.2 A prequantum line bundle for T ∗G ...... 56 5.3 A polarization for T ∗G ...... 57 5.4 T ∗G is foliated by Kahler¨ manifolds...... 63
5.5 Computation of the P¯τ -polarized sections...... 67
5.6 The inner product of Hτ . An unitary isomorphism of Hilbert spaces...... 69
Bibliography 75
A Moser’s Theorem 77 A.1 Lie derivative of time dependent vector field in terms of an isotopy...... 77 A.2 Proof of Moser’s Theorem...... 78
xii List of Tables
4.1 Comparison between classical and quantum mechanics...... 53
xiii xiv Nomenclature
H Hilbert space
N Nijenhuis tensor
P Polarization g Lie algebra
Greek symbols
α Differential form
∆ Laplacian
κ Kahler¨ potential
∇ Connection or gradient (depending on context)
ω Symplectic form
t φX Flow of the vector field X
τ Complex time
θ Symplectic potential
Roman symbols
Ad Adjoint representation of Lie group ad Adjoint representation of Lie algebra f Function on a manifold
G Lie group g or h·, ·i Riemannian metric
J Complex structure
M Manifold
R Curvature of a connection
xv s Section of bundle
U Open set
X Vector field
Subscripts
LX Lie derivative with respect to the vector field X
Superscripts
α# associated vector to the covector α
X[ associated covector to the vector X
xvi Chapter 1
Theoretical Background
1.1 Symplectic Manifolds
Definition 1.1.1. Let M be a manifold. A symplectic form on M is a 2-form ω ∈ Ω2(M) that is closed and nondegenerate. A symplectic manifold is a pair (M, ω), where M is a manifold, and ω ∈ Ω2(M) is a symplectic form.
Definition 1.1.2. A symplectomorphism of two symplectic manifolds (M, ωM ) and (N, ωN ) is a diffeo- ∗ morphism ϕ : M → N such that ϕ ωN = ωM .
Theorem 1.1.3 (Darboux). Let (M, ω) be a symplectic manifold. For all p ∈ M, there exists a coordinate neighborhood of p, (U, x1, ...xn, y1, ..., yn), such that on U
n X ω = dxi ∧ dyi. (1.1) i=1
In particular, the dimension of M is even.
Proof. See [6], page 233.
Remark 1.1.4 (Comparison between Riemannian and Symplectic Geometries).
• Riemannian Geometry: there exist local coordinates centered at p such that g is the Euclidean inner product at p.
• Symplectic Geometry: there exist local coordinates centered at p such that ω is the canonical symplectic form in a neighborhood of p.
• The analogous to Darboux’s Theorem in Riemannian Geometry (i.e. g being the Euclidean inner product in a neighborhood of p) is false. If it were true, the curvature would always be zero.
• Darboux’s Theorem implies that there is no analogue of curvature in symplectic geometry. Since all symplectic manifolds of the same dimension are locally symplectomorphic, there cannot exist any local invariant.
1 Proposition 1.1.5. Let (M, ω) be a symplectic manifold of dimension 2n. Then M is orientable, with volume form ωn = ω ∧ ... ∧ ω. | {z } n times
Proof. We will prove that ∀p ∈ M, (ω ∧ ... ∧ ω)p 6= 0. Let p ∈ M. By Darboux’s Theorem, there exist coordinates around p such that n X ω = dxi ∧ dyi. i=1 Then, ω ∧ ... ∧ ω = dx1 ∧ dy1 ∧ ... ∧ dxn ∧ dyn 6= 0 at p.
Example 1.1.6. Let M be an n-dimensional manifold. We will give the cotangent bundle of M, T ∗M, 1 n 1 n the structure of a symplectic manifold. If (U, x , ..., x ) is a coordinate chart on M, then (dx )p, ..., (dx )p ∗ ∗ Pn i form a basis of Tp M. Then, any element α of Tp M can be given as α = i=1 pi(dx )p. This defines a ∗ ∗ 1 n 1 n local coordinate chart on T M, (T U, x , ..., x , p1, ..., pn), where the (x , ..., x ) are the coordinates of 1 n p = π(α) in M, and (p1, ..., pn) are the coefficients of α in the basis dx , ..., dx . Define θ ∈ Ω1(T ∗M) pointwise by:
∗ θα : Tα(T M) −→ R (1.2) v 7−→ α(dπ · v).
Pn i ∂ Pn ∂ θ is the canonical symplectic potential. If v = a i + bi , then i=1 ∂x i=1 ∂pi
n n X ∂ X ∂ θ (v) = α(dπ · v) = α dπ aj + b α ∂xj j ∂p j=1 j=1 j
n n n X ∂ X X ∂ = α aj = p (dxi) aj ∂xj i p ∂xj j=1 i=1 j=1
| {z∗ } ∈ Tp M n n n n ! X X X ∂ X ∂ = p ai = p (dxi) aj + b . i i α ∂xj j ∂p i=1 i=1 j=1 j=1 j
| {z ∗ } | {z } ∈ Tα(T M) v
Pn i Therefore, in local coordinates θ can be written as θ = i=1 pidx . Define the canonical symplectic form ω ∈ Ω2(T ∗M) by ω = −dθ. The local expression for ω is then
n ! n X i X i ω = −d pidx = dx ∧ dpi. i=1 i=1
ω is closed and nondegenerate. Therefore (T ∗M, ω) is a symplectic manifold.
Definition 1.1.7. Let h ∈ C∞(M). The Hamiltonian vector field defined by h is the unique vector field
Xh satisfying ω(Xh, ·) = dh.
∞ Definition 1.1.8. Let f, g ∈ C (M). Their Poisson bracket, denoted {f, g}, is {f, g} = Xf · g.
2 ∞ Remark 1.1.9. C (M) equipped with the Poisson bracket is a Lie algebra. Also, the map f 7→ Xf is a Lie algebra homomorphism.
1.2 Moser’s Theorem
Let M be a manifold.
Definition 1.2.1. An isotopy is a map ρ: M × R −→ M such that:
•∀ t ∈ R ρt : M −→ M defined by ρt(p) = ρ(p, t) is a diffeomorphism.
• ρ0 = idM
Definition 1.2.2. A time dependent vector field on M is an assignment R 3 t 7→ Xt ∈ X(M) that is smooth in t.
Remark 1.2.3. We can think of an isotopy as being the flow of a time dependent vector field:
• An isotopy defines a time dependent vector field, by the formula
dρ t = X ◦ ρ . (1.3) dt t t
d −1 In other words, Xt is given at a point p by (Xt)p = ds ρs ρt (p) . s=t
• Conversely, if Xt is a time dependent vector field, and if M is compact (or Xt has compact support),
then Xt can be integrated to obtain an isotopy.
The next theorem, by Moser, provides an answer to the question: If M is a manifold and ω0, ω1 are symplectic forms, does there exist a symplectomorphism ϕ :(M, ω0) −→ (M, ω1)?
2 Theorem 1.2.4 (Moser). Let M be a compact manifold, ω0, ω1 ∈ Ω (M) be symplectic forms, such that
2 (i) [ω0] = [ω1] ∈ H (M, R);
(ii) ∀t ∈ [0, 1] ωt := (1 − t)ω0 + tω1 is symplectic.
∗ Then there exists an isotopy ρt : M −→ M such that ρt ωt = ω0 ∀t ∈ [0, 1]
Proof. See appendixA.
1.3 Complex Geometry
1.3.1 Almost complex manifolds. Complex manifolds
Definition 1.3.1. An almost complex structure on a manifold M is a vector bundle isomorphism 2 J : TM −→ TM such that J = −idTM . An almost complex manifold is a pair (M,J) where M is a manifold and J is an almost complex structure on M.
3 Definition 1.3.2. A complex manifold of complex dimension n is given by the data:
• M a topological space (which is Hausdorff and 2nd countable),
n •A = {ϕα : Uα ⊂ M −→ Vα ⊂ C }α∈I a complex atlas, satisfying:
• M = ∪α∈I Uα ;
•∀ α ∈ I, ϕα is an homeomorphism;
•∀ α, β ∈ I such that Uα ∩ Uβ 6= ∅,
−1 n n ϕβ ◦ ϕα : ϕα(Uα ∩ Uβ) ⊂ C −→ ϕβ(Uα ∩ Uβ) ⊂ C
−1 is biholomorphic (this means that both ϕβ ◦ ϕα and its inverse are holomorphic);
n −1 •A is maximal: if ψ : U ⊂ M −→ V ⊂ C is an homeomorphism such that ∀α ∈ I ψ ◦ ϕα is biholomorphic, then ψ ∈ A.
Theorem 1.3.3. Any complex manifold admits an almost complex structure.
Proof. Let M be a complex manifold. We start by giving a local definition of J. Let (U, z1, ..., zn) be a complex coordinate chart on M. If zj = xj + iyj, then for all p ∈ U,
∂ ∂ T M = span , : j = 1, ..., n . p R ∂xj ∂yj
Define
Jp : TpM −→ TpM ∂ ∂ 7−→ ∂xj ∂yj ∂ ∂ 7−→ − . ∂yj ∂xj
2 Then clearly Jp = −id. Now, we prove that J is a globally well defined object. Let (U, z1, ..., zn), (V, w1, ..., wn) be coordinate j j j j neighborhoods such that U ∩ V 6= ∅. We will show that J|U = J|V on U ∩ V . If z = x + iy and w = uj + ivj, since the transition functions are biholomorphic, they satisfy the Cauchy-Riemann equations:
∂uk ∂vk = j j ∂x ∂y (1.4) ∂uk ∂vk = − . ∂yj ∂xj
4 We will now show that J|U = J|V .
n ! ∂ X ∂uk ∂ ∂vk ∂ J| = J| + V ∂xj V ∂xj ∂uk ∂xj ∂vk k=1 n X ∂uk ∂ ∂vk ∂ = − ∂xj ∂vk ∂xj ∂uk k=1 n X ∂vk ∂ ∂uk ∂ ∂ = + = ∂yj ∂vk ∂yj ∂uk ∂yj k=1
∂ ∂ where we used the Cauchy-Riemann equations. Analogously, J|V ∂yj = − ∂xj . Therefore J|U = J|V , and J is well defined globally.
1.3.2 Splittings and forms on complex manifolds
Definition 1.3.4. Let (M,J) be an almost complex manifold. Consider the complexified tangent bun- dle of M, TM ⊗ C. In TM ⊗ C, J has ±i as eigenvalues. Define the following spaces:
T1,0M := (+i) − eigenspace of J = holomorphic tangent vectors, (1.5a)
T0,1M := (−i) − eigenspace of J = anti-holomorphic tangent vectors, (1.5b)
1,0 ∗ T M := (T1,0M) = complex-linear cotangent vectors, (1.5c)
0,1 ∗ T M := (T0,1M) = complex-antilinear cotangent vectors. (1.5d)
Also, define k M l,m Ω (M; C) := Ω (M) = k-forms on M with complex values, (1.6) l+m=k where l ! m ! ^ ^ Ωl,m(M) := sections of the vector bundle T 1,0M ∧ T 0,1M . (1.7)
Consider the projections πl,m :Ωk(M; C) −→ Ωl,m(M). Define ∂,∂¯ by:
∂ := πl+1,m ◦ d:Ωl,m(M) −→ Ωl+1,m(M), (1.8a)
∂¯ := πl,m+1 ◦ d:Ωl,m(M) −→ Ωl,m+1(M). (1.8b)
If M is a complex manifold with almost complex structure J, then this splitting can be written explicitly in terms of local coordinates. Recall that:
∂ ∂ T M = span , : j = 1, ..., n , p R ∂xj ∂yj ∂ ∂ T M ⊗ = span , : j = 1, ..., n , p C C ∂xj ∂yj
∂ ∂ ∂ ∂ J = and J = − . ∂xj ∂yj ∂yj ∂xj
5 Using these equations, the spaces in definition 1.3.4 become:
( ) 1 ∂ i ∂ T M := (+i) − eigenspace of J = span − , (1.9a) 1,0 C 2 ∂xj 2 ∂yj | {z } j=1,...,n := ∂ ∂zj ( ) 1 ∂ i ∂ T M := (−i) − eigenspace of J = span + , (1.9b) 0,1 C 2 ∂xj 2 ∂yj | {z } j=1,...,n := ∂ ∂z¯j 1,0 ∗ j j T M := (T1,0M) = span {dx + idy }j=1,...,n, (1.9c) C | {z } =dzj 0,1 ∗ j j T M := (T0,1M) = span {dx − idy }j=1,...,n. (1.9d) C | {z } =dz¯j
And if (U, z1, ..., zn) is a complex coordinate neighborhood,
l,m X J K ∞ Ω (U) = bJ,K dz ∧ dz¯ : bJ,K ∈ C (U; C) . (1.10) |J|=l,|K|=m
1.3.3 When is an almost complex manifold a complex manifold?
We saw that any complex manifold is an almost complex manifold. Question: Is the converse true? More precisely, given (M,J) an almost complex manifold, does there exist a complex structure A such that A defines J? Answer: Provided by a theorem of Newlander-Nirenberg.
Definition 1.3.5. Let (M,J) be an almost complex manifold. Its Nijenhuis tensor N is a tensor of type (2, 1) given by N (X,Y ) := [JX,JY ] − J[X,JY ] − J[JX,Y ] − [X,Y ].
Definition 1.3.6. Let (M,J) be an almost complex manifold. J is said to be integrable if M admits a complex structure A such that the almost complex structure defined by A is J (in the sense of theorem 1.3.3).
Theorem 1.3.7. (Newlander-Nirenberg) Let (M,J) be an almost complex manifold with Nijenhuis tensor N . Then, the following are equivalent:
(i) J is integrable;
(ii) N = 0;
(iii) d = ∂ + ∂¯;
(iv) ∂¯2 = 0;
2,0 (v) π ◦ d|Ω0,1 = 0.
Proof. See theorem 5.7.4. in [11], and theorem 7.4. and proposition 8.2. in [13].
6 1.3.4 Compatible structures. Kahler¨ manifolds
Proposition 1.3.8. Let M be a manifold. Let ω ∈ Ω2(M) be a symplectic form on M, g be a Riemannian metric on M, and J be an almost complex structure on M. Then, the following are equivalent:
(i) g(·, ·) = ω(·,J·);
(ii) ω(·, ·) = g(J·, ·);
(iii) J(·) =g ˜−1(˜ω(·)).
(Where g˜, ω˜ are the linear isomorphisms given by g˜(u) = g(u, ·) and ω˜ = ω(u, ·)).
Proof. The result follows from computation, and from using the properties of each of the three structures 2 (in particular using that J = −idTM ).
Definition 1.3.9. If the conditions of proposition 1.3.8 are true, (ω, J, g) is called a compatible triple. If (M, ω) is a symplectic manifold, and J is an almost complex structure on M, then J is compatible (with ω) if g(·, ·) := ω(·,J·) is a Riemannian metric.
AKahler¨ manifold is at the same time a symplectic manifold, a complex manifold and a Riemannian manifold, in a way that the three structures are compatible:
Definition 1.3.10. A Kahler¨ manifold is a Manifold M equipped with a symplectic form ω, a complex structure J, and a Riemannian metric g such that (ω, J, g) is a compatible triple. If (M, ω, J, g) is a Kahler¨ manifold, we say that ω is a Kahler¨ form.
Theorem 1.3.11 (A. C. da Silva, [4]). Given a complex manifold (M,J), a form ω ∈ Ω2(M) is Kahler¨ if and only if ∂ω = 0, ∂ω¯ = 0, and if it is given locally by
n i X ω = h dzj ∧ dz¯k, (1.11) 2 jk j,k=1 where ∀p (hjk(p)) is a positive definite hermitian matrix.
Proof. See [4], pages 90-92.
Definition 1.3.12. Let (M, ω, J, g) be a Kahler¨ manifold, and U ⊂ M be open.
• κ ∈ C∞(M; R) is a (global) Kahler¨ potential if ω = i∂∂κ¯ ;
∞ ¯ • κ ∈ C (U; R) is a local Kahler¨ potential if ω|U = i∂∂κ.
∞ 2 2 Lemma 1.3.13. Let (M, ω, J, g) be a Kahler¨ manifold. Let h ∈ C (M). Then, kXhk = k∇hk .
Proof.
h∇h, ∇hi = dh(∇h) = ω(Xh, ∇h) = hJXh, ∇hi
= dh(JXh) = ω(Xh,JXh) = hXh,Xhi
7 1.3.5 Cohomology results on Kahler¨ manifolds
Definition 1.3.14. Let M be a compact Kahler¨ manifold. We define:
• The Hodge-∗ operator: if {ω1, ..., ω2n} is a positively oriented orthonormal coframe, then ∗(ω1 ∧ ... ∧ ωk) = ωk+1 ∧ ... ∧ ω2n.
• An inner product of forms:
k k h·, ·i:Ω (M) × Ω (M) −→ R (1.12) Z (α, β) 7−→ α ∧ ∗β. M
• Adjoints with respect to the inner product: the adjoint of d, d∗, is uniquely specified by the property hdα, βi = hα, d∗βi. Analogously, we define ∂∗ and ∂¯∗.
• The Laplacian, ∆: Ωk(M) −→ Ωk(M), is given by ∆ = dd∗ + d∗d.
Proposition 1.3.15. ∆ = dd∗ + d∗d = 2(∂¯∗∂¯ + ∂¯∂¯∗).
Proof. See [13], page 103.
Proposition 1.3.16. The following are equivalent:
(i) ∆ω = 0;
(ii) dω = 0 and d∗ω = 0;
(iii) ∂ω¯ = 0 and ∂¯∗ω = 0.
Proof. We prove only the first equivalence. The proof of the second one is analogous. (⇐=): If dω = 0 and d∗ω = 0 then ∆ω = 0 by definition. (=⇒): Assume that ∆ω = 0. Then:
0 = h∆ω, ωi = hdd∗ω + d∗dω, ωi
= hdd∗ω, ωi + hd∗dω, ωi = hd∗ω, d∗ωi + hdω, dωi
= kd∗ωk2 + kdωk2.
Therefore, both d∗ω and dω are zero.
Definition 1.3.17. A form is harmonic if its Laplacian is zero. The set of harmonic forms of a given type is denoted Hl,m(M): Hl,m(M) := {α ∈ Ωl,m(M) : ∆α = 0}. (1.13)
Theorem 1.3.18 (Hodge-Dolbeaut decomposition). Let M be a compact Kahler¨ manifold. Then, Ωl,m(M) decomposes as a direct sum of the following subspaces:
Ωl,m(M) = Hl,m(M) ⊕ ∂¯Ωl,m−1(M) ⊕ ∂¯∗Ωl,m+1(M). (1.14)
8 This decomposition is orthogonal with respect to h·, ·i.
Proof. See [8], pages 84-100.
Lemma 1.3.19 (∂¯-Lemma). Let M be a complex manifold. Let ω ∈ Ω0,1(M) be such that ∂ω¯ = 0. Then, ω is locally ∂¯ exact. More preciselly, for all p ∈ M there exists a neighborhood V of p and a complex ∞ ¯ valued function on V , φ ∈ C (V ; C), such that ω|V = ∂φ.
Proof. See [8], pages 25-27.
Lemma 1.3.20 (Global i∂∂¯-Lemma). Let M be a compact Kahler¨ manifold. Let ω be an exact, real, type (1, 1) form. Then, there exists a φ ∈ C∞(M) such that ω = i∂∂φ¯ .
Proof. ω = dα, for some α ∈ Ω1(M). Decomposing α in its (1, 0), and (0, 1) parts, ω = ∂α1,0 + ∂α¯ 1,0 + ∂α0,1 + ∂α¯ 0,1. Since ω is of type (1, 1), ∂α1,0 and ∂α¯ 0,1 are zero. Applying the Hodge-Dolbeaut decomposition theorem to α0,1: α0,1 = γ + ∂η¯ + ∂¯∗ξ.
Where γ ∈ Ω1,0(M) is harmonic, η ∈ C∞(M; C) and ξ ∈ Ω2,0(M). We now prove that ∂¯∗ξ = 0.
0 = ∂α¯ 0,1 = ∂γ¯ + ∂¯∂η¯ +∂¯∂¯∗ξ, |{z} |{z} =0 =0
0 = h∂α¯ 0,1, ξi = h∂¯∂¯∗ξ, ξi = h∂¯∗ξ, ∂¯∗ξi =⇒ ∂¯∗ξ = 0.
So, α0,1 = γ + ∂η¯ , and therefore α = γ + ∂η¯ +γ ¯ + ∂η,¯
ω = dα = dγ +d∂η¯ + dγ¯ +d∂η¯ |{z} |{z} =0 =0 = ∂∂η¯ + ∂∂¯ η¯ = ∂∂η¯ − ∂∂¯η¯ = i∂∂¯(2 Im η).
Lemma 1.3.21 (Existence of local Kahler¨ potentials). Let M be a complex manifold. Let ω be a closed, real, type (1, 1) form. Then, ω is locally i∂∂¯-exact. More precisely, for all p ∈ M there exists a neighbor- ∞ ¯ hood V of p and a function on V φ ∈ C (V ) such that ω|V = i∂∂φ.
Proof. ω is closed, so it can be given locally as ω = dα. α = α1,0 + α0,1, where α1,0 = α0,1 because α is real. Since dα is of type (1, 1), both ∂α1,0 and ∂α¯ 0,1 are zero. The ∂¯-Lemma applied to α0,1 implies that locally, α0,1 = ∂ψ¯ (hence α1,0 = ∂ψ¯). Then, 2 Im ψ is a local Kahler¨ potential:
ω = dα = ∂α¯ 1,0 + ∂α0,1
= ∂∂¯ ψ¯ + ∂∂ψ¯ = −∂∂¯ψ¯ + ∂∂ψ¯
= ∂∂¯(ψ − ψ¯) = i∂∂¯(2 Im ψ).
9 Lemma 1.3.22. Let M be a compact connected Kahler¨ manifold, and φ ∈ C∞(M) such that ∂∂φ¯ = 0. Then φ is constant.
Proof. Recall the following definitions, present in [13], page 101:
L:Ωp,q(M) −→ Ωp+1,q+1(M) (1.15)
α 7−→ ω ∧ α,
Λ = L∗ :Ωp,q(M) −→ Ωp−1,q−1(M). (1.16)
Then, it is a fact, also stated in [13], page 103, that
i∂¯∗ = Λ∂ − ∂Λ. (1.17)
With this information, we can conclude that ∆φ = 0:
∆φ = 2(∂¯∗∂¯ + ∂¯∂¯∗)φ (proposition 1.3.15)
= 2∂¯∗∂φ¯ (φ is of type 0, 0)
= 2i(∂Λ∂¯ − Λ∂∂¯)φ (equation (1.17))
= −2iΛ∂∂φ¯ (φ is of type 0, 0)
= 0.
Using proposition 1.3.16, we conclude that dφ = 0, and since M is connected, φ is constant.
Lemma 1.3.23. Let M be a complex manifold, and φ ∈ C∞(M) such that ∂∂φ¯ = 0. Then φ is locally of the form φ = f + f¯, where f is holomorphic and f¯ is anti-holomorphic.
¯ ¯ ¯ Proof. Since d(∂φ) = ∂∂φ = 0, ∂φ is locally exact. Therefore, we can cover M by open sets {Ui} such ∞ ¯ that for each i there exists a gi ∈ C (Ui; C) such that ∂φ|Ui = dgi. Considering the type of the forms, ¯ we conclude that ∂gi = 0, so gi is anti-holomorphic. Let hi = φ|Ui − gi. Then
¯ ¯ ¯ ∂hi = ∂(φ|Ui − gi) = ∂φ|Ui − dgi = 0,
which implies that hi is holomorphic. Since φ|Ui is real,
¯ ¯ gi + hi = φ|Ui = φ|Ui =g ¯i + hi.
Therefore, ¯ gi − hi =g ¯i − hi.
The left hand side is a anti-holomorphic function. The right hand side is a holomorphic function. ¯ Then, both are holomorphic and anti-holomorphic, and therefore both are locally constant: gi − hi =
ci ¯ g¯i − hi = ci, where ci is a real locally constant function. Define fi = hi + 2 . Then, fi is holomorphic, fi
10 is anti-holomorphic, and
c c f + f¯ = h + i + h + i = h + h¯ + c = h + g = φ| . i i i 2 i 2 i i i i i Ui
1.4 Lie groups and Lie algebras
1.4.1 Representations
Definition 1.4.1. Let V be a finite dimensional vector space. Then
GL(V ) := {T : V −→ V | T is a linear isomorphism} (1.18) is a Lie group under composition, called the general linear group of V . Its Lie algebra is the general linear Lie algebra of V :
gl(V ) = {T : V −→ T | T is linear} = End(V ), (1.19) where the Lie bracket is the commutator as linear operators:
[T,S] = TS − ST.
Definition 1.4.2. A representation of a Lie group G on a vector space V is a Lie group homomor- phism λ: G −→ GL(V ).A representation of a Lie algebra g on a vector space V is a Lie algebra homomorphism λ: g −→ gl(V ).
Definition 1.4.3. A unitary representation of a Lie group is a representation λ: G −→ GL(V ) such that the vector space V is a Hilbert space, and such that λ(g) is unitary ∀g ∈ G.
Definition 1.4.4. Let λ: G −→ GL(V ) be a representation of a Lie group. A subrepresentation of λ is a subspace W of V , such that ∀g ∈ G λ(g)(W ) ⊂ W . A representation λ of a Lie group on V is said to be irreducible if its only subrepresentations are {0} ⊂ V and V itself.
Definition 1.4.5. Let λ: G −→ GL(V ) and µ: G −→ GL(W ) be two unitary representations. λ and µ are equivalent unitary representations if there exists a Hilbert space isomorphism φ : V −→ W such that φ˜ ◦ λ = µ, where φ˜ is the following map:
φ˜: GL(V ) −→ GL(W ) (1.20)
T 7−→ φ ◦ T ◦ φ−1.
It is easy to see that this defines an equivalence relation over the set of unitary representations of G.
11 1.4.2 Adjoint representations
Definition 1.4.6. Let G be a Lie group. Define:
• the action of G on itself by conjugation:
Ψ: G −→ Aut(G) (1.21) Ψg : G −→ G g 7−→ ; h 7−→ ghg−1
• the adjoint representation of G:
Ad: G −→ GL(g) (1.22)
g 7−→ (Adg := deΨg : g −→ g);
• the adjoint representation of g:
ad: g −→ End(g) (1.23) adX : g −→ g X 7−→ . Y 7−→ [X,Y ]
The following proposition is a list of properties of the adjoint representations. In particular, it shows that Ad is a representation of G on g and that ad is a representation of g on g.
Proposition 1.4.7.
• Ψ is a Lie group homomorphism: Ψgh = Ψg ◦ Ψh;
−1 • Ψg is a Lie group isomorphism: Ψg(a)Ψg(b) = Ψg(ab) and (Ψg) = Ψg−1 ;
• Ad is a Lie group homomorphism: Adgh = AdgAdh;
−1 • Adg is a Lie algebra isomorphism: Adg[X,Y ] = [AdgX, Adg,Y ] and (Adg) = Adg−1 ;
• ad is a Lie algebra homomorphism: ad[X,Y ] = [adX , adY ];
• adX is a Lie algebra derivation: adX [Y,Z] = [adX Y,Z] + [Y, adX Z].
Proof. The proof of each equation follows from the definitions. The proofs of the statements about Ad use the properties of the derivative and that Ψg = LgRg−1 , which implies that Adg = dLgdRg−1 . The proof of the statements about ad make use of the Jacobi identity.
12 Remark 1.4.8. A Lie group homomorphism Φ: G −→ H induces an homomorphism on the Lie algebras
φ: g −→ h, given by φ = deΦ. Also, the following diagram commutes:
φ=d Φ g e h
exp exp
G Φ H
For a proof of these facts, see [9] page 60.
The next proposition shows that the homomorphism induced by Ad on the Lie algebras is ad.
Proposition 1.4.9. Let G be a Lie group with Lie algebra g. Then, ad = deAd. Therefore, the homomor- phism induced by Ad on the Lie algebras is ad.
Proof. Let X,Y ∈ g. We will show that deAd(X)(Y ) = ad(X)(Y ) = [X,Y ].
d deAd(X)(Y ) = deψexp(tX)Y dt t=0 d d = ψexp(tX) exp(sY ) dt t=0 ds s=0 d d = exp(tX) exp(sY ) exp(−tX) dt t=0 ds s=0 d d d d = exp(tX) exp(sY ) − exp(sY ) exp(tX) = V. dt t=0 ds s=0 dt t=0 ds s=0
We claim that V = [X,Y ]. To see this, let f ∈ C∞(G) and let X,Y ∈ X(G) be the left invariant vector fields defined by X and Y . Then
d d d d V · f = f exp(tX) exp(sY ) − f exp(sY ) exp(tX) dt t=0 ds s=0 dt t=0 ds s=0 d d d d = f exp(tX) exp(sY ) − f exp(sY ) exp(tX) dt t=0 ds s=0 ds s=0 dt t=0 d d = Y · f (exp(tX)) − X · f (exp(sY )) dt t=0 ds s=0 = X · (Y · f) (e) − Y · (X · f) (e)
= [X,Y ] · f.
1.5 Structure functions and structure constants
Definition 1.5.1. Let M be a differentiable manifold and {X1, ..., Xn} be a field of frames on an open k set of M. The structure functions of this frame, Cij, are defined by the equation:
n X k [Xi,Xj] = CijXk. (1.24) k=1
k k Note that since the Lie bracket is antisymmetric, so are the structure functions: Cij = −Cji. We now give some properties of these functions.
13 1 n Proposition 1.5.2. If {ω , ..., ω } is the field of coframes dual to {X1, ..., Xn}, then
n 1 X dωi + Ci ωj ∧ ωk = 0. (1.25) 2 jk j,k=1
Proof.
i i i i dω (Xl,Xm) = Xl(ω (Xm)) − Xm(ω (Xl)) − ω ([Xl,Xm])
n i i X j = −ω ([Xl,Xm]) = −ω ClmXj j=1 n 1 X = −Ci = − Ci ωj ∧ ωk(X ,X ). lm 2 jk l m j,k=1
Proposition 1.5.3. In the case where M = G is a Lie group and the Xi are left invariant, the structure functions are constant.
k k Proof. We prove that Cij(g) = Cij(e) for all g ∈ G.
n X k Cij(g)(Xk)g = [Xi,Xj]g = [dLgXi, dLgXj]g k=1 n X k = dLg[Xi,Xj]e = dLg Cij(e)(Xk)e k=1 n X k = Cij(e)(Xk)g, k=1
k k which implies that Cij(g) = Cij(e) since the (Xk)g form a basis of TgG.
14 Chapter 2
Geodesics on the Space of Kahler¨ Metrics of a Manifold
In this section we present a brief overview of [14]. In the first sections we explain how it is possible to change the complex structure of a compact Kahler¨ manifold (acting with the ”complex flow” of an Hamiltonian vector field), in such a way that the resulting manifold is still Kahler.¨ In the last sections we explain how this change of Kahler¨ metric can be seen as a change in symplectic form, and how it determines a curve in a space of Kahler¨ metrics. Then we give a proof of the result that this curve is in fact a geodesic with respect to a certain metric on the space of Kahler¨ metrics.
2.1 Lie Series and complexified flows
Let (M,J0) be a compact, complex manifold.
Definition 2.1.1. Let S be a real analytic tensor field on M, and X be a real analytic vector field on M.
Denote by LX the Lie derivative with respect to the vector field X. Let τ ∈ C. We define the exponential of τLX applied to S as the formal Lie Series:
∞ X τ k eτLX S := Lk (S). (2.1) k! X k=0
The definition given above is a formal one. We now prove that the exponential converges to a tensor field on M.
Theorem 2.1.2. For all S real analytic tensor field and X real analytic vector field, there exists a T such that if t is real and |t| < T , then etLX S converges, and
tLX t ∗ e S = (φX ) S, (2.2)
t where φX is the flow of X.
15 dk t ∗ k t ∗ Proof. We start by showing by induction that dtk ((φX ) S) = LX ((φX ) S). (k = 1) : d d d t ∗ t+s ∗ s ∗ t ∗ t ∗ (φX ) S = (φX ) S = (φX ) (φX ) S = LX ((φX ) S) (2.3) dt ds s=0 ds s=0
t ∗ t t s t+s (k =⇒ k + 1) : Note that (φX ) X = X, because the curve φX is tangent to X and φX ◦ φX = φX . t ∗ Therefore, LX and (φX ) commute:
t ∗ t ∗ t ∗ (φ ) LX S = L t ∗ (φ ) S = LX (φ ) S. (2.4) X (φX ) X X X
dk+1 d dk d ((φt )∗S) = ((φt )∗S) = Lk ((φt )∗S) dtk+1 X dt dtk X dt X X d = (φt )∗Lk S = L (φt )∗Lk S = (φt )∗Lk+1S, dt X X X X X X X
dk t ∗ k t ∗ Where we used the induction hypothesis. This proves that dtk ((φX ) S) = LX ((φX ) S). In particular,
k d t ∗ k (φX ) S = LX S. (2.5) dtk t=0
t Since X is real analytic, then φX is real analytic in t and as a map on M. Since S is also real analytic, t ∗ t ∗ (φX ) S is real analytic in t and as a tensor field on M. Let p ∈ M. We can expand (φX ) S around t = 0, p. More precisely, for all p in M there exists Vp an open neighborhood of p and a Tp ∈ R such that
∀t : |t| < Tp ∞ X tk (φt )∗S = S (2.6) X k! k k=0
1 n in Vp, where the Sk are some tensors in Vp whose coefficients are power series in x , ..., x centered in (0, ..., 0), that are convergent in Vp. Let {Vp1 , ..., VpN } be a covering of M by open sets as we just described, and take T = min{Tp1 , ..., TpN }. Let t be such that |t| < T . In each Vpj , it is true that
k k ∞ k k d t ∗ d X t LX S = (φX ) S = Sk = Sk (2.7) dtk t=0 dtk t=0 k! k=0
P∞ tk k t ∗ which implies that k=0 k! LX (S) converges to (φX ) S in each Vpj . We conclude that if |t| < T then P∞ tk k t ∗ k=0 k! LX (S) converges to (φX ) S in all of M, because the Vpj form an open covering of M.
Theorem 2.1.3. For all S real analytic tensor field and X real analytic vector field, there exists a T such that if |τ| < T , then eτLX S converges.
P∞ τ k k Proof. Let T be as in theorem 2.1.2. We start by noticing that k=0 k! LX (S) converges if and only P∞ τ k k 1 m 1 m if k=0 k! LX (S)(X1, ..., Xl, ω , ..., ω ) converges for all p ∈ M, X1, ..., Xl ∈ TpM and ω , ..., ω ∈ ∗ 1 m ∗ Tp M. Let p ∈ M, X1, ..., Xl ∈ TpM and ω , ..., ω ∈ Tp M. For ease of notation, denote ak = 1 k 1 m k! LX (S)(X1, ..., Xl, ω , ..., ω ). As a consequence of the last theorem,
∞ X k |τ| < T and τ ∈ R =⇒ akτ converges. (2.8) k=0
16 P∞ k From complex analysis, the power series k=0 akτ has a radius of convergence R, that satisfies:
∞ X k |τ| < R =⇒ akτ converges, (2.9) k=0 ∞ X k |τ| > R =⇒ akτ diverges. (2.10) k=0
For a proof of this, see the Abel power series theorem, for example in [1] page 38. Conditions (2.8) and (2.10) imply that R ≥ T . Condition (2.9) implies that
∞ X τ k |τ| < T =⇒ Lk (S)(X , ..., X , ω1, ..., ωm) converges, (2.11) k! X 1 l k=0 which completes the proof.
Remark 2.1.4. Applying eτLX to holomorphic tensors can be thought of as performing a pullback by a complexification of the flow of X.
We now prove some properties of the exponential.
Proposition 2.1.5. If all the series in each equation converge, then
• if S, R are tensor fields, eτLX (S ⊗ R) = eτLX S ⊗ eτLX R; (2.12)
• if S is a tensor field of type (l, m),
τLX 1 m τLX τLX τLX τLX 1 τLX m e (S(X1, ..., Xl, ω , ..., ω )) = e S(e X1, ..., e Xl, e ω , ..., e ω ); (2.13)
• if Y,Z are vector fields, eτLX [Y,Z] = eτLX Y, eτLX Z . (2.14)
Proof. We start by proving that the equations are true if τ = t is real. In this case, each exponential is in fact a pullback. The equations we have to prove take the form:
t ∗ t ∗ t ∗ (φX ) (S ⊗ R) = (φX ) S ⊗ (φX ) R, (2.15)
t ∗ 1 m t ∗ t ∗ t ∗ t ∗ 1 t ∗ m (φX ) (S(X1, ..., Xl, ω , ..., ω )) = (φX ) S((φX ) X1, ..., (φX ) Xl, (φX ) ω , ..., (φX ) ω ), (2.16) t ∗ t ∗ t ∗ (φX ) [Y,Z] = (φX ) Y, (φX ) Z . (2.17)
Which are known properties of the pullback. Therefore each equation is true if τ is real. Performing complex analytic continuation on each side of each equation, we conclude that the equations are true for complex τ.
Remark 2.1.6. Equations (2.12), (2.13) and (2.14) can also be proven using the following method. Expand each exponential in a series. On the left side of the equation, use a formula for the Lie derivative
17 of a product (Leibniz rule). Then use Cauchy’s formula for the product of series to write the obtained sum as a product of series, which will be the right side of the equation.
2.2 Action of a complexified analytic flow on a complex structure
Theorem 2.2.1 (Mourao˜ and Nunes, [14]). Let (M,J0) be a compact, complex manifold, and X be a real analytic vector field on M. There exists a T > 0 such that for all τ ∈ B(0,T ), there exists an integrable almost complex structure Jτ satisfying:
1 n (i) (Definition of Jτ ) Let p ∈ M, and let (Uα, z0 , ..., z0 ) be a J0-holomorphic coordinate neighborhood
of p. Then, there exists an open neighborhood Vα,p of p such that:
• p ∈ Vα,p ⊂ Vα,p ⊂ Uα;
• Vα,p has compact closure;
j τX j • the series zτ := e · z0 are uniformly convergent on Vα,p;
1 n • (Vα,p, zτ , ..., zτ ) is a Jτ -holomorphic coordinate neighborhood of p.
(ii) There exists a unique biholomorphism ϕτ :(M,Jτ ) −→ (M,J0) that satisfies the following property: 1 n for all p ∈ M, there exists (Uα, z0 , ..., z0 ) a J0-holomorphic coordinate neighborhood of p and there 1 n exists (Vα,p, zτ , ..., zτ ) a Jτ -holomorphic coordinate neighborhood satisfying the conditions of (i), j j such that ϕτ (Vα,p) ⊂ Uα and zτ = z0 ◦ ϕτ .
Proof. See theorems 2.5 and 2.6. in [14].
Remark 2.2.2.
−1 • ϕτ being a biholomorphism means that Jτ = dϕτ ◦ J0 ◦ dϕτ . This condition can also be stated in
terms of the holomorphic functions: f is J0-holomorphic if and only if f ◦ ϕτ is Jτ -holomorphic.
j j j j τX¯ j • Performing complex conjugation on equation z0 ◦ ϕτ = zτ , we obtain that z¯0 ◦ ϕτ =z ¯τ = e · z¯0.
• The biholomorphisms ϕτ do not depend only on τ, but also on the initial complex structure J0. If J is
any complex structure on M, denote by ϕτ,J the unique biholomorphism ϕτ,J :(M,Jτ ) −→ (M,J) j τX j such that z ◦ ϕτ,J = e · z (in this notation, the biholomorphisms of the previous theorem
are written ϕτ = ϕτ,J0 ). If τ, σ ∈ C, then the following diagram is a commutative diagram of biholomorphisms: ϕτ,Jσ (M,Jτ+σ) (M,Jσ)
ϕτ+σ,J0 ϕσ,Jτ ϕσ,J0
(M,Jτ ) (M,J0) ϕτ,J0
We now note that:
t ∗ j tX j ◦ If t ∈ R, since (φX ) z = e · z , then ϕt is the flow of X:
t ϕt = φX , (2.18)
18 and therefore it does not depend on J0. In this case, ϕt+s = ϕt ◦ ϕs.
◦ If τ, σ ∈ C \ R, then in general it is not true that ϕτ+σ = ϕτ ◦ ϕσ, which means that ϕ is not a
flow. What we can say about about ϕτ+σ is that
ϕτ+σ = ϕτ+σ,J0 = ϕτ,J0 ◦ ϕσ,Jτ (2.19a)
= ϕσ+τ,J0 = ϕσ,J0 ◦ ϕτ,Jσ . (2.19b)
2.3 Restriction to Hamiltonian flows
We now consider the case where (M, ω0,J0, g0) is a compact Kahler¨ manifold, with all three structures an real analytic. Let h ∈ C (M). Consider the Hamiltonian vector field Xh defined by h.
Proposition 2.3.1. Let f ∈ Can(M). Suppose that eτXh ·f is well defined. Then, the Hamiltonian vector
field of the function eτXh · f is given by:
τLXh XeτXh ·f = e · Xf . (2.20)
k Proof. We start by proving by induction on k that L Xf = X k . Xh Xh (f)
(k = 1) : LXh Xf = [Xh,Xf ] = X{h,f} = XXh(f)
k+1 k k k k k+1 (k =⇒ k + 1) : L Xf = LXh LX Xf = LXh XX (f) = [Xh,XX (f)] = X{h,X (f)} = X Xh h h h h Xh (f)
Now we prove the result.
∞ τ k ∞ τ k X X k τLX X τX = X k = X k = L X = e h · X . e h ·f P∞ τ Xk(f) X f Xh f f k=0 k! h k! h k! k=0 k=0
Note that in the case τ ∈ R, equation 2.20 is the known statement that symplectomorphisms preserve Hamiltonian vector fields.
1 n 1 n Proposition 2.3.2. Let (U, z , ..., z ) be a J0-complex coordinate chart on M and let (V, zτ , ..., zτ ) be a 1 n Jτ coordinate chart defined by (U, z , ..., z ). Then on V we have that
τLX e h X j = X j , (2.21a) z zτ
τL¯ X e h X j = X j . (2.21b) z¯ z¯τ
Proof. The result follows from using proposition 2.3.1.
2.4 Action on Kahler¨ structures
In section 2.2 we gave a procedure that changes the complex structure of M from J0 to Jτ . Consider this procedure applied to (M, ω0,J0, g0). We will obtain a new structure (M, ω0,Jτ ).
19 Question: Is the resulting structure Kahler?¨ In other words, is it true that gτ := ω0(·,Jτ ·) is a Rieman- nian metric? Answer: At least for small values of |τ|, the answer is yes, provided by the following theorem.
Theorem 2.4.1 (Mourao˜ and Nunes, [14]). Let (M, ω0,J0, g0) be a compact Kahler¨ manifold, with ω0, J0, an g0 real analytic. Let h ∈ C (M). Then, there exists a T > 0 such that:
(i) For all τ ∈ B(0,T )(M, ω0,Jτ , gτ ) is a Kahler¨ manifold, where:
• Jτ is the complex structure defined by applying theorem 2.2.1 to (M, ω0,J0, g0) with the vector
field Xh;
• gτ := ω0(·,Jτ ·).
(ii)(K ahler¨ potential for (M, ω0,Jτ , gτ )) For all p ∈ M there exists:
1 n • (Uα, z0 , ..., z0 ) a J0-holomorphic coordinate neighborhood of p;
• κ0 : Uα −→ R a local Kahler¨ potential for (M, ω0,J0);
• Vα,p an open set such that:
◦ p ∈ Vα,p ⊂ Vα,p ⊂ Uα;
◦ Vα,p has compact closure;
◦ for all τ ∈ B(0,T ), ϕτ (Vα,p) ⊂ Uα;
◦ for all τ ∈ B(0,T ), κτ defined by
i θ := (∂ − ∂¯ )κ (2.22a) 2 0 0 0 Z t sXh αt := e (θ(Xh))ds (2.22b) 0 ατ := unique complex analytic continuation of αt (2.22c) i ψ := − eτXh · κ + τh − α (2.22d) τ 2 0 τ
κτ := −2 Im ψτ (2.22e)
is well defined on Vα,p (because the Lie series in (2.22b), (2.22c) and (2.22d) are uni-
formly convergent), and is a local Kahler¨ potential for (M, ω0,Jτ ).
Proof. Let T 0 be as in theorem 2.2.1. Along the proof, we will restrict T 0 as we need to. 0 (i): We need to prove that for all τ ∈ B(0,T ), ω0(·,Jτ ·) is symmetric and positive definite. Recall that symmetry is equivalent to ω0 being of type (1, 1). Since ω0 is of type (1,1) with respect to J0 and is preserved by Xh, it is of type (1, 1) with respect to Jτ :
!! ∂ ∂ τL ∂ ∂ ω , = e Xh ω , = 0. (2.23) 0 j ∂zk 0 j k ∂zτ τ ∂z0 ∂z0
0 0 ω0(·,J0·) is positive definite. By continuity, we can restrict T such that for all τ ∈ B(0,T ) ω0(·,Jτ ·) is positive definite.
20 1 n (ii): Let p ∈ M. There exists (Uα, z0 , ..., z0 ) a J0-holomorphic coordinate neighborhood of p that is the domain of a Kahler¨ potential κ0 : U −→ R. Let Vα,p be such that:
• p ∈ Vα,p,τ ⊂ Vα,p,τ ⊂ Uα;
• Vα,p,τ has compact closure.
By a theorem of Grobner and Knapp (theorem 3 of [7]), there exists Tα,p > 0 such that for all τ ∈
B(0,Tα,p) the series defining κτ is uniformly convergent on Vα,p. Restrict Tα,p so that ϕτ (p) ∈ Uα, and shrink Vα,p so that ϕτ (Vα,p) ⊂ Uα. We have just proven that for all p ∈ M, there exists
1 n • (Uα, z0 , ..., z0 ) a J0-holomorphic coordinate neighborhood of p;
• κ0 : Uα −→ R a local Kahler¨ potential for (M, ω0,J0);
• Tα,p > 0, Vα,p an open set such that:
◦ p ∈ Vα,p ⊂ Vα,p ⊂ Uα;
◦ Vα,p has compact closure;
◦ for all τ ∈ B(0,Tα,p), ϕτ (Vα,p) ⊂ Uα;
◦ for all τ ∈ B(0,Tα,p), the series defining κτ is uniformly convergent on Vα,p.
0 Let {Vαj ,pj }j=1,...,N be a finite open cover of M. Take T = min{Tα1,p1 , ..., TαN ,pN ,T }. This is the T of the statement of the theorem, since each p belongs to some Vαj ,pj . We now prove that κτ is a Kahler¨ potential for (M, ω0,Jτ ). The proof of this is a lengthy computation. We will divide it into parts.
Part 1: we prove that θ is real and that dθ = −ω0.
i θ¯ = − (∂¯ − ∂ )κ = θ 2 0 0 0
i i i i θ = ∂ κ − ∂¯ κ =⇒ θ(1,0) = ∂ κ and θ(0,1) = − ∂¯ κ 2 0 0 2 0 0 2 0 0 2 0 0 | {z } | {z } ∈Ω(1,0)(M) ∈Ω(0,1)(M) i i dθ = − (∂ + ∂¯ )(∂¯ − ∂ )κ = − (∂ ∂¯ − ∂¯ ∂ )κ = −i∂ ∂¯ κ = −ω 2 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0
P∞ τ k k−1 Part 2: we prove that ατ = k=1 k! Xh (θ(Xh)).
∞ Z t Z t X uk α = euXh (θ(X ))du = Xk(θ(X ))du t h k! h h 0 0 k=0 ∞ ∞ X 1 Z t X 1 tk+1 = ukdu Xk(θ(X )) = Xk(θ(X )) k! h h k! k + 1 h h k=0 0 k=0 ∞ X tk = Xk−1(θ(X )) k! h h k=1
Therefore, the complex analytic continuation of αt is
21 ∞ X τ k α = Xk−1(θ(X )). τ k! h h k=1
Part 3: we prove that ∀k ≥ 2 : Lk θ = (dι )kθ. We proceed by induction on k. Xh Xh (k = 2) :
L2 θ = (dι + ι d)2θ Xh Xh Xh
= dιX dιX θ + dιX ιX dθ +ιX dd ιX θ + ιX dιX dθ h h h h |{z} h |{z} h h h |{z} =−ω0 =0 =−ω0 2 = (dιXh ) θ − d(ω(Xh,Xh)) − ιXh ddh
2 = (dιXh ) θ
(k =⇒ k + 1) :
Lk+1θ = (dι + ι d)Lk θ = (dι + ι d)(dι )kθ Xh Xh Xh Xh Xh Xh Xh k+1 k k+1 = (dιXh ) θ + ιXh d(dιXh ) θ = (dιXh ) θ
Part 4: we prove that ∀k ≥ 2 : d Xk−1(θ(X )) = Lk (θ). We proceed by induction on k. h h Xh (k = 2) : L2 θ = (dι )2θ = dι dι θ = dι d(θ(X )) = d(X (θ(X ))) Xh Xh Xh Xh Xh h h h
(k =⇒ k + 1) :
Lk+1θ = (dι + ι d)d Xk−1(θ(X )) = dι d Xk−1(θ(X )) = d Xk(θ(X )) Xh Xh Xh h h Xh h h h h
τdιX Part 5: we prove that dατ = e h θ − θ.
∞ ! ∞ ! X τ k X τ k dα = d Xk−1(θ(X )) = τd(θ(X )) + d Xk−1(θ(X )) τ k! h h h k! h h k=1 k=2 ∞ τ k ∞ τ k X k X k τdιX = τd(θ(Xh)) + L (θ) = (dιX ) (θ) = e h θ − θ k! Xh k! h k=2 k=1
τLX Part 6: we prove that e h θ = θ − τdh + dατ .
∞ τ k ∞ τ k τLX X k X k e h θ = L (θ) = θ + τLX (θ) + L (θ) k! Xh h k! Xh k=0 k=2 ∞ X τ k = θ + τ(dι + ι d)θ + (dι )k(θ) Xh Xh k! Xh k=2 ∞ X τ k = θ − τdh + τdι θ + (dι )k(θ) Xh k! Xh k=2 ∞ X τ k = θ − τdh + (dι )k(θ) = θ − τdh + dα k! Xh τ k=1
22 (0,1)τ Part 7: we prove that θ = ∂¯ ψ . This is equivalent to θ(X j ) = dψ (X j ) ∀j = 1, ..., n. τ τ zτ τ zτ
i τXh dψτ (X j ) = − d(e · κ0)X j + τdh(X j ) − dατ (X j ) zτ 2 zτ zτ zτ i τXh τLX = − d(e · κ0)X j + θ(X j ) − (e h θ)(X j ), 2 zτ zτ zτ
which is equal to θ(X j ), since zτ
i τLX τLX τLX τXh τXh ¯ (e h θ)(X j ) = (e h θ)(e h Xzj ) = e (θ(Xzj )) = − e (∂0κ0(Xzj )) zτ 2 i i i τXh τLX τLX τLX = − e (dκ0(Xzj )) = − (e h (dκ0))(e h Xzj ) = − d(e h κ0)X j . 2 2 2 zτ
Part 8: we prove the result.
¯ ¯ ¯ ¯ ¯ i∂τ ∂τ κτ = −i∂τ ∂τ (2 Im ψτ ) = ∂τ ∂τ ψτ − ∂τ ∂τ ψτ
¯ ¯ ¯ ¯ (0,1)τ (0,1)τ = −∂τ ∂τ ψτ − ∂τ ∂τ ψτ = −∂τ θ − ∂τ θ
¯ (1,0)τ (0,1)τ (1,0)τ (0,1)τ ¯ (1,0)τ ¯ (0,1)τ = −∂τ θ − ∂τ θ = −∂τ θ − ∂τ θ − ∂τ θ − ∂τ θ
¯ (1,0)τ (0,1)τ = −(∂τ + ∂τ )(θ + θ ) = −dθ = ω0
Proposition 2.4.2. Let T be as in the previous theorem, and let τ ∈ B(0,T ). Let s ∈ R be such that ∗ τ + s ∈ B(0,T ). Then ϕsκτ = κτ+s.
∗ Proof. We start by proving that ϕsαt = αt+s − αs.
Z t ∗ sXh sXh uXh ϕsαt = e · αt = e e (θ(Xh))du 0 Z t Z t+s (s+u)Xh uXh = e (θ(Xh))du = e (θ(Xh))du 0 s Z t+s Z s uXh uXh = e (θ(Xh))du − e (θ(Xh))du = αt+s − αs 0 0
sX Therefore, by complex analytic continuation, e h · ατ = ατ+s − αs.
To prove the result, we now use the definition of κτ in terms of ατ given by theorem 2.4.1.
1 κ = i(ψ − ψ¯ ) = (eτXh − eτX¯ h ) · κ + i(τ − τ¯)h − i(α − α ) τ τ τ 2 0 τ τ¯
Therefore
1 sXh (s+τ)Xh (s+¯τ)Xh sXh sXh sXh e · κτ = e − e · κ0 + i (τ + s) − (¯τ + s) e · h −i( e ατ − e ατ¯ ) 2 | {z } | {z } | {z } =h =ατ+s−αs =ατ¯+s−αs 1 = e(s+τ)Xh − e(s+¯τ)Xh · κ + i (τ + s) − (¯τ + s) h − i(α − α ) = κ . 2 0 τ+s τ¯+s τ+s
23 2.5 Geodesics on the Space of Kahler¨ metrics
In this subsection we define and describe the space of Kahler¨ metrics. We also explain what are curves in this space, define the Mabuchi metric and give the geodesic equation for this metric.
Definition 2.5.1. The space of Kahler¨ metrics on M in the cohomology class [ω0] is
∗ ∗ ∗ H(ω0,J0) := {ϕ ω0 | ϕ ∈ Diff(M), [ϕ ω0] = [ω0], (M, ϕ ω, J0) is Kahler¨ }. (2.24)
The space of Kahler¨ potentials on M with base point [ω0] is
∞ ¯ K(ω0,J0) := {φ ∈ C (M) | g˜ := (ω0 + i∂0∂0φ)(·,J0·) is positive definite}. (2.25)
Remark 2.5.2.
¯ • As stated in [5], by the ∂∂-lemma any other Kahler¨ metric that is cohomologous to ω0 can be
written using a global Kahler¨ potential. This implies that H(ω0,J0) can be identified with the space of Kahler¨ potentials, modulo constants:
∼ H(ω0,J0) = K(ω0,J0)/R (2.26a) Z ∼ ∞ ¯ n = φ ∈ C (M) | g˜ := (ω0 + i∂0∂0φ)(·,J0·) 0, φω0 = 0 . (2.26b) M
We give a proof that this bijection exists below.
•H (ω0,J0) can be regarded as an infinite dimensional manifold, where a tangent vector δφ0 to ∞ H(ω0,J0) at φ0 is a function on M. To see this, consider a curve in H(ω0,J0), t 7→ φt ∈ C (M) R n (where we have chosen a family of representatives for the equivalence classes such that M φtω0 =
0, so that the map t 7→ φt is smooth). The derivative with respect to t is another function:
d ∞ δφ0 := φt ∈ C (M). (2.27) dt t=0
•H (ω0,J0) can be equipped with a Riemannian metric called the Mabuchi metric, which is given by
Z 1 hδ1φ, δ2φi = (δ1φ · δ2φ) ωφ ∧ ... ∧ ωφ, (2.28) M n! ¯ where ωφ = ω0 + i∂0∂0φ. As proven in [5], the Riemannian manifold H(ω0,J0) admits a unique
Levi-Civita connection. A curve {φt}t∈I is a geodesic for this connection if and only if
¨ 1 ˙ 2 φt = k∇g˜ φtk , (2.29) 2 t g˜t
where ∇g˜t is the gradient, and k · kg˜t is the norm, both with respect to the metric g˜t = (ω0 + ¯ i∂0∂0φt)(·,J0·).
24 We now prove the statement of 2.5.2. Note that the proof of the following theorem is just a description of the mentioned bijection.
Theorem 2.5.3. There exists a bijection
∼ H(ω0,J0) = K(ω0,J0)/R.
∗ ∗ Proof. Part 1: We give a map H(ω0,J0) −→ K(ω0,J0)/R. Let ϕ ω0 ∈ H(ω0,J0). Since [ϕ ω0] = [ω0], ¯ ∞ ∗ ¯ ∗ by the ∂∂-Lemma there exists a φ ∈ C (M) such that ϕ ω0 = ω0 + i∂0∂0φ. Since (M, ϕ ω, J0) is ¯ ∗ Kahler,¨ (ω0 + i∂0∂0φ)(·,J0·) = (ϕ ω0)(·,J0·) is positive definite, and φ belongs to K(ω0,J0)/R. Define ∗ the image of ϕ ω0 to be the equivalence class of φ in K(ω0,J0)/R. We still have to show that this map 0 H(ω0,J0) −→ K(ω0,J0)/R is well defined. Specifically, we have to show that if φ is another function in ∗ ¯ 0 0 ¯ 0 M such that ϕ ω0 = ω0 + i∂0∂0φ , then φ and φ differ by a constant. This is true, since i∂0∂0(φ − φ ) = 0 and M is compact (by lemma 1.3.22). ¯ Part 2: We give a map K(ω0,J0)/R −→ H(ω0,J0) Let φ ∈ K(ω0,J0)/R. Note that i∂0∂0φ does not ¯ depend on the chosen representative for this equivalence class. Since ω0 + i∂0∂0φ is of type (1, 1), g˜ is ¯ symmetric. By hypothesis it is positive definite, hence it is a Riemannian metric. ω0 + i∂0∂0φ is closed. ¯ Since g˜ is nondegenerate and J0 is an isomorphism, ω0 +i∂0∂0φ is symplectic. We have just proven that ¯ ¯ ¯ ¯ (M, ω0 + i∂0∂0φ, J0) is Kahler.¨ Since i∂0∂0φ = d(i∂0φ) is exact, [ω0 + i∂0∂0φ] = [ω0]. We now prove that ¯ ∗ ¯ ω0 + i∂0∂0φ is of the form ϕ ω0, for some ϕ. If this is true, then ω0 + i∂0∂0φ belongs to H(ω0,J0), so we ¯ can define the image of φ to be ω0 + i∂0∂0φ. Define
¯ ωt = (1 − t)ω0 + t(ω0 + i∂0∂0φ).
Then ωt is closed. Also,
¯ ωt(·,J0·) = (1 − t)ω0(·,J0·) + t(ω0 + i∂0∂0φ)(·,J0·) is positive definite, and again since J0 is an isomorphism, ωt is nondegenerate. So, ωt is symplectic, ∞ ∗ ¯ and using Moser’s theorem we conclude that there exists a ϕ ∈ C (M) such that ϕ ω0 = ω0 + i∂0∂0φ. It is easily seen that the given functions are the inverse of the other.
2.6 Symplectic picture and complex picture
Consider the Kahler¨ manifold (M, ω0,J0, g0). We compare two different, but as we will prove equivalent, ways of changing the Kahler¨ structure:
• Symplectic picture: Fixed ω0, varying Jτ .
an Let h ∈ C (M). Using the procedure described in the previous sections with the vector field Xh
we get a new Kahler¨ structure (ω0,Jτ ).
• Complex picture: Fixed J0, varying ωτ .
25 ¯ Let φτ ∈ H(ω0,J0). Consider the new symplectic form ωτ := ω0 + i∂0∂0φτ . We get a new Kahler¨
structure (ωτ ,J0).
Let (M, ω, J) be a Kahler¨ manifold and φ: M −→ M be a diffeomorphism. It is possible to prove that:
(i) φ∗ω is a symplectic form on M;
(ii) φ∗J is a complex structure on M;
(iii) (M, φ∗ω, φ∗J) is a Kahler¨ manifold.
This fact can be used to write the symplectic picture in terms of the complex picture, in the following way: consider a Kahler¨ structure (ω0,Jτ ) obtained from (ω0,J0) by applying the method of the previous sections (symplectic picture). The new Kahler¨ structure comes with a biholomorphism
ϕτ :(M,Jτ ) −→ (M,J0). (2.30)
−1 ∗ Define ωτ = (ϕτ ) ω0, so that ϕτ is a Kahler¨ isomorphism:
ϕτ :(M, ω0,Jτ ) −→ (M, ωτ ,J0). (2.31)
Then, (ωτ ,J0) is a Kahler¨ structure, isomorphic to (ω0,Jτ ), and with the same complex structure as the initial Kahler¨ structure.
2.7 The path of Kahler¨ metrics (generated by a complex flow of an Hamiltonian vector field) is a geodesic
an Let (M, ω0,J0) be Kahler.¨ Let h ∈ C (M), and consider its Hamiltonian vector field Xh. Let T be as in theorem 2.4.1. For all τ = it ∈ B(0,T ), where t ∈ R, by acting with the complexified flow of Xh we obtain new Kahler¨ structures (ω0,Jit) on M. We consider the path {(ω0,Jit)}t∈(−T,T ). Consider the path −1 ∗ −1 ∗ of isomorphic Kahler¨ structures {((ϕit ) ω0, (ϕit ) Jit)}t∈(−T,T ) = {(ωit,J0)}t∈(−T,T ). In short:
itX ϕ e h ∼it (M, ω0,J0, g0) (M, ω0,Jit, git) = (M, ωit,J0, g˜it). (2.32)
Theorem 2.7.1. ωit ∈ H(ω0,J0), for all t ∈ (−T,T ).
∗ Proof. Define Φt := ϕit. Then Φt ωit = ω0. Since Φt is homotopic to the identity, we have that:
∗ ∗ [ω0] = [Φt ωit] = Φt [ωit] = [ωit].
Therefore ωit ∈ H(ω0,J0).
¯ For each t ∈ (−T,T ), ωit ∈ H(ω0,J0), so we can find a φt such that ωit = ω0 + i∂0∂0φt. Writing ω0 in
26 terms of its local Kahler¨ potentials:
¯ ω0 = i∂0∂0κ0, ¯ ω0 = i∂it∂itκit.
∗ Also, recall that ω0 = Φt ωit. Using these formulas,
¯ ∗ ∗ ¯ ¯ i∂it∂itκit = ω0 = Φt ωit = iΦt ∂0∂0(κ0 + φt) = i∂it∂it (κ0 + φt) ◦ Φt .
This suggests that we define −1 φt = κit ◦ Φt − κ0. (2.33)
0 0 However, this definition only works if when we choose a different Kahler¨ potential κ0 (that defines κit), −1 0 −1 0 the functions κit ◦ Φt − κ0 and κit ◦ Φt − κ0 agree on the intersection of their domains of definition.
Definition 2.7.2. For each t ∈] − T,T [, define a function φt on M in the following way. Let p ∈ M. −1 Let U, V be neighborhoods of ϕτ (p) as in theorem 2.4.1, with local Kahler¨ potentials κ0 : U −→ R and
κτ : V −→ R. Then, in the neighborhood ϕτ (V ) of p
−1 φt|ϕτ (V ) = κτ ◦ ϕτ − κ0. (2.34)
Proposition 2.7.3.
(i) φt is well defined.
¯ (ii) ωit = ω0 + i∂0∂0φt.
Proof. (i): To prove this, one must use equations (2.22a) to (2.22e) given in theorem 2.4.1, that define 0 0 0 κτ in terms of κ0, for both κτ and κτ , and then, check that κτ − κτ = (κ0 − κ0) ◦ ϕτ . Performing the computations:
∞ ! X τ k κ0 − κ = Im ieτXh · (κ0 − κ ) + i Xk−1(∂ (κ0 − κ )(X ) − ∂¯ (κ0 − κ )(X )) τ τ 0 0 k! h 0 0 0 h 0 0 0 h k=1 ∞ ! X τ k = Im i(κ0 − κ ) + 2i Xk−1(∂ (κ0 − κ )(X )) 0 0 k! h 0 0 0 h k=1 ∞ X τ k τ¯k = κ0 − κ + Xk−1(∂ (κ0 − κ )(X )) + Xk−1(∂¯ (κ0 − κ )(X )) 0 0 k! h 0 0 0 h k! h 0 0 0 h k=1
0 ¯ 0 κ0 − κ0 is a real analytic function defined on an open subset, such that ∂0∂0(κ0 − κ0) = 0. By lemma 0 ¯ ¯ 1.3.23, κ0 − κ0 is locally of the form f + f, where f is holomorphic and f is anti-holomorphic. On each 0 ¯ smaller open set where κ0 − κ0 = f + f:
∞ ∞ X τ k X τ¯k κ0 − κ = f + Xk−1(df(X )) + f¯+ Xk−1(df¯(X )) τ τ k! h h k! h h k=1 k=1 τXh τX¯ h ¯ 0 = e · f + e · f = (κ0 − κ0) ◦ ϕτ .
27 0 0 Therefore κτ − κτ = (κ0 − κ0) ◦ ϕτ on the whole domain where this expression is defined. (ii): The proof is the following computation:
¯ ¯ ¯ −1 ¯ ω0 + i∂0∂0φt = i∂0∂0κ0 + i∂0∂0(κit ◦ Φt ) − i∂0∂0κ0 ¯ −1 ∗ −1 ∗ ¯ = i∂0∂0((Φt ) κit) = (Φt ) i∂it∂it(κit)
−1 ∗ = (Φt ) ω0 = ωit
The following diagram illustrates the construction and main formulas so far:
eitXh Φ∼t (M, ω0,J0, g0) (M, ω0,Jit, git) = (M, ωit,J0, g˜it) ¯ ¯ ¯ ω0 = i∂0∂0κ0 ω0 = i∂it∂itκit ωit = ω0 + i∂0∂0φt ∈ H(ω0,J0) (2.35)
∗ Φt ωit = ω0
Also, recall the formulas for κit (given in theorem 2.4.1):
i θ := (∂ − ∂¯ )κ (2.36a) 2 0 0 0 ∞ X (it)k α := Xk−1(θ(X )) (2.36b) it k! h h k=1 i ψ := − eitXh · κ + ith − α (2.36c) it 2 0 it
κit := −2 Im ψit (2.36d)
Theorem 2.7.4 (Mourao˜ and Nunes, [14]). φt satisfies:
¨ 1 ˙ 2 φt = k∇g˜ φtk . (2.37) 2 it g˜it
Therefore, the path of Kahler¨ metrics {(ω0,Jit)}t∈(−T,T ) is a geodesic.
Proof. We divide the proof into parts. d Part 1: We show that dt αit = iXh(αit) + iθ(Xh).
∞ ! ∞ d d X (it)k X (it)k−1 α = Xk−1(θ(X )) = i Xk−1(θ(X )) dt it dt k! h h (k − 1)! h h k=1 k=1 ∞ ! X (it)k = iX Xk−1(θ(X )) + iθ(X ) = iX (α ) + iθ(X ) h k! h h h h it h k=1
d ¯ Part 2: We show that dt κit = −Xh(ψit + ψit) − 2h + 2θ(Xh). Using (2.36c) and (2.36d), we conclude that
1 1 κ = eitXh · κ − e−itXh · κ − 2th − iα + iα . it 2 0 2 0 it −it
28 Therefore
d 1 1 d d κ = X ieitXh · κ + X (ie−itXh · κ ) − 2h − i α + i α dt it 2 h 0 2 h 0 dt it dt −it i i = −X − eitXh · κ − e−itXh · κ − α − α − 2h + 2θ(X ) h 2 0 2 0 it −it h ¯ = −Xh(ψit + ψit) − 2h + 2θ(Xh).
Part 3: Let f ∈ C∞(M). We show that:
d (f ◦ ϕ−1) = ∂ f(−iX ) + ∂¯ f(iX ) ◦ ϕ−1. (2.38) dt it it h it h it
We want to compute
d d d d (f ◦ ϕ−1) = df ϕ−1 = ∂ f ϕ−1 + ∂¯ f ϕ−1 . (2.39) dt it dt it it dt it it dt it
j −1 j −itXh j Recall that ϕit is the unique biholomorphism satisfying zit ◦ ϕit = z = e · zit.
d d d dzj ϕ−1 = zj ◦ ϕ−1 = e−itXh · zj = −iX e−itXh · zj it dt it dt it it dt it h it j −1 −1 ∗ j −1 ∗ j = −iXh zit ◦ ϕit = −iXh (ϕit ) zit = d (ϕit ) zit (−iXh) j −1 = dzit d(ϕit )(−iXh)
Performing complex conjugation, we conclude that:
d dzj ϕ−1 = dzj d(ϕ−1)(−iX ) , (2.40) it dt it it it h d dz¯j ϕ−1 = dz¯j d(ϕ−1)(iX ) . (2.41) it dt it it it h
j ¯ j ∂itf is a linear combination of the dzit, and ∂itf is a linear combination of the dz¯it. Because of this, equations (2.39), (2.40) and (2.41), imply that:
d (f ◦ ϕ−1) = ∂ f d(ϕ−1)(−iX ) + ∂¯ f d(ϕ−1)(iX ) dt it it it h it it h −1 ∗ ¯ = (ϕit ) ∂itf(−iXh) + ∂itf(iXh) ¯ −1 = ∂itf(−iXh) + ∂itf(iXh) ◦ ϕit .
˙ d ¯ −1 Part 4: We show that φt = dt κit + ∂itκit(−iXh) + ∂itκit(iXh) ◦ Φt .
˙ d d −1 d −1 d −1 rClφt = φt = κit ◦ Φt = κis ◦ Φt + κit ◦ Φs dt dt ds s=t ds s=t
Now the result follows form using equation (2.38).
29 ˙ −1 Part 5: We show that φt = −2h ◦ Φt .
d φ˙ = κ + ∂ κ (−iX ) + ∂¯ κ (iX ) ◦ Φ−1 t dt it it it h it it h t ¯ ¯ −1 = −Xh(ψit + ψit) − 2h + 2θ(Xh) + ∂itκit(−iXh) + ∂itκit(iXh) ◦ Φt making the following substitutions:
¯ ¯ Xh(ψit) = dψit(Xh) = ∂itψit(Xh) + ∂itψit(Xh) (and analogously for ψit)
(1,0)it (0,1)it ¯ ¯ θ = θ + θ = ∂itψit + ∂itψit
κit = −2 Im ψit
˙ −1 we obtain that φt = −2h ◦ Φt . Part 6: We show that
φ¨ ◦ Φ = 2k∇ hk2 t t git git (2.42)
˙ −1 Using the fact that φt = −2h ◦ Φt ,
d d φ¨ = φ˙ = −2 (h ◦ Φ−1) = −2 ∂ h(−iX ) + ∂¯ h(iX ) ◦ Φ−1, t dt t dt t it h it h t where again we used (2.38). Therefore
¨ ¯ φt ◦ Φt = 2i∂ith(Xh) − 2i∂ith(Xh)
= 2dh(JitXh) = 2ω(Xh,JitXh)
= 2g (X ,X ) = 2k∇ hk2 . it h h git git
Part 7: We show that k∇ hk2 = 1 k∇ φ˙ k2 ◦ Φ . From this and (2.42), the result follows. We git git 4 g˜it t g˜it t note the following equivalences:
2 1 ˙ 2 k∇g hk = k∇g˜ φtk ◦ Φt it git 4 it g˜it 1 ˙ ˙ ⇐⇒ git (∇git h, ∇git h)p = g˜it ∇g˜it φt, ∇g˜it φt 4 Φt(p) ∗ 1 ˙ ˙ ⇐⇒ Φt g˜it (∇git h, ∇git h)p = g˜it ∇g˜it φt, ∇g˜it φt 4 Φt(p) 1 ⇐⇒ g˜ (dΦ ∇ h, dΦ ∇ h) = g˜ ∇ φ˙ , ∇ φ˙ it t git t git Φt(p) it g˜it t g˜it t 4 Φt(p)
˙ Therefore it suffices to prove that ∇g˜it φt = 2dΦt∇git h:
˙ ˙ −1 −1 g˜it ∇g˜it φt, · = dφt = −2dh ◦ dΦt = −2git(∇git h, ·) ◦ dΦt
−1 ∗ −1 = −2git(∇git h, dΦt ·) = −2Φt g˜it(∇git h, dΦt ·)
=g ˜it(−2dΦt∇git h, ·)
30 Chapter 3
Kahler¨ structures on the Cotangent Bundle of a Lie Group
Let G be a compact Lie group, with a left and right invariant Riemannian metric h·, ·i. In this chapter we study possible Kahler¨ structures on T ∗G. We start by giving an overview of known Kahler¨ structures on T ∗G. These are the standard Kahler¨ structure [12], the Kahler¨ structure of Hall and Kirwin [10], and the Kahler¨ structure of Kirwin, Mourao˜ and Nunes [12]. We study if it is possible to strengthen the result of Hall and Kirwin to R = ∞ if the manifold M is a Lie group, and if the condition of h being Ad-invariant is necessary in the theorem of Kirwin, Mourao˜ and Nunes. We also compare the various Kahler¨ structures.
3.1 T ∗G is diffeomorphic to G × g
Each element of T ∗G can be described in terms of its base point in G and the correspondent covector by left translation in g∗. In more precise terms, the following map is a diffeomorphism:
T ∗G −→ G × g∗ (3.1)
∗ α 7−→ (π(α), (dLπ(α)) α).
The metric defines a correspondence between vectors and covectors:
∗ [: g g :# (3.2) Y 7→ hY, ·i.
These two diffeomorphisms can be composed into a diffeomorphism
φ: T ∗G −→ G × g (3.3)
∗ # α 7−→ (π(α), ((dLπ(α)) α) ).
31 3.2 Kahler¨ structures on T ∗G
3.2.1 Standard Kahler¨ structure on T ∗G
We start by explaining some necessary theory about the complexification of a Lie group.
Remark 3.2.1 (Complexification of a Lie group).
• Let G be a Lie group. A universal complexification of G is a complex Lie group GC with a Lie
group homomorphism ι : G −→ GC such that for any Lie group homomorphism ρ: G −→ H (where
H is a complex Lie group), there exists an unique Lie group homomorphism ρC : GC −→ H such
that ρC ◦ ι = ρ:
ι G GC ∃!ρ ρ C H
It is a fact that for any Lie group G the universal complexification exists and is unique up to isomor- phism.
• In the case where G is compact, the universal complexification has a simpler description. If G is compact, as a consequence of representation theory, G can be regarded as a matrix Lie group. iX Then, the matrix Lie group GC = ge : g ∈ G, X ∈ g with product, differential structure and complex structure given by those of the matrices with complex entries, and with inclusion
ι: G −→ GC (3.4) g 7−→ gei0
is the universal complexification of G.
iY • The complex structure of G at the point ge is a map JG : T iY (G ) −→ T iY (G ). By left C C ge C ge C ∼ translation, T iY (G ) = g , which in turn is isomorphic to g × g. Therefore, JG can be regarded ge C C C
as a map JG : g × g −→ g × g, which is given as a block matrix as (see [12], page 1468): C
0 −I JG = . (3.5) C I 0
• If we define
ψτ : G × g −→ GC (3.6) (g, Y ) 7−→ geτY
then for Im τ 6= 0, ψτ are diffeomorphisms (see [12], page 1467).
32 Recall that from example 1.1.6, T ∗G is a symplectic manifold with symplectic form ω. Consider the following diffeomorphism:
∗ φ ψi T G −→ G × g −→ GC. (3.7)
∗ Define a complex structure on T G by pulling back the complex structure of GC:
∗ JST = (ψi ◦ φ) JG . (3.8) C
∗ ∗ Then, (T G, ω, JST ) is a Kahler¨ manifold, and (ω, JST ) is the standard Kahler¨ structure on T G.
3.2.2 The Kahler¨ structure of Hall and Kirwin
In [10], Hall and Kirwin describe a Kahler¨ structure on a tubular open set of T ∗M, where M is a real analytic compact Riemannian manifold.
Theorem 3.2.2 (Hall and Kirwin, [10]). Let M be a compact, real analytic manifold equipped with a real analytic Riemannian metric g. Let T ∗,RM denote the tubular neighborhood of radius R of the zero section:
T ∗,RM := α ∈ T ∗M|g(α#, α#) < R2 . (3.9)
Let
∗ h: T M −→ R (3.10) 1 α 7−→ g(α#, α#). 2
Let β ∈ Ω2(M) be closed, and define
∗ ωβ = ωST − π β, (3.11)
∗ where ωST denotes the standard symplectic form on T M. Then, ωβ is a symplectic form, and there exists an R such that:
∗,R (i) There exists an unique complex structure on T M, JMCS, (called the magnetic complex struc- ture), such that
∗,R −iLX T1,0(T M) = e h (ker dπ ⊗ C), (3.12)
where Xh is the Hamiltonian vector field of h with respect to ωβ.
∗,R (ii) (T M, ωβ,JMCS) is Kahler.¨
33 3.2.3 The Kahler¨ structure of Kirwin, Mourao˜ and Nunes
Let G be a compact Lie Group, equipped with a bi-invariant Riemannian metric h·, ·i. In [12], Kirwin, Mourao˜ and Nunes describe a Kahler¨ structure on T ∗G, that can be obtained by changing the complex structure by means of a Thiemann complexifier function. We present some necessary definitions.
Definition 3.2.3. Let h: T ∗G −→ R be left invariant:
∗ ∗ h((dLg) α) = h(α) ∀α ∈ T G, g ∈ G. (3.13)
We refer to h as the Thiemann complexifier function. We define:
• u, the gradient of h:
u: g −→ g (3.14)
Y 7−→ (dh)Y [ ,
∗ ∗ ∗ where (dh)Y [ ∈ g in the following sense. (dh)Y [ is a map TY [ (T G) −→ R. Note that g = Te G ⊂ ∗ ∗ ∗ TY [ (T G). Restrict (dh)Y [ to g , so that it becomes a linear functional g −→ R:
∗ ∗ ∗ (dh)Y [ : g = Te G ⊂ TY [ (T G) −→ R.
∗∗ Then (dh)Y [ is an element of g , which is naturally isomorphic to g via the double dual isomor- phism.
• H, the jacobian of h:
H : g −→ End(g) (3.15) H(Y ): g −→ g Y 7−→ . W 7−→ (du)Y (W )
• ϕh, the map associated to h:
∗ ∗ ϕh : T G −→ T G (3.16) ϕh(α): Tπ(α)G −→ R α 7−→ , [ Y 7−→ (dh)α(Y )
[ ∗ ∗ where Y ∈ Tπ(α)G ⊂ Tα(T G).
Theorem 3.2.4 (Kirwin, Mourao˜ and Nunes, [12]). Let G be a compact Lie Group, equipped with a bi-invariant Riemannian metric h·, ·i. Let h: T ∗G −→ R satisfy:
∗ ∗ • h is left invariant: h((dLg) α) = h(α) ∀α ∈ T G, g ∈ G;
34 ∗ ∗ • h is right invariant: h((dRg) α) = h(α) ∀α ∈ T G, g ∈ G;
• H is positive definite everywhere: hW, H(Y )W i > 0 ∀Y,W ∈ g; kH(Y )Zk • The norm of H has a greater than 0 lower bound: inf sup kZk > 0. Y ∈g Z∈g
Then,
(i) ϕh is a diffeomorphism.
τ ∗ ∗ τ (ii) Define J = (ψτ ◦ φ ◦ ϕh) JG . If Im τ > 0 then (T G, ω, J ) is Kahler.¨ C
3.2.4 Problems to study
We now pose the questions that we will investigate in this chapter. Question 1: In the case of a compact Lie group with bi-invariant Riemannian metric, what is the relation between the Kahler¨ structures of theorem 3.2.2 (which may only be defined on a tube) and of theorem 3.2.4? Question 2: In the case where M = G is a Lie group and the Riemannian metric is bi-invariant, can we strengthen the result to R = ∞? In theorem 3.2.4, the condition that h is left invariant means that when we consider h to be a function h: G × g −→ R, it does not depend on the point of G, that is, it is a function h: g −→ R. The condition that h is right invariant means that h: g −→ R is Ad-invariant:
h(AdgY ) = h(Y ) ∀g ∈ G, Y ∈ g. (3.17)
Question 3: In theorem 3.2.4, is the condition of Ad-invariance necessary?
3.3 Useful coordinates based on the diffeomorphism T ∗G ∼= G × g
In this section we introduce a basis of vector fields and of forms in G × g, that we will use to study the questions posed in the previous section. Let:
1 n •{ T1, ..., Tn} be a h·, ·i-orthonormal basis of g, with associated coordinates {y , ..., y };
•{ X1, ..., Xn} be the left invariant vector fields on G such that (Xj)e = Tj;
1 n j j •{ ω , ..., ω } be the basis of differential forms in G that is dual to {X1, ..., Xn}, that is ω (Xk) = δk.
Then,
∂ ∂ T (G × g) = span{X , ..., X } ⊕ span , ..., , (3.18) (g,Y ) 1 n ∂y1 ∂yn ∗ 1 n 1 n T(g,Y )(G × g) = span{ω , ..., ω } ⊕ span dy , ..., dy . (3.19)
35 Proposition 3.3.1. In the coordinates that we have just described,
n X θ = yjωj, (3.20) j=1 n n X 1 X ω = = −dθ = ωj ∧ dyj + Cj yjωk ∧ ωl . (3.21) 2 kl j=1 k,l=1
Proof. Recall that from example 1.1.6, θ is given by:
n ! n !! X ∂ X ∂ θ akX + bk = (dL Y )[ dπ akX + bk . (3.22) (x,Y ) k ∂yk x k ∂yk k=1 k=1 As a consequence of the metric being left invariant,
[ ∗ [ (dLxY ) = (dLx−1 ) Y . (3.23)
Therefore,
n ! n !! X ∂ X ∂ θ akX + bk = (dL Y )[ dπ akX + bk (x,Y ) k ∂yk x k ∂yk k=1 k=1 n ! n ! ∗ [ X k [ X k = (dLx−1 ) Y a Xk = Y a Tk k=1 k=1 * n + n X k X j k = Y, a Tk = y a hTj,Tki k=1 j,k=1 n n n ! X X X ∂ = yjaj = yjωj akX + bk , k ∂yk j=1 j=1 k=1
n n X X ω = −dθ = −d yjωj = − dyj ∧ ωj + yjdωj j=1 j=1 n n X 1 X = ωj ∧ dyj + Cj yjωk ∧ ωl . 2 kl j=1 k,l=1
The metric h·, ·i being left invariant means that is is defined by left translation from the metric at the identity.
Proposition 3.3.2. Given that h·, ·i is left invariant, it being right invariant is equivalent to the fact that the metric at the identity is Ad-invariant:
hAdgY, AdgZi = hY,Zi ∀g ∈ G, ∀Y,Z ∈ g. (3.24)
The metric at the identity being Ad-invariant implies that it is ad-invariant:
hadX Y,Zi + hY, adX Zi = 0 ∀X,Y,Z ∈ g. (3.25)
36 Proof. The first statement is a consequence of the fact that Adg = dLgdRg−1 = dRg−1 dLg. Let X,Y,Z ∈ g. Equation (3.24) implies that
hAdexp(tX)Y, Adexp(tX)Zi = hY,Zi.
Differentiating both sides with respect to t and setting t = 0 we obtain equation (3.25).
Proposition 3.3.3. The components of adY in the basis {Tj}j=1,...,n are given by:
j n h i X l j adY = y Clk. (3.26) k l=1
Also, adY is anti-symmetric:
h ij h ik adY = − adY . (3.27) k j
Proof. To prove the first statement, notice that ad is given by Lie brackets, which in turn are determined by the structure constants:
n j n n n ! X h i X l X X l j adY Tj = adY (Tk) = [Y,Tk] = y [Tl,Tk] = y Clk Tj. k j=1 l=1 j=1 l=1
This proves the first statement. The second one is a consequence of h·, ·i being ad-invariant:
h ij h ik adY = hadY Tj,Tki = −hTj, adY Tki = − adY . k j
Consider h, u, H and ϕh as in definition 3.2.3. We now explain how these maps look like in the coordinates and basis that we set in the beginning of this section.
Proposition 3.3.4. Let h, u, H and ϕh be as in definition 3.2.3. Then:
• u is given in coordinates by
∂h ∂h u(y1, ..., yn) = , ..., , (3.28) ∂y1 Y ∂yn Y
which means that hu(Y ), ·i = (dh)Y , where h is seen as a function h: g −→ R.
• The matrix that represents H(Y ) in the basis {T1, ..., Tn} is:
2 2 ∂ h ∂ h ∂y1∂y1 ··· ∂y1∂yn Y Y . . . H(Y ) = . .. . . (3.29) 2 2 ∂ h ∂ h ∂yn∂y1 ··· ∂yn∂yn Y Y
• When seen as a map G × g −→ G × g, ϕh is given by:
ϕh(x, Y ) = (x, u(Y )). (3.30)
37 dL ×id ∼ x−∼1 • Because of the diffeomorphism T(x,Y )(G × g) = TxG ⊕ TY g = g ⊕ g, the derivative of ϕh can be seen as an endomorphism of g ⊕ g. Under this identification, it is given by:
I 0 (dϕh)(x,Y ) = . (3.31) 0 H(Y )
Proof. Proof of the coordinate expression for u: In the basis {T1, ..., Tn},
n X k u(Y ) = (u(Y )) Tk. (3.32) k=1
k ∂h We want to prove that (u(Y )) = k . ∂y Y
k k j (u(Y )) = dy (u(Y )) (y are coord. in the basis {Tj})
k = (dh)Y [ (dy ) (def. of u + double dual iso.) d = h(Y [ + tdyk) dt t=0 d ∞ ∗ ∞ = h(Y + tTk) (h ∈ C (T G) to h ∈ C (g)) dt t=0 ∂h = . ∂yk Y
Proof of the matrix expression for H:
n j n 2 X ∂u k X ∂ h k (du)Y (W ) = dy (W )Tj = w Tj (3.33) ∂yk Y ∂yj∂yk Y j,k=1 j,k=1
∗ Proof of the expression for ϕh : G × g −→ G × g: Let α ∈ T G.