Geometry and Topology of Complex Manifolds - Lecture 1

January 9, 2025

This is the first post in a series on the content from the course taught Winter 2025 at Stanford University on Math 270: Geometry and Topology of Complex Manifolds by Professor Eleny Ionel. The assumed background is the content of Math 215ABC, i.e. the basics of algebraic topology, differential topology, and Riemannian geometry.

The topics we plan to cover include the following.

  • Complex and Kähler Manifolds
  • Complex vector bundles, holomorphic vector bundles (connections and curvature)
  • Dolbeault cohomology, Hodge theorem, Lefschetz theorem
  • Several vanishing theorems (e.g. Kodaira vanishing)
  • Kähler-Einstein metrics on manifolds and Hermitian-Einstein metrics on holomorphic vector bundles.

Second Perspective on Complex Manifolds

One might be able to guess the usual way of defining complex manifolds, which is the usual manifold axioms along with local charts in Cn\bbc^n such that the transition maps are holomorphic, blah blah blah. We will introduce a second perspective on complex manifolds that will prove useful.

Linear Algebra Motivation

We make the following observation. A complex vector space VV is equivalent to a real vector space VV along with some JEndR(V)J \in \End_{\bbr}(V) such that J2=IdVJ^2 = - \id_V, called a complex structure. Given a complex vector space VV, we can let JJ be multiplication by ii. Given (V,J)(V,J) as above, we can get a natural C\bbc-structure on VV by letting

(a+bi)v=av+bJv(a+bi)v = av + bJv

for all a,bRa,b \in \bbr and vVv \in V.

Example Let M=CnM = \bbc^n, pMp \in M and V=TpMV = T_p M with local coordinates z=(z1,,zn)z = (z_1, \ldots, z_n). Then each zk=xk+iykz_k = x_k + i y_k where the x,yx,y are real-valued coordinates, and so TpM=Rn×RnT_p M = \bbr^n \times \bbr^n with the local coordinates

{x1,,xn,y1,,yn}\left\{\frac{\del}{\del x_1}, \ldots, \frac{\del}{\del x_n}, \frac{\del}{\del y_1}, \ldots, \frac{\del}{\del y_n}\right\}

Define the following complex structure on TpMT_pM

J=(0InIn0)End(V)J = \begin{pmatrix}0 & I_n \\ -I_n & 0 \end{pmatrix} \in \End(V)

Unimportant Remark Note that JJ satisfies the following equations which resemble a PDE.

Jxk=ykJyk=xkJ \frac{\del}{\del x_k} = \frac{\del}{\del y_k} \qquad J \frac{\del}{\del y_k} = -\frac{\del}{\del x_k}

We will see later that on manifolds whose tangent bundle has a complex structure JJ, we can always find coordinates around a point such that JJ takes the above canonical form and satisfies these equations. This might not be true in a neighborhood of pp, but if it is then JJ is said to be integrable.

Ignore this for now, but note that this is the reason for the terminology almost complex structure below. If a manifold has an almost complex structure and is integrable, then it has a complex structure (i.e. it is a complex manifold).

Important Remark Note that this extends naturally to families, i.e. to vector bundles, by defining everything fiberwise. That is, a complex vector bundle E\cale over a manifold MM is a real vector bundle E\cale over MM together with a JEnd(E)J \in \End(\cale) such that for every pMp \in M, the map Jp ⁣:EpEpJ_p \colon \cale_p \to \cale_p on the fibers satisfies Jp2=IdJ_p^2 = -\id. Because the notation will get annoying, for vEpv \in \cale_p an element of the fiber over pMp \in M we write JvJv instead of JpvJ_p v.

Almost Complex Structure

Definition An almost complex manifold (ACM) is a manifold MM with an almost complex structure (ACS), i.e. a complex structure JEnd(TM)J \in \End(TM) on the tangent bundle TMTM.

For example, if MM is a complex manifold then multiplying by ii in each local holomorphic chart induces a well-defined almost complex structure on MM. The details are easily obtained from the M=CnM = \bbc^n example in the previous subsection.

Remark Suppose MM is a complex manifold, JJ an ACS on MM, and f ⁣:MCf \colon M \to \bbc a function. Then the following are equivalent.

  1. ff is holomorphic
  2. df ⁣:TMTCdf \colon TM \to T\bbc is complex linear, i.e. dfJ=J0dfdf \circ J = J_0 \circ df where J0J_0 is the canonical ACS on TCT \bbc
  3. f=u+ivf = u + iv are the real and imaginary components, duJ0=dvdu \circ J_0 = dv
  4. The u,vu,v above satisfy the Cauchy-Riemann equations in each local chart, i.e.
uxk=vykuyk=vxk\frac{\del u}{\del x_k} = \frac{\del v}{\del y_k} \qquad \frac{\del u}{\del y_k} = -\frac{\del v}{\del x_k}

Differential Calculus on Almost Complex Manifolds

Note that JEnd(TM)J \in \End(TM) being such that J2=IdJ^2 = -\id on the fibers implies that JJ has eigenvalues ±i\pm i. Considering the eignspaces will allow us to decompose almost complex manifolds into direct sums, which will prove useful. I'll expand on this with some motivation from linear algebra first.

Linear Algebra Motivation

Let VV be a real vector space. We can construct the complexification of VV by taking VC=VRCV^{\bbc} = V \otimes_{\bbr} \bbc. We denote elements of VCV^{\bbc} by v+iwv + iw, which is shorthand for v1+wiv \otimes 1 + w \otimes i. This is naturally a complex vector space.

Let (V,J)(V,J) be a complex vector space, i.e. VV is a real vector space and JJ is a complex structure on VV. Note then that JJ extends to VCV^{\bbc} by J(v+iw)=Jv+iJwJ(v + iw) = Jv + i Jw. As an aside, we denote the conjugate (V,J)(V,-J) by V\bar{V}.

Since JJ has two eigenvalues on VCV^{\bbc}, namely ±i\pm i, there are two eigenspaces

V1,0={vJv=iv}={viJvvV}V^{1,0} = \{v \mid Jv = iv\} = \{v - i Jv \mid v \in V\} V0,1={vJv=iv}={v+iJvvV}V^{0,1} = \{v \mid Jv = -iv\} = \{v + i J v \mid v \in V\}

corresponding to the eigenvalues ii and i-i of JJ, respectively. Note that V1,0=V0,1V^{1,0} = \bar{V^{0,1}} are conjugate.

Important There is a natural decomposition of VCV^{\bbc} into a direct sum of its eigenspaces.

VC=V1,0V1,0V^{\bbc} = V^{1,0} \oplus V^{1,0}

This again extends in families, i.e. to vector bundles by applying the definitions fiberwise. That is, if (E,J)(\cale, J) is a complex vector bundle then we can extend EC=ERC\cale^{\bbc} = \cale \otimes_{\bbr} \bbc and JJ extends on the fibers EpC\cale^{\bbc}_p as before. We then have a decomposition

EC=E1,0E0,1\cale^{\bbc} = \cale^{1,0} \oplus \cale^{0,1}

where

E1,0={(p,viJv)pM,vEp}\cale^{1,0} = \{(p, v - i J v) \mid p \in M, v \in \cale_p\} E0,1={(p,v+iJv)pM,vEp}\cale^{0,1} = \{(p, v + i J v) \mid p \in M, v \in \cale_p\}

are the eigenbundles corresponding to ii and i-i, respectively.

Holomorphic Tangent Bundle

We can apply the above motivation to the setting of a manifold MM with an almost complex structure JJ. Note (TM,J)(TM, J) is a complex vector bundle, so we can complexify it to TMC=TMRCTM^{\bbc} = TM \otimes_{\bbr} \bbc. As above, we have a decomposition

TMC=T1,0MT0,1MTM^{\bbc} = T^{1,0} M \oplus T^{0,1} M

Since the notation will get particularly bad otherwise, let us denote the cotangent bundle TMT^*M by ΩM\Omega M.

Given a complex structure JJ on TMTM, we can obtain a complex structure JJ^* on ΩM\Omega M by Jω=ωJJ^* \omega = -\omega \circ J for ωΩpM=TpM\omega \in \Omega_p M = T^*_p M. The negative sign is not necessary but its a useful convention.

Then (ΩM,J)(\Omega M, J^*) is a complex vector bundle, so we can complexify it just like (TM,J)(TM, J). Let

ΩMC= space of C-valued functions =(TM)C\Omega M^{\bbc} = \text{ space of } \bbc \text{-valued functions } = (T^* M)^{\bbc}

We really want to think of this as the space of complex-valued differential 1-forms on MM. As before,

ΩMC=Ω1,0MΩ0,1M\Omega M^{\bbc} = \Omega^{1,0} M \oplus \Omega^{0,1}M

Let's unravel the definitions here real quick.

Ω1,0M={(p,ωiJω)pM,ωΩpM}={(p,ω+iωJ)pM,ωΩpM}\begin{align*} \Omega^{1,0} M &= \{(p, \omega - i J^* \omega) \mid p \in M, \omega \in \Omega_p M\} \\ &= \{(p, \omega + i \omega \circ J ) \mid p \in M, \omega \in \Omega_p M\} \end{align*} Ω0,1M={(p,ω+iJω)pM,ωΩpM}={(p,ωiωJ)pM,ωΩpM}\begin{align*} \Omega^{0,1} M &= \{(p, \omega + i J^* \omega) \mid p \in M, \omega \in \Omega_p M\} \\ &= \{(p, \omega - i \omega \circ J ) \mid p \in M, \omega \in \Omega_p M\} \end{align*}

Let us extend this to the space of C\bbc-valued differential kk-forms ΩkMC\Omega^k M^{\bbc}. Recall ΩkM=kΩM\Omega^k M = \bigwedge^k \Omega M and naturally we define ΩkMC=ΩkMRC\Omega^k M^{\bbc} = \Omega^k M \otimes_{\bbr} \bbc. Since complexifying commutes with the wedge product, we have

ΩkMC=(kΩM)C=kΩMC=k(Ω1,0MΩ0,1M)=p+q=k(pΩ1,0MqΩ0,1M)=p+q=kΩp,qM\begin{align*} \Omega^k M^{\bbc} &= \left(\wedge^k \Omega M\right)^{\bbc} \\ &= \wedge^k \Omega M^{\bbc} \\ &= \wedge^k \left(\Omega^{1,0} M \oplus \Omega^{0,1} M\right) \\ &= \oplus_{p + q = k} \left(\wedge^p\Omega^{1,0} M \otimes \wedge^q\Omega^{0,1} M \right) \\ &= \oplus_{p+q = k} \Omega^{p,q} M \end{align*}

where Ωp,qM\Omega^{p,q} M is called the space of complex-valued differential forms of type (p,q)(p,q) on MM.

Ωp,qM=Ω1,0MΩ1,0Mp timesΩ0,1MΩ0,1Mq times\Omega^{p,q} M = \underbrace{\Omega^{1,0} M \wedge \ldots \wedge \Omega^{1,0} M}_{p \text{ times}} \otimes \underbrace{\Omega^{0,1} M \wedge \ldots \wedge \Omega^{0,1} M}_{q \text{ times}}

Let's denote AkM=Γ(ΩkMC)A^k M = \Gamma(\Omega^k M^{\bbc}) by the space of smooth sections of complex-valued differential kk-forms on MM, and similarly Ap,q=Γ(Ωp,qM)A^{p,q} = \Gamma(\Omega^{p,q} M) for (p,q)(p,q)-forms.

Definition If MM is a complex manifold, we have the following names for the spaces defined above.

  • T1,0MT^{1,0}M is called the holomorphic tangent bundle,
  • T0,1MT^{0,1}M is the anti-holomorphic tangent bundle,
  • Ω1,0M\Omega^{1,0} M is the holomorphic cotangent bundle, and, lastly
  • Ω0,1M\Omega^{0,1} M is the anti-holomorphic cotangent bundle.

We want to be able to do calculations, so it helps to know what these spaces look like locally. Let MM be a complex manifold with local coordinate z=(z1,,zn)z = (z_1, \ldots, z_n). A complex-valued 11-form can be written locally uniquely as

η=iηidziholomorphic part+jηjdzjanti-holomorphic part\eta = \underbrace{\sum_{i} \eta_i dz_i}_{\text{holomorphic part}} + \underbrace{\sum_j \eta_j d\bar{z}_j}_{\text{anti-holomorphic part}}

where ηi,ηjC(M,C)\eta_i, \eta_j \in C^{\infty}(M, \bbc). Similarly, a complex-valued (p,q)(p,q) form can be written locally uniquely as

η=I,JηI,JdzIdzJ\eta = \sum_{I, J} \eta_{I, J} dz_I \wedge d\bar{z}_J

where II is the multiindex i1<<ipi_1 < \ldots < i_p (and JJ is j1<<jqj_1 < \ldots < j_q), and where dzI=dzi1dzipdz_I = dz_{i_1} \wedge \ldots \wedge dz_{i_p} (and dzJ=dzj1dzjqd\bar{z}_J = d\bar{z}_{j_1} \wedge \ldots \wedge d\bar{z}_{j_q}). As before, ηI,JC(M,C)\eta_{I,J} \in C^{\infty}(M, \bbc).

In the coming lectures we will expand on this. Namely, we can consider what happens for example when ηI,J\eta_{I,J} take values in spaces other than C\bbc, or are holomorphic rather than smooth. We will also define the Dolbeault complex consisting of sections Ap,qA^{p,q} which leads us naturally into the topic of Dolbeault cohomology.