Spring Semester 2018
Math 7721: Kähler Geometry
[KG] refers to the
.pdf notes on Kähler geometry,
[DG] to the
.pdf notes on differential geometry, [FR] to further references.
[Zhengs book] to Professor Zhengs
Differential Geometry.
- JANUARY 8. Almost-complex manifolds. Examples:
complex vector spaces, open submanifolds, Cartesian products. The covariant
2-tensors aJ and Ja arising from
a given covariant 2-tensor a on an almost-complex
manifold. Hermitian (symmetric) 2-tensors a and
skew-Hermitian 2-forms a, defined by requiring that
(1.1)
aJ = Ja.
Hermitian metrics on an almost-complex manifold, characterized, equivalently,
by being those metrics which make J skew-adjoint at each
point, as well as by being the metrics turning J, at
each point, into a linear isometry of the tangent space. The fact that,
when one uses a Hermitian metric g to identify any
covariant 2-tensor a with the endomorphism
A of the tangent bundle TM such that
a(v, ⋅ )
= g(Av, ⋅ ) for all
tangent vector fields v, the 2-tensors
aJ and Ja correspond in the same way
to the composite bundle endomorphisms AJ and
JA. The one-to-one J-correspondence between
Hermitian 2-tensors and skew-Hermitian 2-forms. The Kähler form
Ω = gJ of the
given Hermitian metric g on an almost-complex manifold.
The equality
(1.2)
Ω ∧m = m! dg, where
m = dimCM.
[FR]. [KG]:
Section 3 except formula (3.2), and Remark 4.1(iii) except the last
sentence. Homework #1.
- JANUARY 10. Holomorphic mappings, biholomorphisms.
Integrability of J. Complex manifolds. Kähler
manifolds, examples: open submanifolds, complex vector spaces with Hermitian
inner products, oriented Riemannian surfaces (since an oriented Euclidean
plane is, naturally, a complex line), products of Kähler manifolds.
Almost complex submanifolds, examples: open submanifolds and nonempty
preimages of points under constant-rank holomorphic mappings (including
holomorphic submersions). The Levi-Civita connection of a submanifold metric.
The fact that almost complex submanifolds of Kähler manifolds become
Kähler manifolds when endowed with the submanifold metric. Nonsingular
projective algebraic varieties as examples of Kähler manifolds.
[FR]. [KG]: Section 7 (the first four
lines and the final paragraph), Section 4 (the first three paragraphs).
Homework #2.
- JANUARY 12. Almost-Kähler metrics, including Kähler metrics as a special
case. Finite partitions of unity
[DG
p. 121]. Oriented integration of compactly supported
continuous top degree differential forms
[DG
p. 165]. The Stokes theorem
[DG
p. 166]. The Kähler form
Ω = gJ and the Kähler
class
(3.1) [Ω] ∊ H2(M, R)
of an almost-Kähler manifold (M, g).
Positive and negative cohomology classes in
H2(M, R) on an almost
complex manifold M, and the fact that, by (1.2), positivity
implies being nonzero. Mutual exclusiveness of positivity/negativity/vanishing
in H2(M, R) for a compact
almost complex manifold M, leading to the conclusion that
H2(M, R) ≠ {0}
for a compact almost Kähler manifold
(M, g). An example of a compact complex manifold
M with
H2(M, R) = {0},
provided by a Hopf manifold, that is,
S1 × S2m−1 with
the atlas consisting of the local inverses of the following locally
diffeomorphic surjective mapping (with any fixed real constant
θ ≠ 0):
(3.2) Cm − {0} ∋ x → (e iθ log |x| , x ⁄ |x|) ∊ S1 × S2m−1 .
[FR]. [KG]:
Section 5 (the first paragraph, Remark 5.1, the 4-line paragraph
preceding Theorem 5.3, and the text following the proof of
Theorem 5.3). Homework #3.
- JANUARY 17. The Betti numbers of spheres and the Künneth formula for
S1 × M
[DG
p. 164], derived from the Mayer-Vietoris sequence
[DG
p. 162]. The curvature R of a connection in
a real/vomplex vector bundle
[DG
p. 77]. Existence of connections (and fibre metrics) via partitions of unity
[DG
p. 121]. The first Chern class of a complex line bundle defined with the aid
of a connection and its independence of the connection. The first Chern class
for arbitrary complex vector bundles, and
c1(M) for
an almost complex manifold M. Locally symmetric Kähler
manifolds, and a proof of the fact that they are actually Kähler
manifolds. [FR].
[KG]: Section 2 and, in Section 7, the
paragraph immediately following Remark 7.2.
Homework #4.
- JANUARY 19. Examples of locally symmetric Kähler
manifolds: complex vector spaces with constant metrics, open submanifolds,
complex projective spaces with the Fubini-Study metrics, and complex
hyperbolic spaces. [FR].
[KG]:
Section 7 (the second paragraph following Remark 7.2).
Homework #5.
- JANUARY 22. The curvature tensor R
and Ricci tensor r of a connection in the tangent
bundle
[DG
p. 80]. The first Bianchi identity
[DG
p. 94]. The skew-adjointness property of the
curvature for a connection in a vector bundle compatible with a fibre metric
[DG
p. 103], symmetry of the Ricci tensor r of a
Riemannian manifold
[DG
p. 108]. Index raising and lowering
[DG
p. 105]. The scalar curvature s
[DG
p. 108]. The second Bianchi identity
[DG
pp. 94-95]. The additional symmetry of the Riemannian curvature tensor due to the
fact that the conditions
(6.1) abcd = −bacd = −abdc and
abcd + bcad + cabd = 0
imply abcd = cdab, as one sees
noting that
abcd = −abdc = badc, and hence [DG
p. 108], as illustrated by Milnor's octahedron,
2cdba = (cdba + dcab) = −(bcda + dbca) − (cadb + adcb) = −(cbad + bdac) − (acbd + dabc)
= −(cbad + acbd) − (bdac + dabc) = bacd + abdc = 2bacd.
[FR].
Homework #6.
- JANUARY 24. The relations
[DG
p. 110]
(7.1)
R(u, v)w =
K [g(u, w)v
− g(v, w)u]
and r = Kg
in dimension 2,
where
K = s ⁄ 2.
Einstein metrics and metrics of constant (sectional) curvature
[DG
p. 130]. Complex-linearity of
R(v,w) for Kähler
manifolds. The equality
(7.2) trR J[R(v,w)] = −2 ρ(v,w)
in a Kähler manifold, implying, for the Ricci tensor
r of any Kähler manifold,
(7.3) Hermitian symmetry of r and closedness (dρ = 0) of the
Ricci form ρ = rJ.
Proof of (7.2) based on the identity
(7.4) ρkl = Rskl qJqs.
The conclusion that
(7.5) trC[R(v,w)] = i ρ(v,w),
since 2 trC iA =
i trR A for any
self-adjoint endomorphism A of a finite-dimensional
complex inner-product space (as one sees using a basis formed by eigenvectors
of A). The cohomology relation
(7.6) [ρ] = 2π c1(M)
for the Ricci form ρ of any Kähler
metric.
[KG]: Section 4 (the two lines preceding formula (4.1) plus the 13 lines following it, and Remark 4.1); Section 6
(parts (i), (ii) of Remark 6.1). [FR].
Homework #7.
- JANUARY 26. Examples of Einstein manifolds: vector
spaces with constant metrics (which are flat, hence Ricci-flat); suitable
products; Riemannian surfaces of constant Gaussian curvature. The Einstein
condition as a consequence of irreducibility of the isotropy representation,
further examples of Einstein manifolds (complex projective spaces with the
Fubini-Study metrics, complex hyperbolic spaces, and standard spheres).
Kähler-Einstein metrics and
consequences for the first Chern class, including, in the compact case,
nonnegativity of the integral of its top cup power when the complex dimension
is even. Certain products of closed oriented Riemannian surfaces as examples
showing that a Kähler-Einstein metric may fail to exist despite the
existence of a Kähler metric. The Calabi conjecture (for
c1 < 0, proved independently
by Aubin and Yau):
every compact
almost complex manifold M with
c1(M) < 0,
admitting a Kähler
metric, also admits a Kähler-Einstein metric .
[FR].
[KG]: Section 7 (the third paragraph
following Remark 7.2).
Homework #8.
- JANUARY 29. The Calabi conjecture (for
c1 = 0, proved by Yau):
if
ρ is a closed skew-Hermitian
2-form on a compact almost complex mani-
fold
M admitting a Kähler metric, and
[ρ] = 2π c1(M), then every positive
cohomology class in
H2(M, R) contains the
Kähler form of a Kähler metric
for which
ρ is the Ricci form .
The Goldberg conjecture (still open):
a compact
almost-Kähler Einstein manifold
is necessarily a
Kähler manifold .
The contraction-derived consequences of the
second Bianchi identity [DG
pp. 126-127]:
(9.1) δR = − dr, 2δr = ds,
the latter known as the Bianchi identity for the Ricci tensor. Schur's
lemma
[DG
p. 131]:
if
r = sg ⁄ n on a
connected Riemannian manifold
(M, g) of dimen-
sion n > 2, then
s is constant, that is,
g is an Einstein metric .
The second covariant derivative
∇df of a function f on a
manifold with a connection ∇, also known as the Hessian of
f [DG
p. 89], the coordinate description
(9.2) f, jk = ∂k ∂j f − Γkjl ∂l f
of ∇df and the formula
(9.3) [∇u(df )]v = du dv f − dw f with w = ∇uv,
for any vector fields u and v, where
[∇df ](u,v) = [∇u(df )](v). Symmetry of the Hessian if ∇ is torsionfree. The
gradient of a smooth function on a Riemannian manifold
[DG
p. 106]. A description of the
Hessian in terms of gradient (namely, ∇df
corresponds via g to ∇w for
w = ∇f ). The Laplacian and its
representation as a composition: gradient followed by divergence [DG
p. 125]. The observation that, for any smooth function f
on a Riemannian manifold,
(9.4) ∇df and Δ f are nonnegative / nonpositive at any local minimum / maximum of f.
The operator
i∂∂ on an almost complex manifold
M, sending smooth functions
f : M → R to exact 2-forms, and given by
(9.5) 2i∂∂f = − d [(df )J].
The fact (see Problem 1 in Homework #9) that for
any fixed torsionfree connection ∇ in
TM (not assumed to be in any way related to
J ), and any vector fields
u, v, one then has
(9.6) 2 [i∂∂f ](Ju,v) = − [∇df ]( Ju, Jv) − [∇df ](v, u) − dw f,
where, this time, w = [∇Ju J]v + J [∇v J]u. The conclusion that, if
∇J = 0 (for instance, if the almost
complex manifold M admits a Kähler metric), then,
for every smooth function f, the 2-form
2i∂∂f is skew-Hermitian (as
[i∂∂f ](Ju,v) is symmetric
in u and v). The equality (immediate when
one replaces u in (9.6) with −Ju):
(9.7) 2i∂∂f = (∇df )J + J(∇df ),
for smooth functions f on Kähler manifolds.
[FR]. [KG]:
formula (3.2) in Section 3 and formula (4.3.i) in Section 4.
Homework #9.
- JANUARY 31. The Ricci identity
(∇u[∇ψ])w − (∇w[∇ψ])u = R∇(w,u) ψ, that is,
(10.1) ψa, jk − ψa, kj = Rjkca ψc,
for sections ψ of a vector bundle with an arbitrary
connection ∇ over a manifold M and
vector fields u, w, where any fixed torsionfree
connection ∇ in TM is used to form the
second covariant derivative. The divergence δA of a
bundle morphism
A:TM → TM. The formula
(10.2) Jqpwk ,qp = ρlkwl, that
is, δ[J(∇w)*] = ρ(w, ⋅ ),
easily obtained from (10.1) and (7.2), and valid for all vector fields
w in all Kähler manifolds. The identities, with
A = ∇w : TM → TM,
(10.3) tr JAJA = (tr JA)2 − r(w,w) + δ[JAJw − (tr JA)Jw]
and
(10.4) tr JAJA* = δ(JA*Jw) − r(w,w),
satisfied by any vector field w in any Kähler
manifold, and derived from (10.1) - (10.2).
The μ-divergence
operator δ acting on vector fields, associated
with a fixed positive volume form μ on an oriented
n-dimensional manifold, and given by the local-coordinate formula
(10.5) μ1. . . n δw = ∂j (w jμ1. . . n).
The fact that, if μ = dg is the
volume form of a Riemannian metric g, then
[DG p. 125, Theorem 38.1] for the Levi-Civita connection
∇ of g and any vector field
w,
(10.6) δw = tr ∇w.
The equalities [DG pp. 124-125], used to establish (10.6):
(10.7) μ1. . . n = √ det g ,
gkl
∂j gkl = ∂j det g,
Γkjk = ∂j log√ det g
for det g = det [gkl] and μ = dg.
[FR]. [KG]:
formulae (4.1.b) and (4.2) in Section 4, plus formula (6.1) in Section 6.
Homework #10.
- FEBRUARY 2.
The divergence theorem
[DG
p. 127] and integration by parts. The fact that, for smooth functions
θ, f on a compact oriented Riemannian manifold,
(11.1) (f, Δθ) = − (∇f, ∇θ) = (Δf, θ), (θ, Δθ) = − ||∇θ||2,
where ( , ) and || || denote the
L2 inner product and
L2 norm, leading to Bochners lemma
[DG
p. 128]:
(11.2) if Δθ ≥ 0, then θ must be constant.
The kernel and image of the Laplacian of a compact oriented Riemannian
manifold. The relation
(11.3) trg [i∂∂θ]J = −Δθ
for any smooth function θ on a Kähler manifold. The
∂∂
Lemma:
any exact
skew-Hermitian 2-form on a compact connected Kähler manifold
equals
i∂∂θ for some
θ : M → R, unique up to an additive constant .
[FR]. [KG]:
formulae (4.3.ii), (4.4) and Lemma 4.2 in Section 4.
Homework #11.
- FEBRUARY 5. The
equality of L2 norms:
(12.1)
√2 || ζ || = || trg ζJ ||
for any exact skew-Hermitian 2-form ζ on a compact
Kähler manifold, and the resulting injectivity of the assignment sending
ζ to
trg ζJ. Proof of
the ∂∂
Lemma based on (12.1) combined with the fact that, due to (11.3), the
assignment just mentioned sends
ζ − i∂∂θ to 0 for a function θ with
Δθ = − trg ζJ. The ratio
γ : M → R of the
volume forms of two Riemannian metrics g and
g′ on an oriented manifold M,
defined by the relation
dg′ = γ dg.
The equality
(12.2) ρ′ = ρ − i∂∂ log γ
for the Ricci forms of two Kähler metrics on an almost complex manifold.
The conclusion that the Ricci form (and Ricci tensor) of a Kähler metric
is uniquely determined by its volume form. First part of the proof of (12.2).
[FR].
[KG]: Section 4 (Lemma 4.2 and its
proof), Section 6 (Remark 6.1(iii)).
Homework #12.
- FEBRUARY 7. The remainder of the proof of (12.2), based
on the coordinate-free version
(13.1) detg g′ = γ2
of the coordinate formula
det g′ = γ2 det g. The uniqueness assertions in the Calabi
conjectures, both due to Calabi:
one necessarily
has g = g′ whenever g, g′ are two Kähler
metrics
on a compact almost
complex manifold, such that
ρ = −Ω and ρ′ = −Ω ′,
or
ρ = ρ′ and [Ω] = [Ω ′] .
Proof of the first uniqueness assertion.
[FR].
[KG]: Section 6 (formula (6.2.a) and
Remark 6.1(iii), plus their justifications, and Theorem 6.2).
w Homework #13.
- FEBRUARY 9. Proof of the second uniqueness assertion.
Tensor fields Z on a manifold M,
including sections θ, f of
M × R (that is, functions on
M), sections w,u of
TM (vector fields), ξ of
T*M (1-forms), A, B of
Hom(TM, TM) (bundle
endomorphisms of TM) and, finally, sections
a of Hom(TM, T*M)
(2-tensors, also called twice-covariant tensors). Natural multiplications of
tensor fields. The push-forward [dF]Z of a
tensor field Z on M under a
diffeomorphism
F:M → N between
two manifolds, and the equality
[dF](ZZ′)
= ([dF]Z)([dF]Z′) ,
where ZZ′ stands for any natural
multiplication. The (local) flow
e tw of a vector field
w on a manifold M
[DG
pp. 219-221]. Completeness of a vector field. The Lie derivative
£w Z of a tensor
field Z in the direction of a vector field
w on a manifold, defined to be the derivative with
respect to the real variable t, at
t = 0, of
[de−tw ]Z.
The immediate conclusion that
[de−sw ]£w Z equals the derivative of
[de−tw ]Z
at t = s, and so
(14.1) £w Z = 0 if and
only if the local flow of w leaves
Z invariant.
The Leibniz rule
(14.2) £w (ZZ′)
= (£w Z)Z′ + Z(£w Z′).
The relation
(14.3) £w θ = dw θ
for any smooth function θ. Proofs of (14.2) - (14.3).
[FR].
[KG]: Section 6 (proof of
Theorem 6.2). Homework #14.
- FEBRUARY 12. The identities stating that
£w u, (£w A)u, (£w a)(u, v) are,
respectively, equal to
(15.1) [w, u], [w, Au] − A[w, u], dw [a(u, v)] − a([w, u], v) − a(u, [w, v]),
while
(£w ξ)(u) = dw [ξ(u)] − ξ([w, u])
for vector fields u,v,w, endomorphisms A
of the tangent bundle, twice-covariant tensor fields a, and
1-forms ξ. Proof of (15.1). Holomorphic vector
fields on an almost complex manifold, defined to be those smooth vector fields
w for which
£w J = 0, that
is, the operators J and
[w, ⋅ ]
commute when both treated as endomorphisms of the space of all smooth vector
fields. The equality
(15.2) £w J = [J,∇w]
(commutator of bundle morphisms
TM → TM), for any smooth vector
field on a Kähler manifold, and the resulting the characterization of
holomorphic vector fields w on a Kähler manifold by
the condition [J,∇w] = 0.
Killing vector fields w on a Riemannian manifold
(M, g), defined by requiring that
£w g = 0. The
formula
(15.3) (£w g)(u, v) = g(Bu, v), where B = ∇w + (∇w)*
and u,v,w are any smooth vector fields on a Riemannian
manifold (M, g).
The characterization of Killing fields w by
skew-adjointness of ∇w.
[FR].
[KG]: Section 7 (the three lines
immediately preceding Remark 7.1, parts (a) - (c) of Remark 7.1, and
Remark 7.2). Homework #15.
- FEBRUARY 14. The relation, satisfied by any smooth
vector fields w, u on any manifold:
(16.1) £[w,u] = £w £u − £u £w.
The proof of (16.1) based on the fact that, for
Ψ = £[w,u] − £w £u + £u £w, and any tensor fields
Z, Z′, one has the Leibniz rule
Ψ(ZZ′) = (ΨZ)Z′ + Z(ΨZ′.
The (real) Lie algebra of vector field on a given manifold whose flows
preserve a given tensor field. The real Lie algebra
i(M, g) of Killing vector fields on a
Riemannian manifold (M, g). The complex Lie
algebra h(M) of holomorphic vector
fields on a Kähler manifold (M, g).
The Bochner identity [DG
p. 126], immediate from (10.1):
(16.2) wk, jk − wk, kj = Rjk wk, that is, δ∇w − dδw = r( ⋅ , w),
for vector fields w on any manifold with a fixed
connection ∇.
The Laplacian
Δ acting on sections of a vector bundle with a connection
over a Riemannian manifold (M, g), with
[Δψ]a = g jk ψa, jk . The symbol
aw for a twice-covariant tensor field
a and a vector field w on a
Riemannian manifold (M, g), denoting
the unique vector field with
g(aw, ⋅ ) = a(w, ⋅ ). The linear differential
operator D in a Riemannian manifold
(M, g), sending vector fields to vector
fields, and given by
(16.3) Dw = − Δw − rw
or, equivalently,
(16.4)
g(Dw, ⋅ ) = − δ£w g + dδw,
the equivalence of the two descriptions being due to the coordinate expression
[Dw] j = − wj, k k − Rjk wk, the Bochner formula (16.2), and (15.3). The fact that, as
an obvious consequence of (16.4),
(16.5) Dw = 0 for any Killing vector
field w on a Riemannian manifold.
The identity, satisfied
by any smooth vector field w on a Kähler manifold:
(16.6) δ[J,∇w]* = − g(JDw, ⋅ ).
Proof of (16.6) in local coordinates: the kth component
of the left-hand side is
gkl [Jplw p, q − wl, p Jqp ], q = gkl Jplw p, qq − Jqpwk, pq while (10.2) combined with
skew-adjointness of J gives Jqpwk, pq = ρkl wl. The resulting trivial conclusion that
(16.7) Dw = 0 whenever w is a holomorphic vector field on a Kähler manifold.
The obvious equality
(16.8) |[J, A]|2 = tr [J, A][J, A]* = 2 tr JAJA* + 2 tr AA*,
for any (real) linear endomorphism of the tangent bundle of an almost complex
manifold. The integral formula for smooth vector fields w
on a compact oriented Riemannian manifold (M, g),
obvious from (16.3):
(16.9) (Dw, w) = ||∇w||2 − ∫ M r(w,w) dg and, consequently, Ker D = {0} if r < 0,
the inequality r < 0 meaning that the Ricci tensor
r is negative definite at each point.
[FR]. [KG]:
parts of Section 9 (formulae (9.1) - (9.2) and (9.5), assertions (i)
and (iii) in Remark 9.1).
Homework #16.
- FEBRUARY 16. The relation, satisfied by any compactly
supported smooth vector field w on a Kähler
manifold:
(17.1) 2(Dw, w) = ||B||2 for B = [J,∇w], that is, B = £w J, cf. (15.3),
and obtained by combining (10.4) with (16.8) applied to
A = ∇w. The conclusion that in any compact Kähler manifold
(M, g) the operator D
is nonnegative, while its kernel consists precisely of all of holomorphic
vector fields:
(17.2) Ker D = h(M).
An obvious consequence of (17.2) and (16.5):
(17.3) on a compact
Kähler manifold every Killing field is holomorphic .
The inner product of twice-covariant tensor
fields a, b on a Riemannian manifold
(M, g), given by
(17.4) < a, b > = aj kb jk, that is, < a, b > = tr AB*
A, B being the bundle endomorphisms of
TM characterized by
g(Aw, ⋅ ) = a(w, ⋅ ) and g(Bw, ⋅ ) = b(w, ⋅ ) for all vector
fields w. The observation that
(17.5) < a, b > = 0 if a is symmetric and b
skew-symmetric.
The integral formula
[DG
pp. 128-129] for compactly-supported smooth vector fields
w and functions f on
an oriented Riemannian manifold (M, g),
also due to Bochner, and immediate from (16.2):
(17.6) ∫ M r(w,w) dg = ||δw||2 − ∫ M tr (∇w)2 dg, so that ∫ M r(∇f, ∇f ) dg = ||Δf ||2 − ||∇df ||2.
The fact that the first equality in (17.6) implies nonexistence of nontrivial
Killing fields (or, harmonic 1-forms) on a compact Riemannian manifold having
negative (or, respectively, positive) Ricci curvature, in the sense that its
Ricci tensor r is negative/positive definite at each point. The
Lichnerowicz inequality [DG
p. 128]:
(17.7) (n − 1)τ ≥ nλ
for any positive eigenvalue τ of
−Δ on a compact oriented n-dimensional
Riemannian manifold satisfying the lower bound
r ≥ λg on the
Ricci curvature with a constant λ. The conclusion
that, by (16.9) and the second equality in (17.6), for a positive eigenvalue
τ of −Δ on a compact oriented
Riemannian manifold (M, g),
(17.8) τ ||w||2 = (Dw, w) + 2 ∫ M r(w,w) dg whenever
w = ∇f and Δf = − τf.
The stronger version of the Lichnerowicz inequality (17.7):
(17.9) τ ≥ 2λ,
obvious from (17.8) along with nonnegativity of D (see
(17.1)), and valid for any positive eigenvalue τ of
−Δ on a compact Kähler manifold such that
r ≥ λg for a
constant λ.
[FR]. [KG]:
Section 9 (Lemma 9.2, formula (9.9), and the first part of Theorem
9.4). Homework #17.
- FEBRUARY 19. The theorem stating that
in a compact
oriented Einstein manifold (M, g)
with nonzero Einstein
constant
λ, the assignment which sends
f to ∇f is a
linear isomorphism
of the space of all
functions
f : M → R with
Δf = − 2λf and the space
p of all gradient
vector fields w on M having Dw = 0 ,
where the fact that the assignment takes values in the required space and is
surjective follows from the following identity, valid for any smooth function
f on any Riemannian manifold, and immediate from (16.2) -
(16.3):
(18.1) D∇f = − ∇Δf − 2r∇f,
so that
(18.2) D∇f = − ∇(Δf + 2λf ) when r = λg.
The version of the
above theorem for Kähler manifolds, which, as a consequence of (17.2),
reads
in a compact
Kähler-Einstein manifold (M, g)
with nonzero Einstein
constant
λ, the assignment sending
f to ∇f is a
linear isomorphism
of the space
Ker (Δ + 2λ) of all
functions
f : M → R
with
Δf = − 2λf
and the space p of all holomorphic gradient
vector fields on
M .
The fact that, for any smooth differential 2-form ζ
on a Riemannian manifold, one has
(18.3) δδζ = 0, wj, k jk = wj, k kj,
where δζ denotes the 1-form
ξ with the components
ξj = ζjk ,k and the divergence
δξ of a 1-form ξ
is, by definition, equal to δw for the vector field
w with
g(w, ⋅ ) = ξ.
A proof of (18.3) based on the following equalities of
L2 inner products, resulting from (17.5) via
integration by parts, with any smooth function f having a
(small) compact support:
0 = (ζ, ∇df ) = −(δζ, df ) = (δδζ, f ). Also,
(18.4) δδζ = 0, wj, k jk = wj, k kj,
for smooth vector fields w on a Riemannian manifold,
which is a special case of (18.3), with
ζjk = wj, k − wk, j.
Consequently,
(18.5) δ[Dw + 2rw] = − Δδw,
since adding 2Rjk wk to the coordinate expression
[Dw] j = − wj, k k − Rjk wk, following (16.4), we get
[Dw + 2rw] j = Rjk wk − wj, k k, so that the Bochner formula
Rjk wk = wk, jk − wk, kj in (16.2) gives
[Dw + 2rw] j = wk, j k − wj, k k − wk, kj , and, therefore,
[Dw + 2rw] j , j = −wk, kj j by (18.4), as required in (18.5). In particular, with a constant
λ, (18.5) yields
(18.6) δw ∊ Ker (Δ + 2λ)
and D∇δw = 0 whenever Dw = 0
and r = λg,
the second conclusion being derived from the first via (18.2) for
f = δw. The
equality
(18.7) 2(Dw, w) = ||£w g||2 − ||δw||2
for compactly supported smooth vector fields w on
oriented Riemannian manifolds, with ( , ) and
|| || denoting, as usual, the
L2 inner product and
L2 norm. The proof of (18.7) based on (16.4)
and trivial integration by parts, along with the trivial observation that
|£w g|2 = 2w j, k[wj, k + wk, j].
Matsushimas theorem (the general Riemannian version):
for any
compact oriented Einstein manifold, Ker D
is the L2-orthogonal direct sum
of the Lie
algebra k of all Killing fields, and the space
p of all gradients in
Ker D .
The formulae for the k-component u and
p-component v of any w
in Ker D in the case where the Einstein
constant λ is positive:
(18.8) u = w + ∇φ, v = −∇φ, with
φ given by 2λφ = δw.
Proof of Matsushimas theorem. First, by (16.9),
Ker D = k = p = {0} if
λ < 0, while
p = {0} and
Ker D = k is
the space of all parallel vector fields if
λ = 0. Now let
λ > 0. Then u and
v in (18.8) lie in Ker D,
which is clear for v in view of the second
conclusion in (18.6), and hence obvious for u. Finally,
due to (18.7) with w replaced by u
(and Du = 0), u in (18.8) is a
Killing field, as
δu = δw
+ δ∇φ = 2λφ + Δφ, which vanishes since (18.8)
and (18.6) give
φ ∊ Ker (Δ + 2λ).
[Another argument for λ > 0: as we saw,
v defined by (18.8) lies in
Ker D whenever
w ∊ Ker D.  Thus, assigning
v (18.8) to w we obtain a linear
operator
Π : Ker D → p. The restriction of Π to
p is the identity: if
w = ∇f ∊ Ker D, (18.2) allows us to assume, by adding a constant to
f, that
f lies in
Ker (Δ + 2λ), and so
δw = Δf = −2λf, giving
φ = −f and
v = w in (18.8). Hence
Π is a direct-sum projection onto
p, while, by the divergence theorem, (18.7), and (16.5),
Ker Π = i(M, g).]
[FR]. [KG]:
Section 9 (formula (9.3), Remark 9.1(iv), formula (9.4), the second
part of Theorem 9.4, formula (9.8), and Theorem 9.6).
Homework #18.
- FEBRUARY 21. The Kähler case of Matsushimas
theorem, in which
Ker D = h(M) by
(17.2) and, in addition,
(19.1) Jk = p unless λ = 0,
the equality in (19.1) being immediate if one notes that, for any smooth
vector field w on a Kähler manifold,
(19.2) ∇(Jw) = J∇w.
In fact, given a bundle morphism
A : TM → TM,
it is obvious that
(19.3) [JA]* = −JA* if A and J commute.
Thus, by (19.2), for a holomorphic vector field w,
self-adjointness of ∇(Jw) is equivalent to
skew-adjointness of ∇w, and vice versa.
Complexifications and real forms of Lie algebras. The conclusion that,
in any compact Kähler-Einstein manifold
(M, g) with nonzero Einstein constant, the
Lie algebra k = i(M, g)
of Killing vector fields is a real form of h(M), and
hence
(19.4) dimR i(M, g) = dimC h(M).
Compact Lie algebras, defined by requiring the existence of a Euclidean inner
product < , > that makes
< [u,v],w >
skew-symmetric in u,v,w (or, equivalently, has
< [u,v],v > = 0
for all u,v, which is further equivalent to
skew-adjointness of the operators
Ad u = [u, ⋅ ] for all
u). Compactness of
i(M, g) for a compact oriented
Riemannian manifold (M, g), with
< , > chosen to be the
L2 inner product ( , ).
Specifically,
([u,v],v) = 0 for all Killing
fields u,v since
(19.5) g([u,v],v) = δw, where w = −[g(u,v)]v.
The corollary that h(M) has
a compact real form whenever the compact almost complex manifold
M admits a Kähler-Einstein metric. The obvious fact
that - by (19.2) and (19.3) - if the endomorphisms
J and ∇w of the tangent bundle
of a Kähler manifold commute, then
[∇w + (∇w)*] J = ∇u − (∇u)* for
u = Jw. Hence, in view of (15.2),
(19.6) (£w g) J = d [g(Jw, ⋅ )] for any holomorphic vector field w on a Kähler manifold.
Covariant differentiation of smooth sections defined along a curve in
the base manifold of a real or complex vector bundle endowed with a
connection. Geodesics of a connection in TM, for a
manifold M. The equality, satisfied by every Killing
field u and any smooth vector field
w in any Riemannian manifold:
(19.7) uq, jk = u pRpkj q, that is, ∇w[∇u] = R(u,w).
Relation (19.7) as a trivial consequence of the identity
(19.8) 2uq, jk = 2Rqjk pup + aqj, k + aqk, j − ajk, q , with ajk = uj, k + uk, j ,
valid for any smooth vector field u on a Riemannian
manifold.
[FR]. [KG]:
Section 9 (part (d) of Theorem 9.6).
Homework #19.
- FEBRUARY 23. Proof of (19.8), using the Ricci identity
(20.1) uj, kl − uj, lk = Rlkj pup ,
immediate from (10.1), gives
2uq, jk − aqj, k − aqk, j + ajk, q = (Rqkj p + Rkjq p + Rqjk p)up , which equals 2Rqjk pup due to the first Bianchi
identity. The first-order linear ordinary differential equation
satisfied along any geodesic, in view of (19.7), by the pair
(u, ∇u), for any Killing field
u in a Riemannian manifold. The resulting injectivity of
the linear operator sending u ∊ i(M, g)
to
(ux, (∇u)x), at any fixed point x. (Namely, if
(ux, (∇u)x) = (0,0), then
(u, ∇u) vanishes along every broken
geodesic emanating from x, while such broken geodesics reach every
point in M due to connectivity.) The obvious consequence in the
form of the upper bounds
(20.2) dimR i(M, g) ≤ n(n+1) ⁄ 2, dimR k ≤ m(m+2)
for the Lie algebra i(M, g) of all
Killing vector fields on an n-dimensional Riemannian manifold
(M, g) and, respectively, the Lie algebra
k of holomorphic Killing vector fields on any Kähler
manifold of complex dimension m. Linear
vector fields on finite-dimensional real/complex vector spaces. Linearity of
the flow transformations
e tA of a linear vector field
A on a vector space, completeness of A,
and the commutation relation
Ae tA = e tAA, all derived from the existence and
uniqueness of global solutions to first-order linear ordinary differential
equations
[DG
p. 208], applied to the equation
dΨ ⁄ dt = AΨ
satisfied by Ψ equal to
e tAx, or
e tA(x + y) − e tAx − e tAy, or
e tA(cx) − ce tAy, or
Ae tA − e tAA, where
x, y ∊ V and
c ∊ R.
[FR].
Homework #20.
- FEBRUARY 26. Projectability of vector fields under
smooth mappings between manifolds, with uniqueness and linearity of the
push-forward operation when the mapping is surjective
[DG p. 23]. The fact that all linear vector fields on a
finite-dimensional complex vector space V are projectable
onto P(V )
and their projected images are holomorphic, leading to a complex vector space
of dimension m(m+2) consisting of holomorphic
vector fields on CPm and, consequently,
the inequality dimC h(CPm) ≥ m(m+2). The conclusion
(from (17.3), (19.4), (20.2) and the last inequality) that
(21.1) dimC h(CPm) = m(m+2)
and the holomorphic vector fields on P(V), for any
finite-dimensional complex vector space V, are precisely
the projected images of linear vector fields on V.
The equality
(21.2) 2∇w ζ = ζw + Jζw J, where ζ = hJ and ζw = (dζ )(w, ⋅ , ⋅ ),
for an almost complex manifold M, any torsion-free
connection ∇ in TM, and any Hermitian
tensor field h on M with
∇h = 0, valid under the assumption that
J is parallel relative to some torsion-free connection in
TM. The special case of (21.2) in which
h is a Hermitian metric g, while
∇ and
ζ = Ω denote the Levi-Civita
connection and Kähler form of  g. The conclusion that on any
fixed almost complex manifold the set of all Kähler metrics is empty or
coincides with the set of all almost-Kähler metrics, while the latter set
is a convex cone in a vector space.
[FR]. [KG]:
Section 5 (the statement of Lemma 5.2, and Theorem 5.3).
Homework #21.
- FEBRUARY 28. Proof of (21.2). The line segment of
Kähler metrics joining two given Kähler metrics on a fixed almost
complex manifold. The fact that for any Kähler metric
g and any compactly supported smooth function
f on an almost complex manifold M
the tensor
g + ε(i∂∂f )J
is a Kähler metric as long as ε is
sufficiently close to 0 in R. Time-dependent
Riemannian metrics on a manifold. Relations satisfied by the derivatives with
respect to time - here denoted by
( )′ = d ⁄ dt - with the time-dependent function φ
defined by
dg′ = φ dg
whenever g is a time-dependent Riemannian metric:
(22.1)
δ′ = dφ, trg g′ = 2φ.
[FR]. [KG]:
Section 5 (proof of Lemma 5.2), Section 8 (formula (8.10), the nine
lines immediately preceding it, and the six lines following it).
Homework #22.
- MARCH 2. The existence, on any compact oriented
Riemannian manifold (M, g), of a smooth
function
f : M → R,
unique up to the addition of a constant, such that
(23.1)
Δ f + s is constant
(the latter constant being the average scalar curvature
savg). The resulting differential operator
L, sending smooth vector fields v to
smooth functions, and given by
(23.2) Lv = δv − dv f or, equivalently, Lv = e fδ(e− fv).
The Futaki functional, assigning to any smooth vector field
v on the given compact oriented Riemannian manifold
(M, g) of dimension n
the real number
(23.3) Fv = (savg)n ⁄ 2 ∫ M dv f dg,
where f is chosen as in (23.1). The Futaki invariant
F : h(M) → R of a compact almost complex manifold M
having
c1(M) > 0 or
c1(M) < 0 and
admitting a Kähler metric, defined by (23.3) for a Kähler
metric g on M satisfying the condition
(23.4) i∂∂f + ρ = λΩ with some constant
λ and some smooth function
f : M → R,
where f in (23.4) - necessarily, due to (11.3) - also
satisfies (23.1), while
savg = 2mλ for
m = dimCM.
Futakis theorem stating that, if
c1(M) > 0 or
c1(M) < 0, then
(23.5) g
with (23.4) must exist, and the Futaki invariant
does not depend on its choice,
the existence claim being obvious from the
∂∂
Lemma in the lines following (11.3), and (12.2), since, by the
∂∂
Lemma, (23.4) is equivalent to
(23.6) [ρ] = λ[Ω] in H2(M, R), for some real constant λ.
The fact that, whenever a Kähler manifold
(M, g) satisfies (23.4),
then, with f appearing in (23.4) used to define
L as in (23.2), one has, for any smooth vector
field v on M,
(23.7) ∇Lv − J∇LJv = −2λv − JEA, where A = £v J,
E being the linear differential operator that sends any smooth
endomorphism A of TM to the vector
field EA characterized by
(23.8)
g(EA, ⋅ ) = δA − (df )A or, equivalently, g(EA, ⋅ ) = e fδ(e− fA).
Proof of (23.7), based on the coordinate relation
(Jw)k = −Jkpwp for vector fields w (due to skew-adjointness of
J), (9.7), (15.2) (in its coordinate form
Jkpv k, q =
Aqp + Jqkv p, k), the Ricci identity
(23.9) v k, pq − v k, qp = Rpql k v l,
cf. (10.1), the Bochner formula (16.2), and the equality
JlqJkpRpqs j = Rkls j (see Problem 2 in Homework #6).
[FR]. [KG]:
Section 8 (from the beginning to the seventh line following formula
(8.3) - which is the coordinate version of (23.7) above - except Theorem 8.2
and the last sentence of Theorem 8.1).
Homework #23.
- MARCH 5. The identity, using the same assumptions and
notations as (23.7), which trivially follows from (23.7) (cf. Problem 1 in
Homework #23):
(24.1) ΔLv = −2λ δv − δ(JEA), where A = £v J.
The observation that, for a compact almost complex manifold M,
(24.2) the Futaki invariant vanishes
if M admits a non-Ricci-flat
Kähler-Einstein metric,
as (23.3) yields Fv = 0 for all
v whenever s is constant. (Since (7.6) then
gives
c1(M) > 0 or
c1(M) < 0, the
Futaki invariant is well defined.) Time-dependent Kähler metrics on a
manifold with a fixed (time-independent) almost complex structure
J. The case of a time-dependent Kähler metric with
Ω ′ = 2i∂∂χ for
some time-dependent function χ, in which, with the
time-dependent function
φ defined by
dg′ = φ dg,
(24.3) φ = Δχ, ρ′ = − i∂∂ Δχ.
(The first equality in (24.3) follows from the second part of (22.1) along
with the relation
g = −ΩJ and
(11.3); the second - since (12.2) gives
ρ = ρ0 − i∂∂ log γ for γ with
dg = γ dg0,
the subscript 0 standing for the value
at a fixed time t0, and so
φ = (log γ)′.)
The further conclusion that, if condition (23.4) holds just at one time
t0, and
Ω ′ = 2i∂∂χ as
before, then a time-dependent function f may be chosen so
as to satisfy (23.4) at all times t, and, with
L defined by (23.2),
(24.4) L′ = −2λdχ, f ′ = Δχ + 2λχ, λ′ = 0.
(Namely, f is obtained by solving the equation
f ′ = Δχ + 2λχ with prescribed
f0, and then (24.4) follows from (23.2),
(22.1) and (24.3), while (i∂∂f + ρ − λΩ )′ = 0 by the
second equality in (24.3) with
Ω ′ = 2i∂∂χ.) Proof of the
does not depend clause in (23.5): both metrics may be assumed to
satisfy (23.4) with the same
savg = 2mλ, so that
Ω ′ = 2i∂∂χ for the line
segment joining them,
2i∂∂χ being the difference of their Kähler forms, with a
(time-independent) smooth function χ which exists in
view of the
∂∂
Lemma, and - letting ~ mean differs by a divergence -
one has
dv f ~ −Lv by (23.2), while
(Lv dg)′ = [(Lv)Δχ − 2λdv χ] dg
from the first equalities in (24.3) and (24.4); finally, (11.1) and (24.1)
with A = 0 give
(Lv, Δχ) = (ΔLv, χ) = −2λ(δv, χ) and, at the same time,
−χδv ~
dv χ. Homotopy
equivalence. Simple connectivity. The Poincaré conjecture.
[FR]. [KG]:
Section 8 (the last sentence of Theorem 8.1, and Lemma 8.7).
[FR]. [KG]:
Section 8 (Lemma 8.4(b), the last sentence of Theorem 8.1, and Lemma
8.7). Homework #24.
- MARCH 7.
The Ricci flow on a manifold M, the trajectories of which
are defined to be time-dependent metrics g on
M with g′ = −2r.
Ricci solitons, that is, Riemannian manifolds
(M, g) satisfying the condition
(25.1) £w g + r = λg or, equivalently, wj, k + wk, j + Rjk = λgjk, with a constant λ,
for some smooth vector field w. Kähler-Ricci
solitons, by which one means those Ricci solitons
(M, g) which at the same time are
Kähler manifolds for some almost complex structure J
on M. The observation that
trg applied to (25.1) yields
(25.2) 2 δw + s = nλ, where n = dim M.
The conclusion - obtained by integrating (25.2) and using the divergence
formula - that when M is compact and oriented,
λ is uniquely determined by g
(being equal to 1/n times the average scalar curvature),
which in turn makes w unique up to the addition of a
Killing field. Gradient Ricci solitons: the Ricci solitons
with w in (25.1) which is a gradient,
w = ∇f ⁄ 2,
that is, the Riemannian manifolds
(M, g) satisfying the condition
(25.3) ∇df + r = λg or, equivalently, f, jk + Rjk = λgjk, with a constant λ,
for some smooth function f. The result of Perelman,
which we will not prove or use:
Every compact
Ricci soliton is a gradient Ricci soliton .
Einstein manifolds as the simplest examples of (gradient) Ricci solitons. A
corollary:
(25.4) Dw = 0 whenever £w g + r = λg with a constant λ,
which follows since applying d trg − 2δ to (25.1) we obtain
2g(Dw, ⋅ ) = 0 from
(25.2), (16.4) and (9.1). The immediate consequence that, by (17.2),
(25.5) w in (25.1) is holomorphic for any compact Kähler-Ricci soliton
(M, g).
The conclusion that, due to (25.5) and (19.6), in every compact
Kähler-Ricci soliton, multiplying (25.1) from the right by
J one obtains (23.6) or, equivalently, in view of the
∂∂ Lemma, (23.4). Thus, just as in the Kähler-Einstein case,
c1(M) is positive, negative or zero
for a compact almost complex manifold M admitting a
Kähler-Ricci soliton metric, the sign being the same as that of the
constant λ in (25.1).
The fact that (23.1) follows both from (25.3) (by applying
trg) and from (23.4) (via (11.3)). More on
complexifications of real vector spaces (example: the spaces of real-valued
and complex-valued smooth functions on a manifold). The unique complex-linear
and complex-bilinear/sesquilinear extensions to complexifications of
real-linear and real-bilinear mappings (example: the real and complex
L2 inner products of compactly supported smooth functions
on an oriented manifold with a fixed volume form). The real-part and
conjugation operators. The fact that a complex-linear operator from a complex
space into a complexification is uniquely determined by its real part (which
may be any real-linear operator). The differential operator
P, sending smooth vector fields v on
a compact Kähler manifold to smooth complex-valued functions, and defined
to be the unique complex-linear operator having the real part
L defined by (23.2), so that
(25.6) Pv = Lv − i LJv.
The Tian-Zhu invariant
T : h(M) → C
of a compact almost complex manifold M having
c1(M) > 0 or
c1(M) < 0 and
admitting a Kähler metric, given by
(25.7) Tv = (savg)n ⁄ 2 ∫ M e Pv dg.
for a Kähler metric g on M with
(23.4). The fact that
T : h(M) → C with (25.7) is holomorphic for any compact Kähler manifold
(M, g) due to complex-linearity of
P since, obviously,
(25.8) dTv u = (savg)n ⁄ 2 ∫ M e PvPu dg
whenever v,u ∊ h(M). The
conclusion that, by (25.8) and (23.3),
(25.9) the Futaki invariant equals
−Re dT0 .
[FR]. [KG]: Remark
9.1(ii) in Section 9. Homework #25.
- MARCH 9. According to Problem 1 in
Homework #26, for a vector field
v on a Kähler manifold
(M, g),
(26.1) Jv ∊ i(M, g) whenever v is a holomorphic gradient.
On the other hand, given a Killing vector field u on a
compact oriented Riemannian manifold (M, g),
with f and L as in (23.1) and
(23.2), due to uniqueness of f when it is normalized so
that favg = 0, one has
(26.2) du f = Lu = 0.
Next, if a Kähler manifold (M, g)
satisfies (23.4) with λ ≠ 0 then, by
(23.7), for L and P defined by
(23.2) and (25.6), where f is the function appearing in
(23.4),
(26.3) the restriction of
P to h(M) is injective.
The equality, valid for any
v,u ∊ h(M) whenever
(M, g) is a compact Kähler manifold,
(26.4) (du du T)(v) = (savg)n ⁄ 2 ∫ M e Pv(Pu)2 dg,
which is obvious from (25.8). Constancy of
Δf − |∇f |2 + 2λf under the assumption
(25.3) (see Problem 2 in Homework #25). The resulting
fact that, by (25.8), (25.6) and the second part of (23.2), for any compact
gradient Kähler-Ricci soliton (M, g),
with f and λ ≠ 0 satisfying (25.3),
(26.5) v = ∇f ⁄ (2λ) is a critical point of
the Tian-Zhu invariant T,
where T : h(M) → C
is given by (25.7), and
v ∊ h(M) in view of (25.5).
An immediate consequence of (26.2) - (26.4): if
k denotes the Lie algebra of all Killing fields on (M, g),
then the restriction of T to the space
Jk is real-valued and, in view of (26.4), its
Euclidean Hessian is positive-definite, so that
T : Jk → R
can have at most one critical point. The conclusion, from the last sentence,
(26.1), (24.2), (25.9) and (26.5), that a compact almost complex manifold
cannot simultaneously admit a Kähler-Einstein metric and a non-Einstein
Kähler metric which is a gradient Ricci soliton, since if
λ = 0 we may use Problem 3 in
Homework #26, while for
λ ≠ 0 the existence of the latter
metric implies nonzero Futaki invariant.
Homework #26.
- MARCH 19. The observation that, for any
smooth vector field v on a Kähler manifold
(M, g) with (23.4), defining
L, P and E by (23.2),
(25.6) and (23.8), applying J to both sides of (23.7),
and then taking the divergence, one obtains
(27.1) ∇LJv + J∇Lv = −2λJv + EA, ΔLJv = −2λ δ(Jv) + δEA with A = £v J.
In addition, with the same assumptions and notations as in (27.1),
(27.2) |∇Lv|2 + 2λ dv Lv + g(∇Lv, JEA) = |∇LJv|2 + 2λ dJv LJv − g(∇LJv, EA)
for any smooth vector field v on M,
as well as
(27.3) 2[g(∇Lv, ∇LJv) + λ (dJv Lv + dv LJv)] = g(∇Lv + J∇LJv, EA).
In fact, the left-hand side of (27.2),
g(∇Lv, ∇Lv + 2λv + JEA), equals
g(∇Lv, J∇LJv),
by (23.7), so that it is invariant under the replacement of
v with Jv. This invariance amounts
to equality (27.2), since, whenever v is a smooth vector
field on a Kähler manifold and A is any smooth
endomorphism of its tangent bundle, (15.2), (19.2) and the definition (23.8)
of E give
(27.4) £Jv J = −(£v J)J, E(AJ) = −JEA.
Similarly, the left-hand side of (27.3) is
g(∇LJv, ∇Lv + 2λv) + g(∇Lv, ∇LJv + 2λJv), so that
(27.1) and (23.7) yield (27.3). The equality (see Problem 1 in
Homework #27), valid for smooth complex-valued
functions ψ on any Riemannian manifold:
(27.5) Δeψ = [Δψ + g(∇ψ, ∇ψ)] eψ,
where Δ and g are extended
complex-(bi)linearly, so that
(27.6) Δψ = Δ Re ψ + i Δ Im ψ and g(∇ψ, ∇ψ) = |∇ Re ψ|2 − |∇ Im ψ|2 + 2i g(∇ Re ψ, ∇ Im ψ).
Tian and Zhus theorem: if a compact almost complex manifold
M having
c1(M) > 0 or
c1(M) < 0
admits a Kähler metric, then, for the Tian-Zhu invariant
T : h(M) → C of (M, g) given by (25.7),
(27.7) T does not depend
on the choice of a Kähler metric g with
(23.4).
Note that such g must exist according to the lines
following (23.5). Proof of (27.7):
Homework #27.
- MARCH 21. The spaces
XM of all
smooth vector fields and
Fc M of all smooth complex-valued functions on a given manifold
M. Three important complex-linear operators
(28.1) Θ : Fc M → Fc M, P : XM → Fc M, ∂ : Fc M → XM
in the case of a compact Kähler manifold
(M, g). Namely, using a fixed smooth
function
f : M → R
such that Δ f + s is
constant, we set
(28.2) Θ = Δ − du − i dJu for u = ∇f,
while P and the complex gradient operator
∂ are defined by (25.6) with (23.2) and, respectively
(28.3) ∂ψ = ∇ Re ψ + J∇ Im ψ.
Futakis theorem: for a compact Kähler manifold
(M, g) satisfying (23.4) or,
equivalently, (23.6), with
λ ≠ 0,
(28.4) ∂ maps Ker (Θ + 2λ) isomorphically onto h(M),
and its inverse is −P ⁄ (2λ).
The Kähler version (19.1) of Matsushimas theorem as a special case
of (28.4). The first step in a proof of (28.4): on a Kähler manifold
satisfying (23.4), with
A = £v J
and E as in (23.8),
(28.5) (Θ + 2λ)ψ = − i PEA for ψ = Pv
whenever
v ∊ XM, while, for any
ψ ∊ Fc M
and
v ∊ XM,
(28.6) (i) P∂ ψ = Θψ, (ii) ∂ Pv = −2λv − JEA.
In fact, since
δ = L + df
by (23.2), relations (24.1), (23.7) and (27.1) imply (28.5). Similarly,
(25.6), (23.2), (23.7) and (28.2) - (28.3) yield (28.6.i) and (28.6.ii).
Homework #28.
- MARCH 23. The μ-adjoint
Π* of a real/complex linear operator
Π, characterized by
(29.1) (Πψ, θ) = (ψ, Π*θ) for all ψ and θ,
where μ is a fixed (positive) volume form on an
oriented manifold M and ( , )
denotes the L2 inner product of compactly supported
smooth sections of any given real/complex vector bundle over
M, associated with μ and any
fixed Riemannian/Hermitian fibre metric in the bundle, while
Π sends compactly supported smooth sections of one
such bundle to analogous sections of the other. Uniqueness of the
μ-adjoint Π* (when it
exists). Some trivial observations:
(Π Σ)* =
Σ*Π*, the
composites
Π Π* and
Π*Π are
μ-self-adjoint,
Π** = Π,
(29.2) Π*Π is self-adjoint, nonnegative, and Ker Π*Π = Ker Π.
The fact that, if Π has the
μ-adjoint Π* and
f is a smooth function on M, then
Π has the
e− fμ-adjoint
Πf*, and
(29.3) Πf*θ = e fΠ*(e− fθ).
The weighted L2 inner products
( , )f , related to
( , ), the original ones, by
(29.4) ( ⋅ , θ)f = ( ⋅ , e− fθ),
and the corresponding weighted L2 norms
|| ||f , so that, if
Πf* exists,
(29.5) (Πψ, θ)f = (ψ, Πf*θ)f for all ψ and θ.
The Cauchy-Riemann operator H sending smooth vector
fields on an almost complex manifold to endomorphisms of its tangent bundle,
with
(29.6) Hv = [J, ∇v], that is, Hv = £v J,
cf. (15.2). The dg-adjoint of the Cauchy-Riemann operator
H of a Kähler manifold
(M, g), characterized by
(29.7) g(H*B, ⋅ ) = δ[J, B*].
Justification of (29.7) based on the trivial observation that
(29.8) [J, B]* = [J, B*], for all endomorphisms B of the tangent
bundle,
and integration by parts of the inner product
tr A[J, ∇v]* = tr A[J, (∇v)*].
An obvious algebraic property of the commutator/anticommutator of
J and an endomorphisms A of the
tangent bundle of an almost complex manifold:
(29.9) [J, A] anticommutes, JA + AJ commutes with J.
Homework #29.
- MARCH 26. Given a smooth function f
and a smooth vector field v on a Kähler manifold,
one has
(30.1) Hf*Hv = 2JEA*, where A = £v J
and E is the operator defined by (23.8). In fact, for
B = ∇v and
A = Hv
= £v J = [J, B], (29.3) with
Π = H and (29.7) - (29.9) yield
g(Hf*Hv, ⋅ ) = e fg(H*[J, e− fB], ⋅ ) = e fδ[J, [J, e− fB*]]
= −2e fδ(e− f[J, B*] J) = −2e fδ(e− fA*J)
which, due to the second relations in (23.8) and (27.4), equals
−2g(E(A*J), ⋅ ) = 2g(JEA*, ⋅ ),
proving (30.1). Furthermore, by (29.8), if B is
self-adjoint, so must be both [J, B] and
[J, JB]. Thus, (28.3) gives, in an
arbitrary Kähler manifold (M, g),
(30.2) (£v J)* = £v J whenever v = ∂ψ
for any ψ ∊ Fc M.
Let f, in addition, satisfy (23.4) with a constant
λ. Then, for any smooth function
ψ,
(30.3) Hf*H∂ ψ = −2∂ [(Θ + 2λ)ψ],
since, for v = ∂ψ,
(30.2) gives (30.1) with
A* = A, so that, from
(28.6.ii), the left-hand side of (30.2) equals
−2(∂ Pv + 2λv), and hence the equality follows from (28.6.i). We can now
establish Futakis theorem (28.4): by
and (30.3) and (29.2), ∂ maps
Ker (Θ − 2λ) into h(M), while (28.5) clearly shows that
P maps h(M) into
Ker (Θ + 2λ) . Since
λ ≠ 0, (28.6) implies that the
resulting operators
∂ : Ker (Θ + 2λ) → h(M)
and
−P ⁄ (2λ) : h(M) → Ker (Θ + 2λ)
are each others inverses, which completes the proof of (28.4). Next, for
the gradient operator ∇ of any oriented Riemannian manifold
(M, g) and any smooth function
f on M, defining
L by (23.2), one has
(30.4) δ = −∇*, L = −∇f*.
In other words, δ is the dg-adjoint,
and L the
e− fdg-adjoint of
−∇; in fact, the first relation is obvious from
integration by parts, and the second then follows via the last equality in
(23.2) and (29.3). If, in addition, (M, g)
is a Kähler manifold, then, according to Problem 1 in
Homework #30, for P as in (25.6),
(30.5) P = −∂f*.
Another result of Futaki: in a compact Kähler manifold
(M, g) with (23.4) (that is, having
[ρ] = λ[Ω]),
for a constant λ ≠ 0,
(30.6) Θ = −∂f*∂ is self-adjoint, nonpositive, and Ker Θ = C,
C being the space of constant complex-valued functions. In fact,
(30.6) is clear from (28.6.i), (30.5), (29.2) and Problem 2 in
Homework #28.
Homework #30
- MARCH 28. The conclusion that, for any compact
almost complex manifold,
(31.1) h(M) = {0} if c1(M) < 0 and M admits a Kähler metric.
Proof of (31.1): nonnegativity of −Θ in (30.6) implies
that, when c1(M) < 0 (and
hence λ < 0), 2λ
cannot be an eigenvalue of −Θ. Thus,
Ker (Θ + 2λ) = {0},
and we can use Futakis theorem (28.4). A generalization of
Bochners integral formula (17.6): with hv
defined by
g(hv, ⋅ ) = h(v, ⋅ ),
(31.2) (hv, v)f = || Lv ||2f − (B, B*)f , where h = ∇df + r and B = ∇v,
for any smooth compactly supported vector field v on an
oriented Riemannian manifold (M, g), any
smooth function
f : M → R, and L defined as in (23.2), that is,
Lv = δv − dv f. The Kähler case, still
with arbitrary v and f, in which,
setting ζ = i∂∂f + ρ,
one sees - from (9.7), (7.3) and (1.1) - that
− 2ζJ = ∇df − J(∇df )J + r − JrJ,
where r − JrJ = 2r,
and so − 2ζJ = h − JhJ. Suppose now that one also has
i∂∂f + ρ = λΩ with a constant
λ, as in (23.4) or, equivalently,
(31.3) − 2ζJ = 2λg.
The resulting equality
h − JhJ = 2λg combined with (31.2) and (25.6) gives, for any smooth compactly
supported vector field v,
(31.4) 2λ||v||2f = ||Pv||2f − (B, B*)f − (JB, (JB)*)f , with B = ∇v.
The observation (see Problem 2 in Homework #31) that,
given any smooth function
ψ : M → C on a Kähler manifold (M, g),
(31.5) 2 tr [B2 + (JB)2] = |A|2 for B = ∇v and A = £v J, where v = ∂ψ.
The integration of (31.5) against the volume form
e− fdg yields
2(B, B*)f + 2(JB, (JB)*)f = ||£v J||2f whenever f or
ψ is compactly supported (while one still sets
v = ∂ψ and
B = ∇v). Under the additional
assumption (23.4), this allows us to rewrite (31.4), with the aid of (28.6.i),
as
(31.6) 4λ||v||2f = 2||Θψ||2f − ||£v J||2f if v = ∂ψ for any ψ ∊ Fc M,
Yet another theorem due to Futaki: in a compact Kähler manifold
(M, g) satisfying
[ρ] = λ[Ω]
with a constant λ ≠ 0,
(31.7) τ ≥ 2λ
whenever τ is a nonzero eigenvalue of
−Θ.
(Note that (17.9) is a special case.) Proof of (31.7): if
Θψ = −τψ,
then, by (30.6),
||Θψ||2f = (Θψ, Θψ)f = −τ(ψ, Θψ)f = τ(ψ, ∂f*∂ψ)f = τ(∂ψ, ∂ψ)f = τ||v||2f for v = ∂ψ,
and so (31.6) with v ≠ 0 trivially
implies (31.7). Next: under the assumptions of Futakis theorem (28.4),
for the isomorphism ∂ in (28.4), with
u = ∇f,
(31.8) ∂ maps
Ker (Θ + 2λ) ∩ Ker dJu isomorphically
onto k ⊕ p, and Jk = p.
Here, as in (19.1), k ⊕ p is the
L2-orthogonal direct sum of the Lie algebra
k of all Killing fields, and the space p
of all holomorphic (real) gradients. In addition,
(31.9) Ker (Θ + 2λ) ∩ Ker dJu = {ψ ∊ Fc M : Re ψ, Im ψ ∊ Ker (Θ + 2λ)}.
In fact, (31.9) is obvious from (28.2), while (31.8) follows since an
L2-orthogonality argument (see the hint for Problem 1 in
Homework #28) shows that, if
∂ψ ∊ k ⊕ p,
then the projections of ∂ψ onto
k and p
must coincide with ∇ Re ψ and
J∇ Im ψ.
Homework #31.
- MARCH 30. For a compact oriented
Riemannian manifold (M, g) and a smooth
function
f : M → R,
(32.1) dw is ( , )f -skew-adjoint if w ∊ i(M, g) and dw f = 0.
Let (M, g) now be a compact gradient
Kähler-Ricci soliton with f and
λ ≠ 0 satisfying (25.3), so that
u = ∇f is holomorphic by (25.5),
and Ju is a (holomorphic) Killing field according to
Problem 1 in Homework #26. We proceed to
discuss some observations made, in this case, by Tian and Zhu. As the Lie
derivative £Ju then
commutes with both Θ and the complex gradient
∂ (see Problem 2 in Homework #31),
(32.2) the operator
−i dJu leaves Ker (Θ + 2λ) invariant
(since so does dJu). Recall that
h(M) is a complex Lie algebra, cf. (15.1) and
Problem 2 in Homework #16. Under the
complex-linear isomorphism ∂ in (28.4),
(32.3) −i dJu corresponds to
Ad u = [u, ⋅ ] : h(M) → h(M),
which is immediate from the fact that
dJu corresponds to
Ad Ju. From (28.2),
(32.4) −i dJu = −Δ + du − 2λ on Ker (Θ + 2λ).
On the other hand, by (32.1),
(32.5) −i dJu : Ker (Θ + 2λ) → Ker (Θ + 2λ) is ( , )f -self-adjoint.
Homework #32.
- APRIL 2. Again,
(M, g) is a compact gradient
Kähler-Ricci soliton, with f and
λ ≠ 0 satisfying (25.3). It follows from
(32.3) and (32.5) that, as a vector space, h(M) is
the direct sum of the eigenspaces
hτ corresponding to the eigenvalues
hτ of
Ad u = [u, ⋅ ].
More generally, let
hτ = Ker ([u, ⋅ ] − τ) whenever
τ ∊ R. As
[u, ⋅ ] always is, by the Jacobi identity,
a derivation of the Lie algebra in question, we have, for any real
τ, σ,
(33.1) [hτ, hσ] ⊆ hτ + σ , while h0 is the complex Lie subalgebra
k ⊕ p.
Recall: holomorphic functions
φ : M → C
on an almost complex manifold M are characterized by
(dφ)J = i dφ
(as they are nothing else than holomorphic mappings
M → C). The algebra of
holomorphic functions. Holomorphicity of multiplicative inverses. Vector
bundles over manifolds, local sections, local trivializations and their
compatibility, in the sense of regularity of transition functions
[DG pp. 57-58]. Holomorphic complex vector bundles over almost complex
manifolds, defined analogously to smooth real/complex vector bundles over
smooth manifolds
[DG p. 58], just with holomorphicity of transition functions rather than
their smoothness. Examples: tangent bundles of complex manifolds, and the
tautological line bundle
T over any complex
projective space P(V), cf.
[DG p. 59]. The total space of a vector bundle
[DG p. 66]. The holomorphic complex line bundle
L naturally associated
with a complex submanifold Q of codimension one in a
complex manifold M, and given by
(33.2) Lx = (Tx M ⁄ Tx Q)* if x ∊ Q, and Lx = C otherwise.
The atlas of local trivializations for this
L, consisting of
all (U, φ) such that
(33.3) U ⊆ M  is open,
φ : U → C is holomorphic, and
φ−1(0) = U ∩ Q,
while dφ ≠ 0
everywhere in  U ∩ Q. Here
φ is treated as a local section of
L defined on
 U, with
φx = dφx if x ∊ Q
(cf. Problem 1 in Homework #33) and
φx = φ(x)
otherwise.
Homework #33.
- APRIL 4. The fact that an almost complex submanifold
Q of a complex manifold M is itself
a complex manifold as one sees, locally, using a mapping whose restriction to
Q, due to the inverse mapping theorem, is a
biholomorphisms. The fact, immediate from the holomorphic version of the rank
theorem (see below) that locally, up to biholomorphisms, complex submanifolds
appear as vector subspaces in a complex vector space. The conclusion from an
integral version of the first-order Taylor formula (see Problem 1 in
Homework #34) that a holomorphic function
φ on an open set U in a complex
vector space, vanishing on
U ∩ Ker ξ for a
nonzero linear functional ξ, is holomorphically
divisible by ξ, with the quotient
φ ⁄ ξ equal
on U ∩ Ker ξ to the
ratio
dφ ⁄ dξ
of differentials. The resulting mutual compatibility of the local
trivializations (U, φ) in the line
following formula (33.2). The equality
i∂∂f = 0
valid, in view of (9.5), whenever
f = Re φ and
φ is a holomorphic function. The fact that a
holomorphic function on a compact connected Kähler manifold is
necessarily constant (and, more generally, every holomorphic function on a
Kähler manifold is harmonic). Holomorphic sections of a holomorphic
vector bundle B over an
almost complex manifold, and the space
H o(B )
of all such sections.
Homework #34.
- APRIL 6. The operators
∂v and
∂v acting on smooth functions
U → C, where
U is an open subset of an
m-dimensional complex vector space V and
v ∊ V, and given by
(35.1)
2∂v = dv − i div , 2∂v = dv + i div .
The complex partial derivatives ∂j and
∂j relative to a fixed (complex) basis
e1 , . . . , em
of V, equal, respectively, to
∂v and
∂v for
v = ej .
Holomorphicity of
φ ∊ Fc U as equivalent to
requiring that
∂v φ = 0, or
∂j φ = 0, or
∂v φ = dv φ for all v (or all
j). The complex Jacobian matrix
[∂j F a], representing the differential, at any point
x, of a holomorphic
mapping F from an open set U into a
finite-dimensional complex vector space, where in both spaces fixed bases are
used; in fact,
dFx v = v j(∂j F a)(x). The holomorphic version of the rank
theorem
[DG pp. 33-34]. The identities (see Problem 1 in
Homework #35):
(35.2) [φv, w] = φ[v, w] − (dw φ)v − (Im φ)(£w J)v, (£φv J)u = φ(£v J)u − (dJu φ − i du φ)v,
valid whenever
v, w, u ∊ XM and
φ ∊ Fc M
in any almost complex manifold M. The relation
(35.3) H o(TM) = h(M) if
M is a complex manifold.
Homework #35.
- APRIL 9.
Product bundles. Holomorphic vector-bundle morphisms and (holomorphically)
trivial bundles. Operations on holomorphic
vector bundles: direct sum, Hom, the dual, and the tensor product.
Tensor powers of line bundles with integer exponents. The equalities
(36.1) H o(T *) = V *, H o(T ) = {0}
for the tautological line bundle
T over a complex
projective space P(V), the first of which is a natural
isomorphic identification associating with a linear functional on
V the family of its restrictions to one-dimensional
vector subspaces of V.
Homework #36.
- APRIL 11. Proof of the remaining left-to-right inclusion
in (36.1) (using only Liouvilles theorem, rather than the Hartogs
extension theorem): holomorphic sections of
T * are
naturally identified with degree-one homogeneous holomorphic functions
φ : V − {0} → C, while for such φ
and v, w ∊ V the
holomorphic functions
dv φ and
dw dv φ are homogeneous of degree 0 and,
respectively, of degree −1, so that the latter tends to
0 and infinity and hence, restricted to an complex affine line not
containing 0, must be identically zero by Liouvilles
theorem. Holomorphic subbundles. Quotient bundles. The determinant bundle. The
(anti)canonical bundle of an almost complex manifold. The kernel and image of
a constant-rank holomorphic vector-bundle morphism. Pullbacks of holomorphic
vector bundles under holomorphic mappings, including their restrictions to
complex submanifolds. The normal bundle of a
complex submanifold. The natural structure of a complex manifold on the total
space of a holomorphic complex vector bundle over a complex manifold.
Holomorphic sections as complex submanifolds of the total space, including the
zero section, always identified with the base manifold. The tautological line
bundle
T over a complex
projective space P(V) as a subbundle of the product
bundle
P = P(V) × V, for any finite-dimensional
complex vector space V. The natural isomorphic
identifications
(37.1) T [P(V)] = Hom(T, P ⁄ T ), T ⊗m = det T*[P(V)].
Homework #37.
- APRIL 13. The Picard group Pic(M)
of a complex manifold M, formed by the isomorphism
classes of holomorphic line bundles over M with the
Abelian-group operation induced by the tensor product. Riemann surfaces.
Divisors on a closed Riemann surface S, defined to be
integer-valued functions
D on
S such that the support
S − D−1(0) is finite. The additive group
Div(S) of all divisors on S, the
degree (sum of values) homomorphism
deg : Div(S) → Z,
and the subgroup Pr(S) of all principal divisors
DF
on S, naturally associated with nonconstant holomorphic
mappings
F : S → CP1.
The divisor class group
Cl(S) = Div(S) ⁄ Pr(S). The convention that
(38.1) C is a
subset of CP1 = P(C2), namely, CP1 − {∞},
where ∞ = C(1, 0) (the span of
(1, 0)) and z ∊ C
is identified with C(z, 1). The group
homomorphism
(38.2) Div(S) → Pic(S)
uniquely determined by the requirement that for each
x ∊ Q it send the generator
D[x] ∊ Div(S) having the support
{x} and
D[x](x) = 1 to the isomorphism class of the holomorphic line
bundle D over
S defined as in (33.2) - (33.3) with
M = S and
Q = {x}. The fact that
(38.3) Pr(S) is contained in the kernel of deg.
Homework #38.
- APRIL 16. The degree of a nonconstant holomorphic
mapping
F : S → Q
between closed Riemann surfaces, defined to be the number of
F-preimages, counted with multiplicities, of any
y ∊ Q. Assertion (38.3) as a
trivial consequence of the fact that the degree does not dependent on the
choice of y. The equality between the degree just defined
and the ordinary mapping degree, which uses the oriented-integral isomorphism
Hn(M, R) → R for any (connected) compact oriented
n-dimensional manifold M (and a simple
proof of the latter for a closed Riemann surface
M = S is based on the Stokes theorem
[DG
p. 166]: if μ denotes the area (volume) form
of a fixed Riemannian metric g on S
and
f ∊ FS
is a smooth function
S → R having
favg = 0, then for
θ ∊ FS with
Δθ = f we obtain
2i∂∂θ = f Ω = f μ
due to (11.3) and two-dimensionality, so that
f μ is exact; note that
Ω = μ as
Ωx (u, Ju) = gx (Ju, Ju) = 1
for any x ∊ S and any unit
vector u tangent to S at
x). Holomorphic (local) sections of the line bundle over
S arising from a divisor
D as in (38.2),
identified with (local) holomorphic mappings into
CP1 with the order of zero, at each point
x, not less than
D(x) = 1 (where poles are treated as zeros of negative order). Triviality of the
line bundles associated with a principal divisor
DF
on S, a global trivializing section being
F itself.
Homework #39.
- APRIL 18. The natural biholomorphic identification
(40.1) T * = P(V × C) − {{0} × C}
for the tautological line bundle
T over a complex
projective space P(V), obtained by assigning the graph of
ξ : Λ → C to any pair
(Λ, ξ) ∊ T *, with
Λ ∊ P(V). The holomorphic
Gauss mapping
G : B → V for a holomorphic vector subbundle of the
product bundle
P = M × V over a complex manifold
M, where V is any finite-dimensional
complex vector space. Example:
G : T → V for
T, V as
above. The restriction biholomorphism
G : T − P(V) → V − {0}. Holomorphic fibre bundles, including the
projectivization P(B),
with the fibres
(40.2) [P(B)]x = P(Bx)
for x ∊ M, and
(40.3) the projective
compactification P(B ⊕ [M × C])
of a holomorphic vector bundle
B over a complex
manifold M. The biholomorphism
W − {0} → W* − {0} for any one-dimensional complex
vector space W, given by
(40.4) w → w −1, with w −1(w) = 1.
Gluing of two complex manifolds with the aid of a biholomorphism
between a pair of their open submanifolds, leading to a new complex manifold
as long as the Hausdorff axiom is satisfied. The observation that, for the
total space of the projective
compactification of a given complex line bundle
L over a complex
manifold M,
(40.5) P(L ⊕ [M × C]) arises from the fibrewise gluing of L and L*,
via (40.4), along the complements of the zero section M.
The resulting biholomorphic identification
(40.6) P(L ⊕ [M × C]) = P(L* ⊕ [M × C])
between the projective compactifications of
L and
L*.
Homework #40.
- APRIL 20. The blow-up of a complex manifold
M at a point x, defined to be the
complex manifold N with the underlying set
(41.1) N = (M − {x}) ∪ P(Tx M)
and with the complex manifold structure obtained by choosing a biholomorphism
F : B → U
between a neighborhood B of 0 (a
ball) in Tx M,
and a neighborhood U of x in
M such that
F(0) = x and
dF0 = Id, and then gluing the total
space T of the
tautological line bundle over
P(Tx M) to
M using the following biholomorphism between open
subsets:
(41.2) F º G : G−1(B − {0}) → U − {x}.
Independence of the resulting complex structure of the choice of such
B, U and F. The fact
that, for the total space
T of the tautological
line bundle over P(V), where V is
any finite-dimensional complex vector space,
(41.3) T is the blow-up of V at the point 0,
since the gluing as above with
M = U = B = V = Tx M
and F = Id realizes
M − {x} = V − {0} as a subset of T (namely, T − P(V)). The local nature of blow-ups:
(41.4) the blow-up of U at x is open in the
blow-up of M at x
whenever U is an open submanifold of a complex manifold
M and x ∊ U.
The natural biholomorphic identification
(41.5) [the blow-up of P(V × C) at the point {0} × C] = P(T ⊕ [M × C])
Proof of (41.5) obtained by combining (40.5) for
L = T
with (40.1) and (41.3) - (41.4), where V is identified
with the image of V × {1} under the
holomorphic projection mapping
(V × C) − {(0, 0)} → P(V × C).
Homework #41.
- APRIL 23. Elliptic complexes and the Hodge-de Rham
decomposition theorem. The case of differential forms. Hodges theorem
and the Hodge decomposition for Kähler manifolds [Zhengs book:
Section 8.1]. Ampleness and very-ampleness of holomorphic line bundles over a
complex manifolds [Zhengs book: Section 5.4], and their generalizations
to holomorphic vector bundles of any fibre dimension [Zhengs book:
Section 8.4]. The Chern connection of a Hermitian fibre metric on a
holomorphic vector bundle [Zhengs book: Section 7.3]. Positive line
bundles and the Kodaira embedding theorem [Zhengs book: Section 8.3].
Homework #42.