Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

RICCI-ITERATIONS OF WELL-BEHAVED KÄHLER METRICS

A. LOI AND G. PLACINI


arXiv:2307.11500v1 [math.DG] 21 Jul 2023

A BSTRACT. We introduce a large class of canonical Kähler metrics, called in this paper well-
behaved, extending metrics induced by complex space forms. We study Kähler–Ricci iterations
of well-behaved metrics on compact and non-compact Kähler manifolds. That is, we are interested
in well-behaved metrics for which the iteration of the Ricci operator is a multiple of a Kähler metric,
i.e., ρkω = λΩ. In particular, when k = 1, under some condition on the maximal domain of defi-
nition of canonical coordinates, we show that λ is forced to be positive. Moreover, for arbitrary k,
we prove two additional results. Namely, if ω and Ω are induced by a flat metric, then ω is Ricci-
flat. Finally, if a Kähler-Ricci soliton Ω arises as Kähler–Ricci iteration of a metric ω induced by a
complex space form, then the Kähler–Ricci soliton is forced to be trivial, that is, Kähler–Einstein.
These three theorems extend well known results on Kähler–Einstein metrics to higher iterations of
the Ricci operator and a larger class of metrics.

C ONTENTS
1. Introduction and main results 1
2. Well-behaved Kähler metrics 6
3. Proof of Theorem 1 and a toy example 12
4. Proofs of Theorem 2 and Theorem 3 16
References 19

1. I NTRODUCTION AND MAIN RESULTS

Yau’s solution [42] of Calabi’s conjecture is a cornerstone in the study of complex Monge-
Ampère equations. It shows that, given a Kähler class α on a compact complex manifold M, any
form representing 2πc1 (M) is the Ricci form of a unique Kähler form ω ∈ α. Namely, if ρω
denotes the Ricci form of ω, the complex Monge-Ampère equation
(1) ρω = λΩ
where [λΩ] = 2πc1 (M) admits a unique solution. Solving similar complex Monge-Ampère equa-
tions is a central problem in Kähler geometry, see [3, 9, 10, 11, 39] among many. A notable case
is when λω ∈ 2πc1 (M). Then equation (1) is the Einstein equation ρω = λω and its solutions are
Kähler–Einstein metrics. In this case, there are obstructions to the existence of Kähler–Einstein
Date: July 24, 2023; ©A. Loi and G. Placini 2023.
2010 Mathematics Subject Classification. 32W20; 32Q15; 53C42; 32Q20.
Key words and phrases. Kähler–Ricci iterations, Diastasis function, Bochner coordinates, Kähler–Einstein metrics,
Compactifications of Cn .
The authors are supported by INdAM and GNSAGA - Gruppo Nazionale per le Strutture Algebriche, Geometriche e
le loro Applicazioni and by GOACT - Funded by Fondazione di Sardegna. We acknowledge financial support by PNRR
e.INS Ecosystem of Innovation for Next Generation Sardinia (CUP F53C22000430001, codice MUR ECS00000038).
1
metrics when c1 (M) > 0, cf. [16, 38]. It is very natural to study solutions of equation (1) and
obstructions to their existence when ω and Ω satisfy additional hypotheses. As notable examples, if
ω is a Kähler–Einstein metric induced by a non-elliptic complex space form S, then M is a totally
geodesic submanifold in S [40] while λ > 0 if ω is a projectively induced Kähler–Einstein metric
[18].
The first part of this paper is concerned with the study of equation (1) when M is not necessarily
compact and ω belongs to a large class of Kähler metrics including metrics induced by complex
space forms or, more broadly, by generalized flag manifolds or toric manifolds. In particular we
find that, under certain conditions on the Kähler potentials of the metrics ω and Ω, one necessarily
has λ > 0. In order to state such hypotheses and our first result, let us recall some known facts on
Kähler metrics.
Given a real analytic Kähler metric g on a complex manifold M, one can introduce a very
special Kähler potential Dpg for the metric g in a neighbourhood of a point p ∈ M, which Calabi
christened diastasis [8]. Moreover, one can always find local Bochner coordinates centred at p such
that the series expansion of the diastasis function Dpg has no terms of degree < 2. See Section 2
and references therein for details on the diastasis function and Bochner coordinates. We can now
introduce the class of Kähler metrics we are interested in, see Section 2 for a detailed discussion.
Definition 1. We call a real analytic Kähler metric g on a n-dimensional complex manifold M
well-behaved at p ∈ M if there exists a measure zero (possibly singular or empty) real submanifold
Hpg of M such that the following conditions are satisfied.
(A) Calabi’s diastasis function Dpg around p is globally defined and non-negative on M \ Hpg ;
(B) the Bochner coordinates (z) = {z1 , . . . zn } around p extend to holomorphic functions
f1 , . . . , fn on M \ Hpg .
Remark 1. It is not clear to us whether the definition of well-behaved Kähler metric does depend
on the point p. In the compact case we even lack examples of metrics which are not well-behaved at
some point, see Remark 4. In the non-compact case, where examples of such metrics are available
(cf. Example 11), it is non-trivial to check whether they are well-behaved at some point. On the
other hand, we see no reason, other than lack of examples, that suggests independence on the point.
Notice that conditions (A) and (B) are always satisfied on a small enough neighbourhood U
of any point p where Bochner coordinates are defined and Dpg is positive (except at p where it
vanishes) by setting Hpg = ∅. Thus, roughly speaking, a Kähler metric g is well-behaved at a point
p ∈ M, if U can be enlarged to the dense open subset M \ Hpg in such a way that Dpg remains
non-negative and Bochner coordinates extend to global holomorphic functions (not necessarily
coordinates). Obviously Hpg 6= ∅ if M is compact.
In contrast with this, we will require a global condition on the metrics. Such condition was first
considered in [1], see also [2].
Definition 2. Let M be a n-dimensional complex manifold and g be a well-behaved Kähler metric
at a point p ∈ M. We define the Bochner-Euclidean volume of g at p to be
Z
g
(2) voleucl (M \ Hp , g) := dµ(f )
M \Hpg

where f1 , . . . , fn are as in Definition 1 and dµ(f ) := in


df
2n 1
∧ df¯1 ∧ · · · dfn ∧ df¯n .
2
Notice that voleucl (M \ Hpg , g) strongly depends on the Bochner coordinates and so on the met-
ric g. One can think of the Bochner–Euclidean volume as the volume of M \ Hpg with respect
of the pull-back of the standard Lebesgue measure of Cn via the map M \ Hpg → Cn , q 7→
(f1 (q), . . . , fn (q)).
We are now ready to state our first result, Theorem 1. Although its hypotheses appear technical
at first glance, they are satisfied by a large class of Kähler metrics, see Section 2 for several exam-
ples. In fact, when specialized to certain canonical metrics, Theorem 1 yields several corollaries
(Corollary 1-3 below) which extend well known results.
Theorem 1. Let M be a complex n-dimensional manifold and let g, G be two well-behaved Kähler
metrics at the same point p ∈ M satisfying
ρω = λΩ,
where ω and Ω are the Kähler form associated to g and G respectively. Assume that g and G
satisfy the following conditions:
n
(i) vol(M, ω) = M ωn! < ∞;
R
(ii) voleucl (M \ Hpg , g) = ∞;
(iii) codimR (Hpg ) ≥ 2.
with Hpg as in Definition 1 . Assume that also G is well-behaved at p and
(iv) codimR (HpG ) ≥ 2.
Then λ > 0.
The simplest case, which motivated Theorem 1, is that of compact homogeneous Kähler mani-
folds. Indeed, the Ricci tensor of a homogeneous Kähler metric is always a homogeneous Kähler–
Einstein metric, cf. Proposition 2 below. If in addition the homogeneous Kähler form is integral,
it is projectively induced (see [35, 30]). To the best of our knowledge, this is the only example of
projectively induced Kähler forms ω, Ω satisfying the hypotheses of Theorem 1. Inspired by this
case we study projectively induced (radial) metrics g on CP1 and show in Proposition 3 that their
Ricci tensor is projectively induced if and only if it is Kähler–Einstein. Moreover, we give a very
explicit example: Ricci forms of well-behaved metrics on CP1 , see Example 12. In particular,
the positivity of the first Chern class allows us to construct well-behaved, not projectively induced
metrics g and G on CP1 which satisfy the hypotheses of Theorem 1.
Some comments on the necessity of the hypotheses (i)-(iii) are in order.
- Assumption (i) is a necessary condition for Theorem 1 to be true. For example the hyperbolic
metric ghyp on the complex hyperbolic space is well-behaved and satisfies conditions (ii) and (iii)
(see Example 3) and being Kähler-Einstein it satisfies (1) with g = G and λ = −2(n + 1) < 0.
- Also assumption (ii) is necessary. Take for example any bounded domain in Cn with the flat met-
ric. Examples of compact manifolds not satisfying (ii) nor (iii) are complex tori with the flat metric
(Example 9) or Riemann surfaces with the hyperbolic metric and products thereof (Example 10).
- Concerning the necessity of property (iii), we do not know of any compact Kähler manifold sat-
isfying (i) and (ii) but not (iii), see Examples 9 and 10 below. In particular, it would be interesting
to understand the interplay between conditions (ii) and (iii).
Notice that the assumptions (i)-(iii) of Theorem 1 are satisfied, for instance, by the complex
projective space CPn . Moreover, properties (i)-(iii) are hereditary for projective varieties or, more
generally, compact Kähler submanifolds of certain compactifications of Cn , see Proposition 1.
From this we obtain the following results.
3
Corollary 1. Let M be a compact complex manifold endowed with two projectively induced Kähler
forms ω, Ω satisfying ρω = λΩ. Then M is Fano.
Corollary 2. Let M be a compact complex manifold equipped with two Kähler forms ω, Ω induced
by immersions into classical flag manifolds. If ρω = λΩ, then M is Fano.
Observe that Corollary 1 is in fact implied by Corollary 2. On the other hand, not all classical
flag manifolds are projectively induced, even among irreducible ones and up to homothety, cf.
[27].
Corollary 3. Let M be a compact complex manifold equipped with two Kähler forms ω, Ω induced
by immersions ϕ1 , ϕ2 into a toric manifold such that ϕ1 (p) = ϕ2 (p) is fixed by the torus action
for some p ∈ M. If ρω = λΩ, then M is Fano.
As mentioned above, the Einstein equation ρω = λω is a special case of equation (1). In light of
this, if ω = Ω in Corollary 1 and Corollary 3 we recover some known results on Kähler–Einstein
submanifolds. Namely, the main theorem of [18] and [2, Theorem 2.6] respectively. In fact, the
proof of Theorem 1 is inspired by these results and their proofs.
One also derives the following consequence for a Kähler-Einstein metric g on a compact com-
plex manifold M. If c1 (M) < 0 or c1 (M) = 0, the existence of Kähler-Einstein metrics is
guaranteed by [3] and [42], but explicit examples are extremely rare. By Theorem 1 such metrics
necessarily satisfy at least one of the following conditions for a point p ∈ M:
• g is not well-behaved at p;
• voleucl (M \ Hpg , g) < ∞;
• codimR (Hpg ) < 2.
Let us now comment on the necessity of the condition of well-behaviour of the metrics g and
G in Theorem 1. We observe that requiring both g and G to be well-behaved is necessary in the
compact case. For instance, one can consider a compact complex manifold with c1 (M) < 0 and
a Kähler form Ω such that λΩ represents the first Chern class (up to a 2πi factor). By Calabi’s
conjecture, in a given Kähler class there exists a unique Kähler form ω whose Ricci form ρω equals
λΩ. By Tian’s theorem [37], if we choose the class [ω] to be integral, we can approximate the form
ω with (multiples of) projectively induced metrics ωk . For k large enough ρωk /λ is still positive,
hence a Kähler form. We then have projectively induced Kähler forms ωk satisfying
ρωk = λΩk
for Kähler forms Ωk which cannot be well-behaved by Theorem 1. In fact one can see that the
necessary condition is weaker than G being well-behaved, cf. Corollary 4.
Analogously, if we pick λ ∈ 2πZ we can find projectively induced metrics Ωk satisfying
ρωk = λΩk
for Kähler forms ωk which cannot be well-behaved by Theorem 1.
In the second part of this paper we focus on a generalization of equation (1) related to the so-
called Kähler–Ricci iterations. Namely, if ±ρω is again a Kähler metric we can consider its Ricci
form ρ2ω or, more generally, we can define inductively
(
ω, if k = 0
(3) ρkω :=
ραk−1 , if αl := ±ρlω > 0, for all l < k.
4
This iteration was introduced by Nadel in [31] where he also proved that periodic points of order
two or three must be Kähler-Einstein metrics. This was generalized [20, 33] to periodic points ω
of any order, that is, those satisfying ρkω = λω for any k ∈ N. The iteration (3) can be regarded as
a discretization of the Kähler–Ricci flow. Observe that, if the manifold M is compact, by Calabi’s
conjecture one does not need positivity assumptions to reverse the construction above and define
the inverse Ricci–Kähler iterations ρ−kω . Both of this discrete dynamical systems have been studied
in the literature, see for instance [5, 13].
Here we focus on the following generalized Monge-Ampère equation on a complex manifold
M
(4) ρkω = λΩ
where Ω is again a Kähler form. In particular we study equation (4) for a special class of well-
behaved Kähler metrics. Our second result is the following theorem dealing with Kähler metrics
induced by the flat metric (which are well-behaved by Example 4 below).
Theorem 2. Let M be a complex manifold with two metrics g and G induced by the flat metric. If
the corresponding Kähler forms ω, Ω satisfy ρkω = λΩ for some k ≥ 1 and λ ∈ R, then (M, g) is a
totally geodesic submanifold of the flat ambient space.
Observe that this result can be seen as a generalization of [40, Theorem 2.1]. Namely, when
k = 1 and g = G, Theorem 2 shows that a Kähler–Einstein submanifolds of Cn with the flat
metric is necessarily Ricci-flat, hence a totally geodesic submanifold. It is worth pointing out that
the metric g is assumed to be induced by a flat one for simplicity, but the hypothesis on g can be
sensibly relaxed, cf. Remark 11 below.
Our third and last result deals with Kähler–Ricci solitons (KRS). Recall that a Kähler metric G
is a KRS if there exists a holomorphic vector field X (the solitonic vector field) such that Ric(G) =
µG + LX G where Ric(G) denotes the Ricci tensor of G and LX the Lie derivative in the direction
of X. Kähler–Ricci solitons are important generalizations of Kähler–Einstein metrics which arise
in the study of the Kähler–Ricci flow. Our next result show that KRS cannot arise as Kähler–Ricci
iterations of metrics induced by complex space forms.
Theorem 3. Let M be a complex manifold with two real analytic Kähler metrics g and G. Assume
that g is induced by a complex space form and G is a KRS. If the corresponding Kähler forms ω, Ω
satisfy ρkω = λΩ for some λ 6= 0 and k ≥ 0, then G is trivial, i.e. Kähler–Einstein.
Observe that, when λ = 0, one cannot draw any conclusion. Take for instance g to be the flat
metric on C and G to be Hamilton’s cigar KRS [17]. It is worth mentioning that we do not know of
any Kähler–Einstein metric G and Kähler metric g induced by a complex space form which satisfy
ρkω = λΩ for λ < 0 unless g = G and k = 1. In that case M is forced to be a totally geodesic
submanifold in CHn [40]. On the other hand, such examples for λ > 0 are discussed in Section 3,
see in particular Proposition 2.
Also in this case we generalize some recent results on KRS induced by complex space forms
which indeed served as partial motivation for this paper. In particular, if we assume k = 0 (and
λ > 0) in Theorem 3, we recover [25, Theorem 1.1]. If we restrict to compact complex manifolds,
then this result can be rephrased in terms of the inverse Kähler–Ricci iterations. Namely, Theo-
rem 3 shows that none of the inverse Kähler–Ricci iterations ρ−k
ω of a non-trivial KRS (g, X) on a
compact complex manifold M can be induced by a complex space form.
5
Finally, observe that the metric g is assumed to be induced by a complex space form, but the
proof of the theorem shows that this hypothesis can be relaxed, cf. Remark 11 below.
Organization of the paper. In Section 2 we discuss well-behaved Kähler metrics as well as the
conditions (i)-(iii) of Theorem 1. Moreover, we prove that metrics induced by suitable compact-
ifications of Cn are both well-behaved and satisfy conditions (i)-(iii), see Proposition 1. This is
crucial in the proofs of Corollaries 1-3 which complete Section 2. In Section 3 we give the proof of
Theorem1 and a discussion on Kähler–Ricci iterations on compact homogenous Kähler manifolds.
In particular, we give a very explicit example: Kähler–Ricci iterations of well-behaved metrics on
CP1 . Finally, Section 4 is devoted to the proofs of Theorem 2 and Theorem 3.

2. W ELL - BEHAVED K ÄHLER METRICS

In order to discuss well-behaved metrics, we briefly recall some basic facts on Calabi’s diastasis
functions and Bochner coordinates. Given a complex manifold M endowed with a real analytic
Kähler metric g, one can introduce on a neighbourhood of a point p ∈ M, a very special Kähler
potential Dpg for the metric g, which Calabi christened diastasis. Recall that a Kähler potential is
an analytic function Φ defined in a neighbourhood of a point p such that ω = 2i ∂ ∂Φ, ¯ where ω is
the Kähler form associated to g. In a complex coordinate system (z) = {z1 , . . . , zn } around p one
has
∂ ∂ ∂2Φ
(5) gαβ̄ = 2g( , )= .
∂zα ∂ z̄β ∂zα ∂ z̄β
A Kähler potential is not unique: it is defined up to summing with the real part of a holomorphic
function. By duplicating the variables z and z̄, a potential Φ can be complex analytically continued
to a function Φ̃ defined on a neighbourhood U ⊂ M ×M of the diagonal containing (p, p̄) (here M
denotes the conjugated manifold of M). The diastasis function is the Kähler potential Dpg around
p defined by
Dpg (q) = Φ̃(q, q̄) + Φ̃(p, p̄) − Φ̃(p, q̄) − Φ̃(q, p̄).
The diastasis function is characterized among potentials by the property that in every coordinates
system (z) centred in p
X
Dpg (z, z̄) = ajk z j z̄ k ,
|j|,|k|≥0

with aj0 = a0j = 0 for all multi-indices j. More generally, for later use, we give the following (see
also [25, 26])
Definition 3. Let M be a complex manifold and f : U → R a real analytic function defined on
a neighbourhood U of a point p ∈ M. We say that f is of diastasis type if in one (and hence
any) coordinate system {z1 , . . . , zn } centred at p the expansion of f in z and z̄ does not contain
non-constant purely holomorphic or anti-holomorphic terms (i.e. of the form z j or z̄ j with |j| > 0).
Remark 2. By its very definition Calabi’s diastasis function Dpg is of diastasis type.
Remark 3. Calabi’s procedure can be applied to any (not necessarily Kähler) real-analytic real-
valued closed form of type (1, 1). Namely, if Γ is a such a form, on a neighbourhood of a point
¯ with γ of diastasis type.
p ∈ M one can write Γ = 2i ∂ ∂γ,
6
One can always find local complex coordinates (z) centred at p such that
X
Dpg (z, z̄) = |z|2 + bjk z j z̄ k .
|j|,|k|≥2

These coordinates, called Bochner coordinates around p, are uniquely defined up to a unitary
transformation (cf. [6, 8] or [29] for an updated account on the subject). The following result
summarizes some fundamental properties of Calabi’s diastasis function and Bochner coordinates.
Theorem 4 ([8]). Let ϕ : (S, h) → (M, g) be a holomorphic isometric immersion between Kähler
manifolds of complex dimension m and n respectively. If g is real analytic, then h is real analytic
and for every point q ∈ S
g
Dqh = ϕ∗ Dϕ(q) .
Moreover, if (w1 , . . . , wm ) is a system of Bochner coordinates in a neighbourhood U of q, then
there exists a system of Bochner coordinates (z1 , . . . , zn ) centred at ϕ(q) such that
(6) z1 ◦ ϕ|U = w1 , . . . , zm ◦ ϕ|U = wm .
We provide now several examples of well-behaved Kähler metrics both on non-compact and
compact complex manifolds aimed at showing that they are fairly common among real analytic
Kähler metrics.
Example 1. (Products and linear combinations) Let M be any complex manifold. It is not hard
to verify that if gj , j = 1, . . . , k, are well-behaved Kähler metrics at a point p ∈ M and αj are
positive real numbers such that g = kj=1 αj gj is still a Kähler metric, then g satisfies property
P
(A) in Definition 1. On the other hand, we cannot describe the Bochner coordinates of g in terms
of those of the metrics gj . Thus, we do not have any information on whether linear combinations
of well-behaved metrics satisfy condition (B) in Definition 1.
A different behaviour is exhibited by Kähler products. In fact, if (Mj , gj ) with j = 1, . . . , k are
Kähler manifolds with gj well-behaved at pj ∈ Mj for all j , then the Kähler metric g1 ⊕ · · · ⊕ gk
on M1 ×· · ·×Mk is well-behaved at (p1 , . . . , pk ). Moreover, it is easy to see that products preserve
assumptions (i)-(iii) in Theorem 1.
Example 2. (Hereditary property) By Theorem 4 we get many examples of well-behaved Kähler
metrics induced by well-behaved ones. More precisely, suppose that ϕ : S → M is a holomorphic
isometric immersion between Kähler manifolds (S, h) and (M, g). If g is well-behaved at p = ϕ(q)
for some q ∈ S, then h is well-behaved at q. Moreover, we have Hqh = ϕ−1 (Hpg ).
It is easy to exhibit examples showing that conditions (i) and (ii) are not hereditary, see also
Remark 6. Therefore, if (M, g) is a Kähler manifold satisfying conditions (i)-(iii) in Theorem 1,
we cannot conclude that a Kähler submanifold (S, h) inherits those properties.
Example 3. (Hyperbolic disc) The hyperbolic metric ghyp on CHn = {z = (z1 , . . . zn ) | ∈
Cn | |z|2 < 1} is well-behaved at any point of CH n . This can be verified at the origin 0 ∈ CH n .
Indeed, the Kähler form associated to ghyp is given by ωhyp = − 2i ∂ ∂¯ log(1 − |z|2 ) whose diastasis
g
at the origin 0 ∈ Cn reads as D0hyp (z) = − log(1 − |z|2 ). Hence, the Bochner coordinates
g
are the restriction of the Euclidean coordinates to CH n and H0 hyp = ∅. This shows that ghyp
is well-behaved and that its Euclidean volume voleucl (CHn , ghyp ) is finite although the volume
vol(CHn , ghyp ) is infinite. More generally, one can find in [14] a description of Calabi’s diastasis
function and Bochner coordinates for bounded symmetric domains with their Bergman metrics
which yields the same conclusion.
7
Example 4. (Flat Cn ) The flat metric g0 on Cn is well-behaved at any point of Cn . This too
can be easily verified at the origin 0 ∈ Cn . Indeed, the Kähler form associated to g0 is given by
¯ 2 whose diastasis at the origin reads as D g0 (z) = |z|2 . Thus, the Bochner coordinates
ω0 = 2i ∂ ∂|z| 0
are the Euclidean coordinates and H0g0 = ∅. It follows by Example 2 that the Kähler metric g
induced on a Stein manifold M by the flat metric through an embedding into a complex Euclidean
space is indeed well-behaved at any p ∈ M with Hpg = ∅. Clearly, the Bochner-Euclidean volume
coincides with the Kähler volume in this case and they are infinite.
Example 5. (Fubini-Study metric) Let gF S be the Fubini-Study metric of CP n whose associated
Kähler form is given in homogeneous coordinates by ωF S = 2i ∂ ∂¯ log(|Z0 |2 + · · · + |Zn |2 ). It is
easy to see that the diastasis at the point p0 = [1, 0, . . . , 0] is globally defined on CP n \ HpgF S ,
Z
where Hpg0F S is the hyperplane {Z0 = 0}. Moreover, the affine coordinates zj = Z0j , j = 1, . . . n
are indeed global Bochner coordinates on CP n \ HpgF S because the diastasis in these coordinates is
given by Dpg0F S (z) = log(1+|z|2 ). Hence gF S is well-behaved at p0 (in fact at all points because ωF S
is homogeneous). It is then clear that the Bochner-Euclidean volume voleucl (CPn \ HpgF S , gF S ) is
infinite while vol(CPn , gF S ) is clearly finite. Thus (CPn , gF S ) is the simplest example of compact
Kähler manifold with a well-behaved metric which satisfies (i)-(iii) of Theorem 1.
Example 6. (Projectively induced metrics) A Kähler metric g on a (not necessarily compact)
complex manifold M is said to be projectively induced if there exists a holomorphic immersion
ϕ : (M, g) → (CP N ≤∞ , gF S ) such that ϕ∗ gF S = g (see [8]). Here CP N ≤∞ denotes the finite
or infinite dimensional complex projective space equipped with the Fubini-Study metric gF S . It
is not hard to see that the second part of Calabi Theorem 4 still holds when the ambient space is
the infinite dimensional complex projective space (CP ∞ , gF S ). Therefore, as in Example 2, one
sees that a projectively induced metric g on a complex manifold M is well-behaved at any point
p ∈ M. Many examples of compact and non-compact projectively induced Kähler metrics have
been considered in the literature (see [29] and references therein). Moreover, when N < ∞ and
M is compact, a projectively induced metric g on M has infinite Bochner–Euclidean volume (and
finite volume), cf. Proposition 1 below.
In the following we deal with two examples of well-behaved Kähler metrics on compactifica-
tions of Cn which are not necessarily projectively induced.
Example 7. (Generalized flag manifolds) Let (M, g) be a generalized flag manifold, that is, a
compact and simply-connected homogeneous Kähler manifold. If the Kähler form associated to g
(respectively one of its multiples) is integral then the Kähler metric g (resp. its integral multiple)
is projectively induced (see [30]) and hence well-behaved at any point by Example 6. Typical
examples are homogeneous Kähler metrics of Hermitian symmetric spaces of compact type or
Kähler-Einstein metrics on generalized flag manifolds. On the other hand, there exist generalized
flag manifolds (M, g) where g and all its multiples
√ are not integral, but g is well-behaved. For
1 1
example (CP × CP , g), where g = gF S ⊕ 2gF S is well-behaved by Example 1.
More generally, there exists a large class of irreducible classical generalized flag manifolds
(M, g) with g well-behaved and not necessarily projectively induced (see [28, Theorem 1]). Such
manifolds are compactifications of Cn , namely M = Cn ⊔ H, where H is a complex (possibly
singular) submanifold of complex codimension one. Moreover, the diastasis function D0g at 0 ∈ Cn
is defined on Cn and positive away from 0 and the Bochner coordinates are globally defined on
Cn . This shows that the Euclidean volume voleucl (M \ H, g) is infinite while, by compactness
8
vol(M, g) < ∞. Therefore, such classical generalized flag manifolds satisfy conditions (i)-(iii) of
Theorem 1.
Notice that for the Hermitian symmetric spaces of compact type Hpg is in fact the cut locus
with respect to the point p (see for instance [23, 24, 36]). We believe (see also [28, Conjecture
p. 499]) that this property should characterize Hermitian symmetric spaces among generalized flag
manifolds.
Example 8. (Toric manifolds) Recall that a toric manifold M is a complex manifold which con-
tains an open dense subset biholomorphic to (C∗ )n and such that the canonical action of (C∗ )n on
itself extends to a holomorphic action on the whole manifold M. A toric Kähler metric ω on M is a
Kähler metric which is invariant under the action of the real torus T n = {(eiθ1 , . . . , eiθn ) | θi ∈ R}
contained in the complex torus (C∗ )n . That is, for every fixed θ ∈ T n the diffeomorphism
fθ : M → M given by the action of (eiθ1 , . . . , eiθn ) is an isometry.
We have the following well known fact on toric manifolds, see for example [15, Section 2.2.1],
[4, Proposition 2.18] or refer to [2, Appendix] for a self-contained proof. If M is a projective,
compact toric manifold, then there exists an open dense subset X ⊆ M which is algebraically
biholomorphic to Cn . More precisely, for every point p ∈ M fixed by the torus action there are
an open dense neighbourhood Xp of p and a biholomorphism φ : Xp → Cn such that p is sent to
the origin and the restriction of the torus action to Xp corresponds via φ to the canonical action
of (C∗ )n on Cn . Moreover, one can show [2, Theorem 2.6] that the diastasis function Dpg of the
toric metric g is defined and positive on Xp and that the coordinates on Xp given by φ are Bochner
coordinates for g. Therefore invariant metrics on toric manifolds are well-behaved and have infinite
Bochner–Euclidean volume. Since they have finite volume by compactness, toric metrics satisfy
conditions (i)-(iii) of Theorem 1.
Example 9. (Flat torus) Let Tn = Cn /Z2n be the n-dimensional complex torus endowed with the
flat metric g0 . Let
1 1 1 1
M0 = {(z1 , . . . zn ) ∈ Cn | − < ℑ(zj ) < , − < ℜ(zj ) < , j = 1, . . . , n}
2 2 2 2
be the fundamental domain of the Z -action containing the origin 0 ∈ Cn where ℜ(z) and ℑ(z)
2n

are the real and imaginary part of z ∈ C respectively. The diastasis D0g0 of g0 with respect to 0 reads
as D0g0 (z, z̄) = |z|2 = nj=1 |zj |2 and the Euclidean coordinates (z1 , . . . , zn ) are in fact Bochner
P

coordinates for g0 around 0. Notice that M0 = M \ H0g0 is the maximal domain of extension of
D0g0 so that the Bochner–Euclidean volume voleucl (M \ H0g0 , g0 ) is finite. Moreover, H0g0 is given
by
n   n  
g0
[
n 1 [ [ n 1
H0 = (z1 , . . . zn ) ∈ T | ℜ(zj ) = (z1 , . . . zn ) ∈ T | ℑ(zj ) =
j=1
2 j=1
2

which is a singular submanifold of M of real codimension 1. Hence g0 is well-behaved but not


projectively induced by Example 6 (or even locally projectively induced by [8]).
Example 10. (Riemann surfaces) The hyperbolic metric ghyp on a compact Riemann surface Σg of
g
genus g ≥ 2 is well-behaved. Indeed, consider Σg = CH1 /Γg where Γg = π1 (Σg ) and let D0hyp be
the diastasis function at 0 ∈ CH1 . Then its maximal domain of definition M0 is the fundamental
g g
domain of the Γg -action and satisfies M0 = Σg \ H0 hyp where H0 hyp is a bouquet of 2g circles.
Also in this case the euclidean coordinate z is the Bochner coordinate for ghyp around 0. Hence
9
ghyp is well-behaved, but does not satisfy condition (ii) in Theorem 1 because its Euclidean volume
g
is finite. Moreover, ghyp does not satisfy property (iii) because H0 hyp has real codimension 1 in Σg .
Hence ghyp is well-behaved but not projectively induced by Example 3 (or even locally by [8]).
Similar arguments apply to a compact quotient of CH n equipped with the hyperbolic metric
such as products of compact Riemann surface Σgj of genus gj ≥ 2.
Now we exhibit the existence of not well-behaved Kähler metrics on non-compact manifolds.
Example 11. (Metrics which are not well-behaved at a point) A necessary condition for g to satisfy
property (A) of Definition 1 at a point p is that the maximal domain of definition of its diastasis
Dpg at p, say Mp , is dense on M. Simple examples of complete and non-compact Kähler manifolds
where Mp is not dense in M can be found in [34, Theorem 1]. For instance, it is proven in [34]
that there exists a real analytic Kähler metric on C such that M0 is a disk of finite radius. Notice
also that one can exhibit examples of real analytic Kähler metrics on M where, for some p ∈ M,
Dpg is negative at some point of its domain of definition. The first example of such metrics was
constructed by Calabi himself see [8, p. 23].
Remark 4. We do not know any example of a Kähler metric on a compact complex manifold
which is not well-behaved at some point. Moreover, in all known examples with c1 (M) > 0 the
maximal domain of definition of Calabi’s diastasis function of a given Kähler metric at p is given by
the complement of a complex analytic hypersurface Sp of M homologous to the dual of the Kähler
form ω associated to g. Hence in these cases condition (A) in Definition 1 is always achieved by
taking Hpg = Sp . Notice that by Example 6 this condition is satisfied by all projectively induced
Kähler metrics regardless of the sign of c1 (M). Calabi himself (in [8, p.4] comments as: “One
could formulate several conjectures over the behaviour of the diastasis in the large, which in the
author’s opinion would furnish ample material for further studies.”
Remark 5. We also lack examples (compact or non-compact) where condition (A) is valid while
(B) is violated.
Further examples of well-behaved Kähler metrics satisfying conditions (i)-(iii) can be obtained
by the following proposition on compactifications of Cn which will be a key ingredient in the
proofs of Corollary 1-3. By compactification of Cn we mean a compact complex manifold M
containing an analytic subvariety H such that X = M \ H is biholomorphic to Cn . We refer the
reader to [32] for a survey on the topic.
Proposition 1. Let M be a smooth projective compactification of X, such that X is algebraically
biholomorphic to Cn and let g be a Kähler real-analytic metric on M. Assume that there exists a
point p∗ ∈ M where the following conditions are satisfied:
(A′ ) Calabi’s diastasis function Dpg∗ around p∗ is globally defined and non-negative on X;
(B ′ ) the Bochner coordinates (z) = {z1 , . . . zn } around p∗ are globally defined on X.
If ϕ : (S, h) → (M, g) is a holomorphic isometric immersion from a compact Kähler manifold
(S, h) such that ϕ(p) = p∗ , then h is well-behaved and satisfies (i)-(iii) of Theorem 1.
Proof. Condition (A′ ) and (B ′ ) are clearly stronger than (A) and (B) of Definition 1. Hence g is
well-behaved at p∗ . By the hereditary property of well-behaved Kähler metrics (cf. Example 2) h
turns out to be well-behaved at p. Condition (i) is satisfied since M is compact and condition (iii)
is obviously hereditary.
10
The least immediate part of the proof is to show the validity of (ii), namely the fact that the
Bochner-Euclidean volume of h at p is infinite. This follows by a modification of an argument
given in the proof of [2, Proposition 2.3]) valid for Kähler–Einstein submanifolds. Namely, let
w1 , . . . , wm be Bochner coordinates for h defined in a neighbourhood U of p. By Theorem 4, we
can assume that these are given as

z1 ◦ ϕ|U = w1 , . . . , zm ◦ ϕ|U = wm .

Now consider the connected open subset Ŝ := ϕ(S) ∩ X of ϕ(S). Denote by ωg and ωh the Kähler
ωm m
forms associated to g and h respectively. Now the two m-forms m!g and 2i m det(gαβ̄ )dz1 ∧ dz̄1 ∧
· · · ∧ dzm ∧ dz̄m defined on X coincide on ϕ(U) ⊂ Ŝ as they both restrict to the volume form
associated to h. Since they are real analytic, they must agree on the whole connected open set Ŝ.
ωm ωm
Since m!h = ϕ∗ m!g is a volume form on S \ Hph = ϕ−1 (Ŝ), the form dz1 ∧ dz̄1 ∧ · · · ∧ dzm ∧ dz̄m
cannot vanish on Ŝ. We deduce that the restriction to Ŝ of the projection on the first m coordinates
π : Cn −→ Cm is open. Since it is also algebraic (because the biholomorphism between X and
Cn is algebraic), its image π(Ŝ) contains a Zariski open subset of Cm , see [7, Theorem 13.2]. We
conclude that the Bochner–Euclidean volume is infinite because

im
Z
voleucl (S \ Hph , h) ≥ dz1 ∧ dz̄1 ∧ · · · ∧ dzm ∧ dz̄m = ∞.
π(Ŝ) 2m

Remark 6. The compactness assumption of S in Proposition 1 is necessary since (i) of Theorem 1,


unlike (ii) and (iii), is not hereditary for complete Kähler manifold admitting a holomorphic isom-
etry into finite dimensional complex projective space. Take, for example, a complex 2-dimensional
abelian variety T ⊂ CP n and choose a copy of C dense in T . Restrict the Fubini-Study metric
of CP n to T and to C denoting the later by g. Thus (C, g) is a non-compact and complete Kähler
manifold of infinite volume, which admits a holomorphic isometric immersion into (CP n , gF S )
(see [22, Example 6.4, p. 18] for details).

Remark 7. It is worth pointing out that property (ii) is not hereditary when the ambient space is
not a compactification of Cn . Take, for example two complex manifolds M1 and M2 equipped with
two Kähler metrics g1 and g2 which are well-behaved at p1 and p2 respectively. Choose (M1 , g1 )
and (M2 , g2) satisfying voleucl (M1 \ Hpg11 ) < ∞ and voleucl (M2 \ Hpg22 ) = ∞. Then the product
M1 × M2 equipped with the Kähler metric g1 ⊕ g2 satisfies (iii) while g1 (which is induced by the
natural inclusion M1 → M1 × M2 ) does not.

The prototype example of compactifications of Cn satisfying (A′ ) and (B ′ ) is of course the


finite dimensional complex projective space CP n . Indeed, we can write CP n = Cn ⊔ Hpg0F S (cf.
Example 5). Hence, combining Example 6 with Proposition 1 proves Corollary 1.
Notice that by Proposition 1 all the Kähler metrics g in Example 7 and Example 8 respectively
and all their Kähler submanifolds are (well-behaved) at a suitable p and satisfy (i)-(iii) of Theo-
rem 1. Namely, we have proven Corollary 2 and Corollary 3.
11
3. P ROOF OF T HEOREM 1 AND A TOY EXAMPLE

Proof of Theorem 1. Take Bochner coordinates {z1 , . . . , zn } for the metric g on a contractible
neighbourhood U of p. Then
iX i X ∂ 2 Dpg
ω= gαβ̄ dzα ∧ dz̄β = dzα ∧ dz̄β
2 2 ∂zα ∂ z̄β
on U and ρω = −i∂ ∂¯ log det gαβ̄ is the local expression of its Ricci form. From the ∂ ∂-lemma
¯ and
the equation ρω = λΩ it follows that the volume form of (M, g) reads on U as:
 2 g 
ωn ∂ Dp λ G
(7) = det dµ(z) = e− 2 Dp +F +F̄ dµ(z) ,
n! ∂zα ∂ z̄β
n
where dµ(z) := 2i n dz1 ∧ dz̄1 ∧ · · · ∧ dzn ∧ dz̄n , F is a holomorphic function on U and Dpg (resp.
DpG ) is the diastasis function with respect to p for the Kähler metric g (resp. G).
We claim that F + F̄ = 0. In order to prove our claim notice that, by the very definition of
∂ 2 Dg
Bochner coordinates, one can easily check that log det( ∂zα ∂ z̄pβ ) is of diastasis type. Moreover,
formula (2) yields
λ ∂ 2 Dpg
F + F̄ = DpG + log det( )
2 ∂zα ∂ z̄β
where the right hand side is of diastasis type because DpG is of diastasis type by Remark 2. This
forces F + F̄ to vanish, proving our claim.
By assumption (B) of Definition 1 there exist f1 , . . . , fn holomorphic functions on M \ Hpg ex-
tending the Bochner coordinates {z1 , . . . , zn } for the metric g. On the other hand, since assumption
n λ G
(A) of Definition 1 is satisfied both for g and G, the real analytic n-forms ωn! and e− 2 Dp dµ(f ) are
globally defined on the connected open set
X := (M \ Hpg ) ∩ (M \ HpG ) = M \ (Hpg ∪ HpG ).
Notice that X is connected by assumptions (iii) and (iv). Since, by formula (2) with F + F̄ = 0,
these n-forms agree on U, they must agree on X, i.e.
ωn λ G
(8) = e− 2 Dp dµ(f ).
n!
Now assume by contradiction that λ ≤ 0 . Combining (8) and the fact that DpG is non-negative
on X (by assumption (A) for the metric G) yields
ωn ωn
Z Z Z Z
−λ DpG
vol(M, g) := = = e 2 dµ(f ) ≥ dµ(f ).
M n! X n! X X
Moreover, since g has infinite Bochner-Euclidean volume (assumption (ii)), we get
Z Z
vol(M, g) ≥ dµ(f ) = dµ(f ) = voleucl (M \ Hpg , g) = ∞,
X M \Hpg

which is the desired contradiction, being vol(M, g) finite, by assumption (i). 


Remark 8. Notice that in the proof of Theorem 1 we do not use the non-negativity of Dpg of
condition (A) for the metric g, but only the fact that Dpg is globally defined on M \Hpg . Moreover, in
12
the proof we are only requiring DpG to satisfy condition (A) and not necessarily (B) of Definition 1.
Therefore we get:
Corollary 4. Let g be a projectively induced metric on a compact Kähler manifold M with
c1 (M) < 0. If −ρω is associated to a Kähler metric G, then either G does not satisfy (A) of
Definition 1 or codimR (HpG ) ≤ 1.
As a first example we specialize to Kähler-Ricci iterations on simply connected compact ho-
mogeneous Kähler manifolds, that is, generalized flag manifolds. This choice is based on the fact
that homogeneous Kähler manifolds are a precious source of examples of Kähler manifolds with
positive first Chern class in the compact realm. In particular, to our best knowledge, the only ex-
amples of projectively induced Kähler forms ω, Ω satisfying equation (1) and the hypotheses of
Theorem 1 are homogeneous Kähler forms. In fact, we can say something more for generalized
flag manifolds.
Proposition 2. Let M = G/K be a generalized flag manifold endowed with a Kähler metric
g and associated Kähler form ω. Then ω is homogeneous if and only if its Ricci form ρω is a
Kähler-Einstein form.
Proof. Let ω be homogeneous, i.e., for any h ∈ G we have h∗ ω = ω. Then we have h∗ ρω =
ρh∗ ω = ρω so that Ω := ρω is a homogeneous form. Moreover, Ω is positive at a point because
it represents a positive class. This, together with homogeneity, shows that Ω is a Kähler form.
Thus Ω Kähler-Einstein because it is has constant scalar curvature and satisfies [ρΩ ] = [Ω], see for
instance [21, Section 2].
Vice versa, suppose that ω is any Kähler form on M such that ρω is homogeneous. Then we have
ρh∗ ω = h∗ ρω = ρω . Since G is connected, h∗ = Id : H ∗ (M) −→ H ∗ (M) because h is homotopic
to the identity in G as a map M −→ M. This in particular implies [h∗ ω] = h∗ [ω] = [ω] and we
can conclude that ω = h∗ ω by Calabi’s conjecture. 
Observe that an integral homogeneous Kähler metric on a generalized flag manifold is always
projectively induced, see for instance [30, 35]. It seems natural to ask whether, relaxing the homo-
geneity condition on the Kähler metric, the statement of Proposition 2 still holds true. Namely, we
ask the following question:
Question: Let g and G be projectively induced Kähler metrics on a generalized flag manifold. Is
ρω = λΩ equivalent to G being Kähler–Einstein?
As a special case, Proposition 3 below gives an answer in the affirmative for (radial) metrics
on CP1 . In order to prove this result, we carry out in full detail the Kähler-Ricci iterations on
projectively induced metrics on CP1 . Namely, suppose that a Kähler metric g is induced by a
holomorphic embedding ϕ : CP1 −→ CPn in a higher dimensional complex projective space, that
is, g = ϕ∗ gF S where gF S is the Fubini-Study metric of CPn .
For homogeneous coordinates [z0 : z1 ] consider the affine chart {z0 6= 0} with coordinate
z = zz10 . Recall that the embedding ϕ is determined by sections of a bundle O(k) which can be
identified with homogeneous polynomials of degree k in z0 and z1 . Without loss of generality
we can assume that ϕ is full, i.e., the embedding is determined by a basis of H 0 (CP1 , O(n)).
Moreover, let us assume that the induced metric ϕ∗ (gF S ) is radial in z. This is equivalent to the
fact that the coordinates of ϕ are monomials so that in homogeneous coordinates (possibly after
reordering) ϕ is of the form
ϕ([z0 : z1 ]) = [α0 z0n : α1 z1 z0n−1 : · · · : αn z1n ].
13
Notice that we can impose αh0 = αn = 1 after i precomposing ϕ with f ∈ P GL(2, C) given in
1 1 ∗
coordinates by f ([z0 : z1 ]) = √n α z0 : n α z1 . Then the Kähler form ω = ϕ (ωF S ) associated to
0

n
the projectively induced metric g is given in the chosen chart by
i
(9) ω = ∂∂ log(1 + a1 |z|2 + · · · + aj |z|2j + · · · + |z|2n )
2
where aj = |αj | . Denote the argument of the logarithm in (9) by P (x) where x = |z|2 . Then the
2

form can be rewritten as


i P ′P + xP ′′ P − x(P ′ )2
(10) ω= dz ∧ dz̄.
2 P2
Writing D := P ′ P + xP ′′ P − x(P ′ )2 for conciseness, the Ricci form ρω is given by
D P2
(11) ρω = −i∂∂ log = i∂∂ log .
2P 2 D
First we want to prove the following result.
Proposition 3. Consider a projectively induced metric g on CP1 with associated Kähler form ω
n

as in (9). Then its Ricci form ρω is projectively induced if and only if aj = j , i.e., g = ngF S .
In order to prove Proposition 3 we need the following auxiliary result.
f (x)
Lemma 5. Let ω = 2i ∂∂ log be a Kähler form on CP1 where z is the affine coordinate on the
h(x)
chart U0 = {z0 6= 0}, x = |z|2 and f, h ∈ R[x] are polynomials. If g is projectively induced, then
f /h = Q ∈ R[x].
Proof. Suppose there exists an embedding ϕ : CP1 −→ CPn such that ω = ϕ∗ ωF S . Since ω is a
radial form on U0 , its potential must be radial on U0 . This means that we can write
i i f (x)
ω = ∂∂ log Q(x) = ∂∂ log
2 2 h(x)
|ϕj (z)|2 . Thus, log(f /h) differs from log Q by
P
where ϕ(z) = [ϕ0 (z) : · · · : ϕn (z)] and Q =
the real part of a holomorphic function F : U0 −→ C, that is, Q = f /h + eF +F̄ . Now it is clear
from the series expansion of Q that F + F̄ must be constant and this proves the claim. 
n
Proof of Proposition 3. Clearly, if P (x) = (1 + x)n = (1 + |z|2 ) then ρω = ρ(nωF S ) = 4ωF S
which is again projectively induced.
P2
Vice versa, suppose ρω = i∂∂ log is a projectively induced Kähler form. Then by Lemma 5
D
we can assume that
P2
(12) = λQ
D
for a monic polynomial Q and λ ∈ R. Notice that 2πi ρω represents c1 (CP1 ). Therefore deg Q = 2
for topological reasons.
We want to show that P = (1 + x)n by contradiction. If P 6= (1 + x)n , then P must have more
than one root. Consequently there exists a root of P , say α, such that Q 6= (x − α)2 . To conclude
the proof we show that this implies P = (x − α)n which is the desired contradiction.
14
We do so by proving that if (x − α)k divides P for k < n, then also (x − α)k+1 divides P . First,
it is evident from (12) that if (x − α)k divides P , then (x − α)2k−1 divides D because α is at most a
simple root of Q. Now it is easy to see that the (2k − 2)-th derivative D (2k−2) of D can be written
as
2
(13) D (2k−2) = −x P (k) + R
where P (k) is the k-th derivative of P and R is a sum of polynomials each of which is divided by
a derivative of P of order < k. Therefore we get
2 2
(14) 0 = D (2k−2) (α) = −α P (k) (α) + R(α) = −α P (k) (α)
Since P = (1 + · · · ) we have α 6= 0 and P k (α) = 0, that is, (x − α)k+1 divides P . This concludes
the proof. 
We want to specialize now to the case of embeddings ϕ : CP1 −→ CP2 in order to give some
explicit examples.
Example 12. By the discussion above, a radial embedding ϕ : CP1 −→ CP2 can be written in
homogeneous coordinates in the form
ϕ([z0 : z1 ]) = [z02 : αz1 z0 : z12 ].
The Kähler form ω associated to the projectively induced metric g = ϕ∗ (gF S ) metric on CP1 in an
affine chart is given as
i i (a + 4x + ax2 )
(15) ω = ∂∂ log(1 + ax + x2 ) = dz ∧ dz̄
2 2 (1 + ax + x2 )2
where a = |α|2 and x = |z|2 . Therefore its Ricci form is
(a + 4x + ax2 ) 2(a + 4x + ax2 )3 − 4a(1 + ax + x2 )3
(16) ω1 = ρω = −i∂∂ log =i dz ∧ dz̄.
2(1 + ax + x2 )2 (a + 4x + ax2 )2 (1 + ax + x2 )2
By Proposition 3 the metric g1 associated to the Ricci form ω1 := ρω is projectively induced
if and only if a = 2. Moreover, one can easily check that √ the coefficients of the polynomial
2(a + 4x + ax2 )3 − 4a(1 + ax + x2 )3 are all positive for 2 ≤ a ≤ 2 so that for those values
g1 is indeed a Kähler metric. In addition, the expression (16) of the Kähler form ω1 is written in
Bochner coordinates and can be used to compute the Euclidean volume voleucl (CP1 \ [0 : 1], g1 ):

2(a + 4x + ax2 )3 − 4a(1 + ax + x2 )3
Z
1
voleucl (CP \ [0 : 1], g1 ) = 2π dx
0 (a + 4x + ax2 )2 (1 + ax + x2 )2
where we made use of the equality dx = dr 2 = 2rdr for z = re−iθ . Looking at the degrees of the
numerator and denominator of the integrand it is immediate that voleucl (CP1 \[0 : 1], g1) = ∞. We
conclude that g is a projectively induced metric, hence well-behaved at all points and with infinite
euclidean volume, whose Ricci form is associated √ to a well-behaved, not projectively induced
Kähler metric g1 with infinite Euclidean volume for 2 ≤ a < 2.
Based on this example we want to show that the second Ricci iteration of g is again a Kähler form
ω2 associated to a well-behaved Kähler metric g2 with infinite Euclidean volume when a belongs
to an suitable interval. To show this we iterate the computations carried above. It is evident that
the complexity of such calculations increases with the number of iterations. For this reason we
15
resorted to the use of the software Mathematica [19] to compute the Ricci form ω2 := ρω1 = ρ2ω as
well as the values of a for which ω2 is associated to a Kähler metric. The calculations yield
2(a + 4x + ax2 )3 − 4a(1 + ax + x2 )3 6N
ω2 = ρ2ω = −i∂∂ log 2 2 2 2
=i dz ∧ dz̄
(a + 4x + ax ) (1 + ax + x ) D
where N and D are polynomials in x with D > 0. One√ can check numerically that the coefficients
of N are all positive when a lies in an interval I ⊂ ( 2, 2). Therefore, for a ∈ I, both g1 and
g2 are positive definite. Moreover, from deg N = 18 and deg D = 20 follows that the Euclidean
volume of g2 is
Z ∞
1 6N
voleucl (CP \ [0 : 1], g2 ) = 2π dx = ∞.
0 D
Therefore, for a ∈ I, g1 is an example of well-behaved, not projectively induced Kähler metric
with infinite Euclidean volume whose Ricci form is associated to a well-behaved, not projectively
induced Kähler metric with infinite Euclidean volume. This provides a pair of well-behaved (not
projectively induced) metrics on CP1 satisfying the hypotheses of Theorem 1.

4. P ROOFS OF T HEOREM 2 AND T HEOREM 3


In order to prove Theorem 2 we briefly recall some basic facts on Umehara algebra and its field
of fractions. Let M be a complex manifold. Fix a point p ∈ M and let Op be the algebra of germs
of holomorphic functions around p. Denote by Rp the germs of real numbers. The Umehara algebra
(see [25, 41] and references therein) is defined to be the R-algebra Λp generated by elements of the
form hk̄ + h̄k, for h, k ∈ Op . Let

Ôp = {α = (α1 , . . . , αm ) | αj ∈ Op , αj (p) = 0, ∀j = 1, . . . , m, m ≥ 1} .


For α = (α1 , . . . , αm ) ∈ Ôp and ℓ ∈ N such that ℓ ≤ m we set

X m
X
hα, αiℓ(z) = |αj (z)|2 − |αk (z)|2 .
j=1 k=ℓ+1

Since hk̄ + h̄k = |h + k|2 − |h|2 − |k|2 , it is not hard to see that each f ∈ Λp can be written as
f = h + h̄ + hα, αiℓ
for some h ∈ Op , α = (α1 , . . . , αm ) ∈ Ôp , ℓ ≤ m and such that α1 , . . . , αm are linearly indepen-
dent over C.
Consider the R-algebra Λ̃p ⊂ Λp given by
n o
(17) Λ̃p = a + hα, αiℓ | a ∈ Rp , α ∈ Ôp , ℓ ≤ m .

Notice that the germ of the real part of a non-constant holomorphic function h ∈ Op belongs to Λp
but not to Λ̃p .
The key elements in the proof of Theorem 2 and Theorem 3 are the following two lemmata. The
first lemma descents from [26, Theorem 2.1] and we specialize it here to our needs.
Lemma 6. Let K̃p be the field of fractions of Λ̃p . Let ξ ∈ K̃p , then eξ 6∈ K̃pα K̃p \ Rp , ∀α ∈ R.
16
Lemma 7. Let M be an n-dimensional complex manifold and p ∈ M and let fh ∈ Kp , where
Kp denotes the field of fractions of the Umehara algebra Λp . Then, for any system of complex
coordinates {z1 , . . . , zn } around p one has:
" #
2 f
∂ log
f n+1 hn+1 det h
∈ Λp , with α, β = 1, . . . , n.
∂zα ∂ z̄β

Proof. Set fα = ∂f /∂zα , hα = ∂h/∂zα , fβ̄ = ∂f /∂ z̄β , hβ̄ = ∂h/∂ z̄β and fαβ̄ = ∂ 2 f /∂zα ∂ z̄β ,
hαβ̄ = ∂ 2 h/∂zα ∂ z̄β , α, β = 1, · · · , n. Then
∂ 2 log(f /h)
= fαβ̄ /f − hαβ̄ /h − fα fβ̄ /f 2 + hα fβ̄ /h2 ,
∂zα ∂ z̄β
so that
∂ 2 log(f /h)  
(18) fh = h fαβ̄ − fα fβ̄ /f − f hαβ̄ − hα hβ̄ /h .
∂zα ∂ z̄β
h 2 fi
n+1 n+1 ∂ log
Now some linear algebra shows that f h det ∂zα ∂ z̄βh is the determinant of a matrix whose
entries are generated by holomorphic and anti-holomorphic functions. Namely, we have
2 log(f /h) ∂ 2 log(f /h)
f h ∂ ∂z
 
1 ∂ z̄ 1
· · · f h ∂z 1 ∂ z̄ n
0 0
 .. .. .. .. .. 
∂ 2
log(f /h)  . . . . . 
f n+1 hn+1 det = f h det  f h ∂ 2 log(f /h) · · · f h ∂ 2 log(f /h) 0 0  .
 
∂zα ∂ z̄β  ∂zn ∂ z̄1 ∂zn ∂ z̄n 
 hf /f 1̄ ··· hf /f
n̄ 1 0 
f h1̄ /h ··· f hn̄ /h 0 1
Denote by Rj the j-th row of the matrix above. Replacing Rj by Rj + fj Rn+1 − hj Rn+2 for all
j = 1, . . . , n and making use of (18) we get
f11 h − f h11 · · · f1n̄ h − f h1n f1 −h1
 
.. .. .. .. .. 
∂ 2 log(f /h)
 . . . . . 
n+1 n+1

f h det = f h det  f h − f h · · · fnn̄ h − f hnn f n −h .
n 
∂zα ∂ z̄β  n1 n1
 hf1 /f ··· hfn̄ /f 1 0 
f h1 /h ··· f hn̄ /h 0 1
Finally, multiplying the last two rows by f and h respectively we get
f11 h − f h11 · · · f1n̄ h − f h1n f1 −h1
 
.. .. .. .. .. 
∂ 2 log(f /h)
 . . . . . 
n+1 n+1

f h det = det  f h − f h · · · fnn̄ h − f hnn f n −h .
n 
∂zα ∂ z̄β  n1 n1
 hf1 ··· hfn̄ f 0 
f h1 ··· f hn̄ 0 h
2
Hence f n+1 hn+1 det ∂ ∂zlog(f /h)
α ∂ z̄ β is finitely generated by holomorphic or anti-holomorphic functions.
2
In addition, it is real-valued, because the matrix ∂ ∂zlog(f /h)
α ∂ z̄ β is Hermitian and this proves the lemma.

Remark 9. Lemma 7 is an extension of [40, Lemma 2.2], which is valid for Λp , to its field of
fractions Kp .
17
The following result, needed both in the proofs of Theorem 2 and Theorem 3, follows by induc-
tion on k ≥ 1, by the definition of the Ricci form associated to a real analytic Kähler metric and
by Lemma 7.
Corollary 5. Let g be a real analytic Kähler metric on a complex manifold M. Choose U around
¯ k,
p ∈ M such that Calabi’s diastasis function Dpg : U → R is defined and such that ρkω = 2i ∂ ∂Ψ
where Ψk is of diastasis type (cf. Remark 3). If Dpg belongs either to K̃p or log K̃p then Ψk ∈
log K̃p .
Proof of Theorem 2. Since g is induced by a flat metric there exists an open neighbourhood U of a
point p ∈ M and a holomorphic isometry ϕ : U → CN such that ϕ∗ g0 = g|U , where g0 is the flat
metric on CN . Let
ϕ|U : U → CN , q 7→ (ϕ1 (q), . . . , ϕN (q))
where ϕj ∈ Op and ϕj (p) = 0, j = 1, . . . , N, be the local expression of ϕ.
Notice that, by the hereditary property of the diastasis function we have
N
X
(19) Dpg = |ϕi |2 ∈ Λ̃p .
i=1

By shrinking U if necessary, Corollary 5 yields


i ¯
(20) ρkω = ∂ ∂Ψ k
2
where ω is the Kähler form associated to g and Ψk ∈ log K̃p . Now equation ρkω = λΩ, together
with the fact that G is induced by the flat metric gives
M
X
Ψk = λDpG =λ |φi |2 ∈ Λ̃p ,
i=1

where φk ∈ Op and φk (p) = 0, such that φk is a non-constant function for all k = 1, . . . , M.


Thus Lemma 6 forces λ = 0. Then the Ricci tensor of g vanishes. Since the immersion of M in
its flat ambient space lifts to a globally defined holomorphic isometry ϕ : M → CN , the thesis
follows. 
Remark 10. In contrast with the proof of Theorem 3, the previous proof only uses a weak version
of Lemma 6, that is, the fact that ef 6∈ K̃p \ Rp , ∀f ∈ Λ̃p . This has been proven in [12, (i) of
Theorem 2.1)].
Proof of Theorem 3. Fix a point p ∈ M and consider an open coordinate neighbourhood centred
at p.. Since the solitonic vector field X for G can be assumed to be the real part of a holomorphic
vector field, we can write locally
n  
X ∂ ¯ ∂
X= fj + fj
j=1
∂zj ∂ z̄j
for some holomorphic functions fj , j = 1, . . . , n, on U. Thus, by the definition of Lie derivative,
after a straightforward computation, we can write on U
i ¯
(21) LX Ω = ∂ ∂ξ
2
18
where
n
∂DpG ∂DpG
¯
X
(22) ξ= fj + fj .
j=1
∂zj ∂ z̄j

The equation ρkω = λΩ together with the KRS equation for Ω, i.e. ρΩ = µΩ + LX Ω, gives
(23) ρk+1
ω = µρkω + LX Ω,
¯ k and ρk+1 = i ∂ ∂Ψ
By Remark 3 we can restrict U so that ρkω = 2i ∂ ∂Ψ ¯ k+1 with Ψk and Ψk+1 of
ω 2
G
diastasis type. Thus (21), (22) and (23) together with λDp = Ψk yield
ξ
(24) Ψk+1 = λµΨk + .
λ
Notice now that the assumption that g is induced by a complex space form implies that Dpg ∈ K̃p
(in the flat case) or Dpg ∈ log K̃p (for the hyperbolic or projective space). Thus, by Corollary 5 Ψk
ξ
and Ψk+1 belong to log K̃p and, consequently, = nj=1 fj ∂Ψ + f¯j ∂Ψ
P
∂zj
k k
∂ z̄j
∈ K̃p . It follows by (24)
λ
that eξ ∈ K̃p−λµ K̃p . Hence ξ is a constant by Lemma 6 forcing G to be Kähler–Einstein. 
Remark 11. The proofs of Theorems 2 and 3 show that the same conclusions can be achieved
taking any metric g such that Dpg ∈ log K̃p . Examples of such metrics are given by the Kähler
metrics induced by those in Examples 6-10 above.

R EFERENCES
1. C. Arezzo and A. Loi, A note on Kähler-Einstein metrics and Bochner’s coordinates, Abh. Math. Sem. Univ.
Hamburg 74 (2004), 49–55. MR 2112820
2. Claudio Arezzo, Andrea Loi, and Fabio Zuddas, Some remarks on the symplectic and Kähler geometry of toric
varieties, Ann. Mat. Pura Appl. (4) 195 (2016), no. 4, 1287–1304. MR 3522347
3. Thierry Aubin, Équations du type Monge-Ampère sur les variétés kählériennes compactes, Bull. Sci. Math. (2)
102 (1978), no. 1, 63–95. MR 494932
4. Victor V. Batyrev, Quantum cohomology rings of toric manifolds, no. 218, 1993, Journées de Géométrie
Algébrique d’Orsay (Orsay, 1992), pp. 9–34. MR 1265307
5. Robert J. Berman, Sebastien Boucksom, Philippe Eyssidieux, Vincent Guedj, and Ahmed Zeriahi, Kähler-Einstein
metrics and the Kähler-Ricci flow on log Fano varieties, J. Reine Angew. Math. 751 (2019), 27–89. MR 3956691
6. S. Bochner, Curvature in Hermitian metric, Bull. Amer. Math. Soc. 53 (1947), 179–195. MR 19983
7. Armand Borel, Linear algebraic groups, second ed., Graduate Texts in Mathematics, vol. 126, Springer-Verlag,
New York, 1991. MR 1102012
8. Eugenio Calabi, Isometric imbedding of complex manifolds, Ann. of Math. (2) 58 (1953), 1–23. MR 57000
9. Xiuxiong Chen, Simon Donaldson, and Song Sun, Kähler-Einstein metrics on Fano manifolds. I: Approximation
of metrics with cone singularities, J. Amer. Math. Soc. 28 (2015), no. 1, 183–197. MR 3264766
10. , Kähler-Einstein metrics on Fano manifolds. II: Limits with cone angle less than 2π, J. Amer. Math. Soc.
28 (2015), no. 1, 199–234. MR 3264767
11. , Kähler-Einstein metrics on Fano manifolds. III: Limits as cone angle approaches 2π and completion of
the main proof, J. Amer. Math. Soc. 28 (2015), no. 1, 235–278. MR 3264768
12. Xiaoliang Cheng, Antonio J. Di Scala, and Yuan Yuan, Kähler submanifolds and the Umehara algebra, Internat.
J. Math. 28 (2017), no. 4, 1750027, 13. MR 3639045
13. Tamás Darvas and Yanir A. Rubinstein, Convergence of the Kähler-Ricci iteration, Anal. PDE 12 (2019), no. 3,
721–735. MR 3864208
14. Antonio J. Di Scala and Andrea Loi, Kähler maps of Hermitian symmetric spaces into complex space forms,
Geom. Dedicata 125 (2007), 103–113. MR 2322543
19
15. Simon K. Donaldson, Conjectures in Kähler geometry, Strings and geometry, Clay Math. Proc., vol. 3, Amer.
Math. Soc., Providence, RI, 2004, pp. 71–78. MR 2103718
16. A. Futaki, An obstruction to the existence of Einstein Kähler metrics, Invent. Math. 73 (1983), no. 3, 437–443.
MR 718940
17. Richard S. Hamilton, The Ricci flow on surfaces, Mathematics and general relativity (Santa Cruz, CA, 1986),
Contemp. Math., vol. 71, Amer. Math. Soc., Providence, RI, 1988, pp. 237–262. MR 954419
18. Dominique Hulin, Kähler-Einstein metrics and projective embeddings, J. Geom. Anal. 10 (2000), no. 3, 525–528.
MR 1794575
19. Wolfram Research, Inc., Mathematica, Version 13.3, Champaign, IL, 2023.
20. Julien Keller, Ricci iterations on Kähler classes, J. Inst. Math. Jussieu 8 (2009), no. 4, 743–768. MR 2540879
21. Shoshichi Kobayashi, Hypersurfaces of complex projective space with constant scalar curvature, J. Differential
Geometry 1 (1967), 369–370. MR 229183
22. A. Loi, G. Placini, and M. Zedda, Immersions into Sasakian space forms, 9 May 2023, preprint arXiv:2305.05509
[math.DG].
23. Andrea Loi, Calabi’s diastasis function for Hermitian symmetric spaces, Differential Geom. Appl. 24 (2006),
no. 3, 311–319. MR 2216943
24. Andrea Loi and Roberto Mossa, The diastatic exponential of a symmetric space, Math. Z. 268 (2011), no. 3-4,
1057–1068. MR 2818743
25. , Kähler immersions of Kähler-Ricci solitons into definite or indefinite complex space forms, Proc. Amer.
Math. Soc. 149 (2021), no. 11, 4931–4941. MR 4310116
26. , Holomorphic isometries into homogeneous bounded domains, Proc. Amer. Math. Soc. 151 (2023), no. 9,
3975–3984. MR 4607641
27. , Rigidity properties of holomorphic isometries into homogeneous Kähler manifolds, 9 May 2023, preprint
arXiv:2305.05244 [math.DG].
28. Andrea Loi, Roberto Mossa, and Fabio Zuddas, Bochner coordinates on flag manifolds, Bull. Braz. Math. Soc.
(N.S.) 50 (2019), no. 2, 497–514. MR 3955252
29. Andrea Loi and Michela Zedda, Kähler immersions of Kähler manifolds into complex space forms, Lecture
Notes of the Unione Matematica Italiana, vol. 23, Springer, Cham; Unione Matematica Italiana, [Bologna], 2018.
MR 3838438
30. Andrea Loi and Fabio Zuddas, On the Gromov width of homogeneous Kähler manifolds, Differential Geom. Appl.
47 (2016), 130–132. MR 3504923
31. Alan M. Nadel, On the absence of periodic points for the Ricci curvature operator acting on the space of Kähler
metrics, Modern methods in complex analysis (Princeton, NJ, 1992), Ann. of Math. Stud., vol. 137, Princeton
Univ. Press, Princeton, NJ, 1995, pp. 277–281. MR 1369142
32. Thomas Peternell and Michael Schneider, Compactifications of Cn : a survey, Several complex variables and com-
plex geometry, Part 2 (Santa Cruz, CA, 1989), Proc. Sympos. Pure Math., vol. 52, Amer. Math. Soc., Providence,
RI, 1991, pp. 455–466. MR 1128563
33. Yanir A. Rubinstein, Some discretizations of geometric evolution equations and the Ricci iteration on the space
of Kähler metrics, Adv. Math. 218 (2008), no. 5, 1526–1565. MR 2419932
34. Osamu Suzuki, Remarks on continuation problems of Calabi’s diastatic functions, J. Fac. Sci. Univ. Tokyo Sect.
IA Math. 29 (1982), no. 1, 45–49. MR 657871
35. Masaru Takeuchi, Homogeneous Kähler submanifolds in complex projective spaces, Japan. J. Math. (N.S.) 4
(1978), no. 1, 171–219. MR 528871
36. Hiroyuki Tasaki, The cut locus and the diastasis of a Hermitian symmetric space of compact type, Osaka J. Math.
22 (1985), no. 4, 863–870. MR 815454
37. Gang Tian, On a set of polarized Kähler metrics on algebraic manifolds, J. Differential Geom. 32 (1990), no. 1,
99–130. MR 1064867
38. , Kähler-Einstein metrics with positive scalar curvature, Invent. Math. 130 (1997), no. 1, 1–37.
MR 1471884
39. , K-stability and Kähler-Einstein metrics, Comm. Pure Appl. Math. 68 (2015), no. 7, 1085–1156.
MR 3352459
40. Masaaki Umehara, Einstein Kaehler submanifolds of a complex linear or hyperbolic space, Tohoku Math. J. (2)
39 (1987), no. 3, 385–389. MR 902577
20
41. , Diastases and real analytic functions on complex manifolds, J. Math. Soc. Japan 40 (1988), no. 3, 519–
539. MR 945350
42. Shing Tung Yau, On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation.
I, Comm. Pure Appl. Math. 31 (1978), no. 3, 339–411. MR 480350

D IPARTIMENTO DI M ATEMATICA E I NFORMATICA , U NIVERSIT Á DEGLI STUDI DI C AGLIARI , V IA O SPEDALE


72, 09124 C AGLIARI , I TALY
Email address: loi@unica.it

D IPARTIMENTO DI M ATEMATICA E I NFORMATICA , U NIVERSIT Á DEGLI STUDI DI C AGLIARI , V IA O SPEDALE


72, 09124 C AGLIARI , I TALY
Email address: giovanni.placini@unica.it

21

You might also like