geometry of warped products as riemannian submanifolds and related
TRANSCRIPT
SOOCHOW JOURNAL OF MATHEMATICS
Volume 28, No. 2, pp. 125-156, April 2002
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN
SUBMANIFOLDS AND RELATED PROBLEMS
BY
BANG-YEN CHEN
Abstract. The warped product N1 ×f N2 of two Riemannian manifolds (N1, g1)
and (N2, g2) with warping function f is the product manifold N1×N2 equipped with
the warped product metric g1+f2g2, where f is a positive function on N1. It is well-
known that the notion of warped products plays some important roles in differential
geometry as well as in physics. In this article we survey recent results on warped
products isometrically immersed in real or complex space forms as Riemannian,
Lagrangian, or CR-submanifolds. Moreover, we also present some recent related
results concerning convolution of Riemannian manifolds, constant-ratio submani-
folds, T -constant submanifolds, N -constant submanifolds and rectifying curves.
1. Introduction
Let B and F be two Riemannian manifolds of positive dimensions equipped
with Riemannian metrics gB and gF , respectively, and let f be a positive function
on B. Consider the product manifold B × F with its projection π : B × F → B
and η : B × F → F . The warped product M = B ×f F is the manifold B × F
equipped with the Riemannian structure such that
||X||2 = ||π∗(X)||2 + f2(π(x))||η∗(X)||2, (1.1)
Communicated November 10 2001.AMS Subject Classification. 53C40, 53C42, 53C50.Key words. warped product, warped product immersion, CR-product, Segre embedding,
CR-warped product, constant-ratio submanifold, T -constant submanifold, N -constant subman-ifold, rectifying curve, convolution manifold, convolution Riemannian manifold.
This is the written version of author’s talk delivered at the International Conference onDifferential Geometry held at Tokyo Metropolitan University, December 17-19, 2001, celebratingthe 60th anniversary of Professor Koichi Ogiue.
125
126 BANG-YEN CHEN
for any tangent vector X ∈ TxM . Thus, we have g = gB + f2gF . The function f
is called the warping function of the warped product (cf. [4]). It is well-known
that the notion of warped products plays some important roles in differential
geometry as well as in physics ([2, 49]). For instance, the best relativistic model
of the Schwarzschild space-time that describes the out space around a massive
star or a black hole is given as a warped product (cf. [49], pp.364-367).
One of the most fundamental problems in the theory of submanifolds is the
immersibility (or non-immersibility) of a Riemannian manifold in a Euclidean
space (or, more generally, in a space form). According to a well-known theorem
of J. F. Nash, every Riemannian manifold can be isometrically immersed in some
Euclidean spaces with sufficiently high codimension. Nash’s theorem implies in
particular that every warped product N1×f N2 can be immersed as a Riemannian
submanifold in some Euclidean space. Moreover, many important submanifolds
in real and complex space forms are expressed as warped products submanifolds,
see, for instance, ([1, 4, 5, 6, 12, 14, 36, 37, 38, 39, 49, 50]).
In view of these simple facts, we propose the following very simple problem:
Problem 1.
∀ N1 ×f N2
isometric
−−−−−−−−→immersion
Em or Rm(c) =⇒ ??? (1.2)
where Rm(c) denotes a Riemannian m-manifold of constant sectional curvature
c.
In particular, we ask the following.
Problem 1.1. Given a warped product N1 ×f N2, what are the necessary
conditions for the warped product to admit a minimal isometric immersion in a
Euclidean m-space Em (or in Rm(c))?
Problem 1.2. Let N1 ×f N2 be an arbitrary warped product immersed in
a Euclidean space Em as a Riemannian submanifold. What are the relationships
between the warped product structure of N1 ×f N2 and extrinsic structures of
N1 ×f N2 in Em (or in Rm(c)) as a Riemannian submanifold ?
Notice that the main warped structure of a warped product is warping func-
tion.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 127
In this article, we present recent results and related results concerning these
basic problems. More precisely, we survey recent results concerning warped prod-
ucts isometrically immersed in real or complex space forms as Riemannian, La-
grangian, or CR-submanifolds. Moreover, we also present some related results
on convolution of Riemannian manifolds, constant-ratio submanifolds, T -constant
submanifolds, N -constant submanifolds, as well as rectifying curves.
2. Basic Formulas
Let N be an n-dimensional submanifold of a Riemannian m-manifold Rm(c)
of constant sectional curvature c. We choose a local field of orthonormal frame
e1, . . . , en, en+1, . . . , em in Rm(c) such that, restricted to N , the vectors e1, . . . , en
are tangent to N and en+1, . . . , em are normal to N .
Let K(ei ∧ ej), 1 ≤ i < j ≤ n, denote the sectional curvature of the plane
section spanned by ei, ej . Then the scalar curvature of N is given by
τ =∑
i<j
K(ei ∧ ej). (2.1)
For a submanifold N in Rm(c) we denote by ∇ and ∇ the Levi-Civita con-
nections of N and Rm(c), respectively. The Gauss and Weingarten formulas are
given respectively by
∇XY = ∇XY + σ(X,Y ), (2.2)
∇Xξ = −AξX +DXξ, (2.3)
for vector fieldsX,Y tangent toN and vector field ξ normal toN , where σ denotes
the second fundamental form, D the normal connection, and A the shape operator
of the submanifold. Let {σrij}, i, j = 1, . . . , n; r = n+ 1, . . . ,m, denote the coef-
ficients of the second fundamental form with respect to e1, . . . , en, en+1, . . . , em.
The mean curvature vector−→H is defined by
−→H =
1
ntrace σ =
1
n
n∑
i=1
σ(ei, ei), (2.4)
where {e1, . . . , en} is a local orthonormal frame of the tangent bundle TN of
N . The squared mean curvature is given by H2 = 〈 −→H,−→H 〉, where 〈 , 〉 denotes
128 BANG-YEN CHEN
the inner product. A submanifold N is called minimal in Rm(c) if the mean
curvature vector of N in Rm(c) vanishes identically.
Denote by R the Riemann curvature tensors of N . Then the equation of
Gauss is given by (see, for instance, [7])
R(X,Y ;Z,W ) = c{〈X,W 〉 〈Y,Z〉 − 〈X,Z〉 〈Y,W 〉}+ 〈σ(X,W ), σ(Y,Z)〉 − 〈σ(X,Z), σ(Y,W )〉 , (2.5)
for vectors X,Y,Z,W tangent to N .
Let M be a Riemannian p-manifold and e1, . . . , ep be an orthonormal frame
fields on M . For differentiable function ϕ on M , the Laplacian ∆ϕ of ϕ is defined
by
∆ϕ =p
∑
j=1
{(∇ejej)ϕ− ejejϕ}. (2.6)
3. Riemannian Products, Moore’s Lemma and Nolker’s Theorem
Suppose that M1, . . . ,Mk are Riemannian manifolds and that
f : M1 × · · · ×Mk → EN
is an isometric immersion of the Riemannian productM1×· · ·×Mk into Euclidean
N -space. J. D. Moore [42] proved that if the second fundamental form h of f
has the property that h(X,Y ) = 0 for X tangent to Mi and Y tangent to Mj,
i 6= j, then f is a product immersion, that is, there exist isometric immersions
fi : Mi → Emi , 1 ≤ i ≤ k, such that
f(x1, . . . , xk) = (f(x1), . . . , f(xk)), (3.1)
when xi ∈Mi for 1 ≤ i ≤ k.
Let M0, · · · ,Mk be Riemannian manifolds, M = M0×· · ·×Mk their product,
and πi : M →Mi the canonical projection. If ρ1, · · · , ρk : M0 → R+ are positive-
valued functions, then
〈X,Y 〉 := 〈π0∗X,π0∗Y 〉 +k
∑
i=1
(ρi ◦ π0)2 〈πi∗X,πi∗Y 〉 (3.2)
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 129
defines a Riemannian metric on M , called a warped product metric. M endowed
with this metric is denoted by M0 ×ρ1M1 × · · · ×ρk
Mk.
A warped product immersion is defined as follows: LetM0×ρ1M1×· · ·×ρk
Mk
be a warped product and let fi : Ni →Mi, i = 0, · · · , k, be isometric immersions,
and define σi := ρi ◦ f0 : N0 → R+ for i = 1, · · · , k. Then the map
f : N0 ×σ1N1 × · · · ×σk
Nk →M0 ×ρ1M1 × · · · ×ρk
Mk (3.3)
given by f(x0, · · · , xk) := (f0(x0), f1(x1), · · · , fk(xk)) is an isometric immersion,
which is called a warped product immersion.
S. Nolker [47] extended Moore’s result to the following.
Nolker Theorem. Let f : N0×σ1N1 ×· · ·×σk
Nk → RN (c) be an isometric
immersion into a Riemannian manifold of constant curvature c. If h is the sec-
ond fundamental form of f and h(Xi, Xj) = 0, for all vector fields Xi and Xj,
tangent to Ni and Nj respectively, with i 6= j, then, locally, f is a warped product
immersion.
Warped product immersions was applied in [38] by Dillen and Nolker to clas-
sify semi-parallel immersions with flat normal connection. For a recent extensions
of Nolker theorem, see [42, 51].
4. Warped Products in Real Space Forms
In this section we present some recent general results concerning warped
products isometrically immersed in real space forms. These results provide some
solutions to Problem 1.
Theorem 4.1.([28]) For any isometric immersion φ : N1 ×f N2 → Rm(c) of
a warped product N1 ×f N2 into a Riemannian manifold of constant curvature c,
we have∆f
f≤ n2
4n2H2 + n1c, (4.1)
where ni = dimNi, n = n1 +n2, H2 is the squared mean curvature of φ, and ∆f
is the Laplacian f on N1.
130 BANG-YEN CHEN
The equality sign of (4.1) holds identically if and only if there exists a warped
product representation M1 ×ρ M2 of Rm(c) such that φ : N1 ×f N2 →M1 ×ρ M2
is a minimal warped product immersion.
Theorem 4.2.([27]) For any isometric immersion φ : N1 ×f N2 → Rm(c),
the scalar curvature τ of the warped product N1 ×f N2 satisfies
τ ≤ ∆f
n1f+n2(n− 2)
2(n− 1)H2 +
1
2(n+ 1)(n− 2)c. (4.2)
If n = 2, the equality case of (4.2) holds automatically.
If n ≥ 3, the equality sign of (4.2) holds identically if and only if either
(a) N1 ×f N2 is of constant curvature c, the warping function f is an eigen-
function with eigenvalue c, i.e., ∆f = cf , and N1 ×f N2 is immersed as a totally
geodesic submanifold in Rm(c), or
(b) locally, N1 ×f N2 is immersed as a rotational hypersurface in a totally
geodesic submanifold Rn+1(c) of Rm(c) with a geodesic of Rn+1(c) as its profile
curve.
Remark 4.1. Every Riemannian manifold of constant curvature c can be
locally expressed as a warped product whose warping function satisfies ∆f = cf .
For examples, Sn(1) is locally isometric to (0,∞) ×cos x Sn−1(1); E
n is locally
isometric to (0,∞) ×x Sn−1(1); and Hn(−1) is locally isometric to R ×ex E
n−1.
Remark 4.2. There are other warped product decompositions of Rn(c)
whose warping function satisfies ∆f = cf . For example, let {x1, . . . , xn1} be a
Euclidean coordinate system of En1 and let ρ =
∑n1
j=1 ajxj+b, where a1, . . . , an1, b
are real numbers satisfying∑n1
j=1 a2j = 1. Then the warped product E
n1 ×ρSn2(1)
is a flat space whose warping function is a harmonic function.
In fact, those are the only warped product decompositions of flat spaces whose
warping functions are harmonic functions.
Remark 4.3. Ejiri constructed in [39] many examples of minimal warped
product immersions into Riemannian manifolds of constant curvature. Moreover,
he also proved the following.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 131
Theorem 4.3.([40]) There exist countably many immersions of S1 × Sn−1
into Sn+1 such that the induced metric on S1 × Sn−1 is a warped product of
constant scalar curvature n(n− 1).
Now, we provide some immediate applications.
Corollary 4.1.([28]) Let N1 and N2 be two Riemannian manifold and f be
a nonzero harmonic function on N1. Then every minimal isometric immersion
of N1 ×f N2 into any Euclidean space is a warped product immersion.
Remark 4.4. There exist many such minimal immersions. For example, if
N2 is a minimal submanifold of Sm−1(1) ⊂ Em, then the minimal cone C(N2)
over N2 with vertex at the origin is a warped product R+ ×s N2 whose warping
function f = s is a harmonic function. Here s is the coordinate function of the
positive real line R+.
Corollary 4.2.([28]) Let f 6= 0 be a harmonic function on N1. Then for
any Riemannian manifold N2 the warped product N1 ×f N2 does not admits any
minimal isometric immersion into any hyperbolic space.
Corollary 4.3.([28]) Let f be a function on N1 with ∆f = λf with λ > 0.
Then for any Riemannian manifold N2 the warped product N1 ×f N2 admits no
minimal isometric immersion into any Euclidean space or hyperbolic space.
Remark 4.5. In views of Corollaries 4.2 and 4.3, we point out that there
exist minimal immersions from N1 ×f N2 into hyperbolic space such that the
warping function f is an eigenfunction with negative eigenvalue. For example,
R ×ex En−1 admits an isometric minimal immersion into the hyperbolic space
Hn+1(−1) of constant curvature −1.
Corollary 4.4.([28]) If N1 is a compact, then N1 ×f N2 does not admit a
minimal isometric immersion into any Euclidean space or hyperbolic space.
Remark 4.6. For Corollary 4.4, we point out that there exist many minimal
immersions of N1 ×f N2 into Em with compact N2. For examples, a hypercater-
noid in En+1 is a minimal hypersurfaces which is isometric to a warped product
R ×f Sn−1. Also, for any compact minimal submanifold N2 of Sm−1 ⊂ E
m, the
minimal cone C(N2) is a warped product R+ ×sN2 which is also a such example.
132 BANG-YEN CHEN
Remark 4.7. Contrast to Euclidean and hyperbolic spaces, Sm admits
minimal warped product submanifolds N1 ×f N2 with N1 and N2 being compact.
The simplest such examples are minimal Clifford tori defined by
Mk,n−k = Sk
√
k
n
× Sn−k
√
n− k
n
⊂ Sn+1, 1 ≤ k < n.
Remark 4.8. In [54] Suceava construct a family of warped products of
hyperbolic planes which do not admit any isometric minimal immersion into
Euclidean space, by applying the invariants introduced in [15].
5. Warped Product Real Hypersurfaces in Complex Space Forms
For Riemannian product real hypersurfaces in complex space forms, we have
the following non-existence theorem.
Theorem 5.1.([32]) There do not exist real hypersurfaces in complex pro-
jective and complex hyperbolic spaces which are Riemannian products of two or
more Riemannian manifolds of positive dimension.
In other words, every real hypersurface in a nonflat complex space form is
irreducible.
A contact manifold is an odd-dimensional manifold M 2n+1 with a 1-form η
such that η ∧ (dη)n 6= 0. A curve γ = γ(t) in a contact manifold is called a
Legendre curve if η(β ′(t)) = 0 along β. Let S2n+1(c) denote the hypersphere
in Cn+1 with curvature c centered at the origin. Then S2n+1(c) is a contact
manifold endowed with a canonical contact structure which is the dual 1-form
of the characteristic vector field Jξ, where J is the complex structure and ξ the
unit normal vector on S2n+1(c).
Legendre curves are known to play an important role in the study of contact
manifolds, e.g. a diffeomorphism of a contact manifold is a contact transformation
if and only if it maps Legendre curves to Legendre curves.
Contrast to Theorem 5.1, there exist many warped product real hypersurfaces
in complex space forms as given in the following three theorems.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 133
Theorem 5.2.([24]) Let a be a positive number and γ(t) = (Γ1(t),Γ2(t)) be
a unit speed Legendre curve γ : I → S3(a2) ⊂ C2 defined on an open interval I.
Then
x(z1, . . . , zn, t) = (aΓ1(t)z1, aΓ2(t)z1, z2, . . . , zn), z1 6= 0, (5.1)
defines a real hypersurface which is the warped product Cn∗∗ ×a|z1| I of a complex
n-plane and I, where Cn∗∗ = {(z1, . . . , zn) : z1 6= 0}.
Conversely, up to rigid motions of Cn+1, every real hypersurface in C
n+1
which is the warped product N ×f I of a complex hypersurface N and an open
interval I is either obtained in the way described above or given by the product
submanifold Cn × C ⊂ C
n × C1 of C
n and a real curve C in C1.
Let S2n+3 denote the unit hypersphere in Cn+2 centered at the origin and
put U(1) = {λ ∈ C : λλ = 1}. Then there is a U(1)-action on S2n+3 defined by
z 7→ λz. At z ∈ S2n+3 the vector V = iz is tangent to the flow of the action.
The quotient space S2n+3/ ∼, under the identification induced from the action,
is a complex projective space CP n+1 which endows with the canonical Fubini-
Study metric of constant holomorphic sectional curvature 4. The almost complex
structure J on CP n+1 is induced from the complex structure J on Cn+2 via the
Hopf fibration: π : S2n+3 → CP n+1. It is well-known that the Hopf fibration π
is a Riemannian submersion such that V = iz spans the vertical subspaces.
Let φ : M → CP n+1 be an isometric immersion. Then M = π−1(M) is a
principal circle bundle over M with totally geodesic fibers. The lift φ : M →S2n+3 of φ is an isometric immersion so that the diagram:
M CP n+1
M S2n+3
................................................................................................................
...............
φ
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
π
................................................................................................................
...............
φ
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
π
commutes.
Conversely, if ψ : M → S2n+3 is an isometric immersion which is invariant
under the U(1)-action, then there is a unique isometric immersion ψπ : π(M ) →
134 BANG-YEN CHEN
CP n+1 such that the associated diagram commutes. We simply call the immer-
sion ψπ : π(M ) → CP n+1 the projection of ψ : M → S2n+3.
For a given vector X ∈ Tz(CPn+1) and a point u ∈ S2n+2 with π(u) = z,
we denote by X∗u the horizontal lift of X at u via π. There exists a canonical
orthogonal decomposition:
TuS2n+3 = (Tπ(u)CP
n+1)∗u ⊕ Span {Vu}. (5.2)
Since π is a Riemannian submersion, X and X∗u have the same length.
We put
S2n+1∗ =
{
(z0, . . . , zn) :n
∑
k=0
zkzk = 1, z0 6= 0}
, CP n0 = π(S2n+1
∗ ). (5.3)
Theorem 5.3.([24]) Suppose that a is a positive number and γ(t) = (Γ1(t),
Γ2(t)) is a unit speed Legendre curve γ : I → S3(a2) ⊂ C2 defined on an open
interval I. Let x : S2n+1∗ × I → C
n+2 be the map defined by
x(z0, . . . , zn, t) = (aΓ1(t)z0, aΓ2(t)z0, z1, . . . , zn),n
∑
k=0
zkzk = 1. (5.4)
Then
(i) x induces an isometric immersion ψ : S2n+1∗ ×a|z0| I → S2n+3.
(ii) The image ψ(S2n+1∗ ×a|z0| I) in S2n+3 is invariant under the action of U(1).
(iii) the projection ψπ : π(S2n+1∗ ×a|z0|I) → CP n+1 of ψ via π is a warped product
hypersurface CP n0 ×a|z0| I in CP n+1.
Conversely, if a real hypersurface in CP n+1 is a warped product N ×f I of
a complex hypersurface N of CP n+1 and an open interval I, then, up to rigid
motions, it is locally obtained in the way described above.
In the complex pseudo-Euclidean space Cn+21 endowed with pseudo-Euclidean
metric
g0 = −dz0dz0 +n+1∑
j=1
dzjdzj , (5.5)
we define the anti-de Sitter space-time by
H2n+31 = {(z0, z1, . . . , zn+1) : 〈z, z〉 = −1}. (5.6)
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 135
It is known that H2n+31 has constant sectional curvature −1. There is a U(1)-
action on H2n+31 defined by z 7→ λz. At a point z ∈ H2n+3
1 , iz is tangent to the
flow of the action. The orbit is given by zt = eitz with dzt
dt= izt which lies in
the negative-definite plane spanned by z and iz. The quotient space H 2n+31 / ∼
is the complex hyperbolic space CHn+1 which endows a canonical Kahler metric
of constant holomorphic sectional curvature −4. The complex structure J on
CHn+1 is induced from the canonical complex structure J on Cn+21 via the totally
geodesic fibration: π : H2n+31 → CHn+1.
Let φ : M → CHn+1 be an isometric immersion. Then M = π−1(M) is a
principal circle bundle over M with totally geodesic fibers. The lift φ : M →H2n+3
1 of φ is an isometric immersion such that the diagram:
M CHn+1
M H2n+31
................................................................................................................
...............
φ
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
π
................................................................................................................
...............
φ
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
π
commutes.
Conversely, if ψ : M → H2n+31 is an isometric immersion which is invariant
under the U(1)-action, there is a unique isometric immersion ψπ : π(M ) →CHn+1, called the projection of ψ so that the associated diagram commutes. We
put
H2n+11∗ = {(z0, . . . , zn) ∈ H2n+1
1 : zn 6= 0}, (5.7)
CHn∗ = π(H2n+1
1∗ ). (5.8)
Theorem 5.4.([24]) Suppose that a is a positive number and γ(t) = (Γ1(t),
Γ2(t)) is a unit speed Legendre curve γ : I → S3(a2) ⊂ C2. Let y : H2n+1
1∗ × I →C
n+21 be the map defined by
y(z0, . . . , zn, t) = (z0, . . . , zn−1, aΓ1(t)zn, aΓ2(t)zn), (5.9)
z0z0 −n
∑
k=1
zkzk = 1. (5.10)
Then
136 BANG-YEN CHEN
(i) y induces an isometric immersion ψ : H2n+11∗ ×a|zn| I → H2n+3
1 .
(ii) The image ψ(H2n+11∗ ×a|zn| I) in H2n+3
1 is invariant under the U(1)-action.
(iii) the projection ψπ : π(H2n+11∗ ×a|zn| I) → CHn+1 of ψ via π is a warped
product hypersurface CHn∗ ×a|zn| I in CHn+1.
Conversely, if a real hypersurface in CHn+1 is a warped product N ×f I of a
complex hypersurface N and an open interval I, then, up to rigid motions, it is
locally obtained in the way described above.
6. Warped Products as Lagrangian Submanifolds
A submanifold N in a Kahler manifold M is called a totally real submanifold
if the almost complex structure J of M carries each tangent space TxN of N
into its corresponding normal space T⊥x N [34]. The submanifold N is called a
holomorphic submanifold (or Kahler submanifold) if J carries each TxN into itself
[48].
A totally real submanifold N in a Kahler manifold M is called a Lagrangian
submanifold if dimRN = dimC M . (For the most recent survey on differential
geometry of Lagrangian submanifolds, see [17, 22].)
For Lagrangian immersions into complex Euclidean n-space Cn, a well-known
result of M. Gromov [41] states that a compact n-manifold M admits a La-
grangian immersion (not necessary isometric) into Cn if and only if the complex-
ification TM ⊗ C of the tangent bundle of M is trivial. In particular, Gromov’s
result implies that there exists no topological obstruction to Lagrangian immer-
sions for compact 3-manifolds in C3, because the tangent bundle of a 3-manifold
is always trivial.
From the Riemannian point of view, it is natural to ask the following basic
question.
Problem 6.1. When a Riemannian n-manifold admits a Lagrangian iso-
metric immersion into Cn?
Not every warped product N1 ×f N2 can be isometrically immersed in a
complex space form as a Lagrangian submanifold. However, for warped products
of curves and the unit (n − 1)-sphere Sn−1, we have the following existence
theorem.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 137
Theorem 6.1.([13, 18]) Every simply-connected open portion of a warped
product manifold I ×f Sn−1 of an open interval I and a unit (n − 1)-sphere
admits an isometric Lagrangian immersion into Cn.
The Lagrangian immersions given in Theorem 6.1 with vanishing scalar cur-
vature have been recently determined explicitly in [5].
The Lagrangian immersions given in Theorem 6.1. are expressed in terms of
complex extensors in the sense of [13] which are defined as follows:
Let G : Mn−1 → Em be an isometric immersion of a Riemannian (n − 1)-
manifold into Euclidean m-space Em and F : I → C a unit speed curve in the
complex plane. We extend the immersion G : M n−1 → Em to an immersion of
I ×Mn−1 into complex Euclidean m-space Cm given by
φ = F ⊗G : I ×Mn−1 → C ⊗ Em = C
m, (6.1)
where F ⊗G is the tensor product immersion of F and G defined by
(F ⊗G)(s, p) = F (s) ⊗G(p), s ∈ I, p ∈Mn−1. (6.2)
Such an extension F ⊗G of the immersion G is called a complex extensor of G
(or of submanifold Mn−1) via F .
Those Lagrangian immersions obtained in Theorem 6.1 show how to embed
a time slice of the Schwarzschild space-time that models the out space around a
massive star or a black hole as Lagrangian submanifolds (cf. [5, 49]).
Since rotation hypersurfaces and real space forms can be at least locally
expressed as the warped products of curves and a unit sphere, Theorem 6.1
implies immediately the following
Corollary 6.1. Every rotation hypersurface of En+1 can be isometrically
immersed as Lagrangian submanifold in Cn.
Corollary 6.2. Every Riemannian n-manifold of constant sectional curva-
ture c can be locally isometrically immersed in Cn as a Lagrangian submanifold.
Corollary 6.3. Every rotation Every simply-connected surface equipped with
a Riemannian metric g = E2(x)(dx2+dy2) always admits a Lagrangian isometric
immersion into C2.
138 BANG-YEN CHEN
Remark 6.1. Not every Riemannian n-manifold of constant sectional curva-
ture can be globally isometrically immersed in Cn as a Lagrangian submanifold.
For instance, it is known from [15] that every compact Riemannian n-manifold
with positive sectional curvature (or positive Ricci curvature) does not admit any
Lagrangian isometric immersion into Cn.
Remark 6.2. In views of Theorem 6.1 it is interesting to ask the following.
Problem 6.2. Determine necessary and/or sufficient conditions for warped
product manifolds to admit Lagrangian isometric immersions into complex Eu-
clidean spaces or more generally into complex space forms.
7. Segre Embedding and Its Extensions
Let (zi0, . . . , z
iαi
) (1 ≤ i ≤ s) denote the homogeneous coordinates of CP αi .
Define a map:
Sα1···αs : CPα1 × · · · × CP αs → CPN , N =s
∏
i=1
(αi + 1) − 1, (7.1)
which maps a point ((z10 , . . . , z
1α1
), . . . , (zs0, . . . , z
sαs
)) of the product Kahler man-
ifold CP α1 × · · · × CP αs to the point (z1i1· · · zs
ij)1≤i1≤α1,...,1≤is≤αs in CPN . The
map Sα1···αs is a Kahler embedding which is called the Segre embedding. The
Segre embedding was constructed by C. Segre in 1891 [52].
The following results obtained in 1981 can be regarded as “converse” to Segre
embedding constructed in 1891.
Theorem 7.1.([19, 32]) Let Mα1
1 , . . . ,Mαss be Kahler manifolds of dimen-
sions α1, . . . , αs, respectively. Then every Kahler immersion
f : Mα1
1 × · · · ×Mαss → CPN , N =
s∏
i=1
(αi + 1) − 1,
of Mα1
1 ×· · ·×Mαss into CPN is locally the Segre embedding, that is, Mα1
1 , . . . ,Mαss
are open portions of CP α1 , . . . , CP αs , respectively, and moreover, the Kahler
immersion f is congruent to the Segre embedding.
This theorem was proved in [44] under two additional assumptions; namely,
s = 2 and the Kahler immersion f has parallel second fundamental form.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 139
Let ∇kσ denote the k-th covariant derivative of the second fundamental form
for k = 0, 1, 2, · · ·. Denoted by ||∇kσ||2 the squared norm of ∇kσ.
Theorem 7.2.([19, 32]) Let Mα1
1 ×· · ·×Mαss be a product Kahler submanifold
of CPN . Then we have
||∇k−2σ||2 ≥ k! 2k∑
i1<···<ik
α1 · · ·αk, (7.2)
for k = 2, 3, · · · .The equality sign of (7.2) holds for some k if and only if M α1
1 , . . . ,Mαss
are open portions of CP α1 , . . . , CP αs , respectively, and the Kahler immersion is
congruent to the Segre embedding.
In particular, if k = 2, Theorem 7.2 reduces to
Theorem 7.3.([19, 32]) Let Mh1 ×Mp
2 be a product Kahler submanifold of
CPN . Then we have
||σ||2 ≥ 8hp. (7.3)
The equality sign of (7.3) holds if and only if M h1 and Mp
2 are open portions
of CP h and CP p, respectively, and the Kahler immersion is congruent to the
Segre embedding Sh,p.
8.CR-Products
A submanifold N in a Kahler manifold M is called a CR-submanifold [3] if
there exists on N a holomorphic distribution D whose orthogonal complement
D⊥ is a totally real distribution, i.e., JD⊥x ⊂ T⊥
x N .
A CR-submanifold of a Kahler manifold M is called a CR-product [8] if it is
a Riemannian product NK ×N⊥ of a Kahler submanifold NK and a totally real
submanifold N⊥.
For CR-products in complex space forms, the following are known.
Theorem 8.1.([8]) We have
(i) A CR-submanifold in Cm is a CR-product if and only if it is a direct sum
of a Kahler submanifold and a totally real submanifold of linear complex
subspaces.
140 BANG-YEN CHEN
(ii) There do not exist CR-products in complex hyperbolic spaces other than
Kahler submanifolds and totally real submanifolds.
CR-products NK×N⊥ in CP h+p+hp are obtained from the Segre embedding;
namely, we have the following.
Theorem 8.2.([8]) Let NhT ×Np
⊥ be the CR-product in CPm with constant
holomorphic sectional curvature 4. Then
m ≥ h+ p+ hp. (8.1)
The equality sign of (8.1) holds if and only if
(a) NhT is a totally geodesic Kahler submanifold,
(b) Np⊥ is a totally geodesic totally real submanifold, and
(c) the immersion is given by
NhT ×Np
⊥
t.g.
−−−−→CP h ×CP pShp
−−−−−−−−→Segre imbedding
CP h+p+hp. (8.2)
Theorem 8.3.([8]) Let NhT ×Np
⊥ be the CR-product in CPm with constant
holomorphic sectional curvature 4. Then the squared norm of the second funda-
mental form satisfies
||σ||2 ≥ 4hp. (8.3)
The equality sign of (8.3) holds if and only if
(a) NhT is a totally geodesic Kahler submanifold,
(b) Np⊥ is a totally geodesic totally real submanifold, and
(c) the immersion is given by
NhT ×Np
⊥
t.g.
−−−−→CP h ×CP pShp
−−−−−−−−→Segre imbedding
CP h+p+hp ⊂ CPm. (8.4)
9. Warped Products as CR-Submanifolds
For warped product CR-submanifolds in complex space forms, we propose
the following simple question:
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 141
Problem 9.1.
∀
NK ×f N⊥
or
N⊥ ×f NN
CR-immersion
−−−−−−−−→ Cm or Mm(4c) =⇒ ???
where Mm(4c) denotes a Kahler manifold of constant holomorphic sectional cur-
vature 4c.
First we mention the following.
Theorem 9.1.([19]) If N⊥×f NK is a warped product CR-submanifold of a
Kahler manifold M such that N⊥ is a totally real and NK a Kahler submanifold
of M , then it is a CR-product.
Theorem 9.1 shows that there does not exist warped productCR-submanifolds
of the form N⊥×fNK other than CR-products. So, we shall only consider warped
product CR-submanifolds of the form: NK ×f N⊥, by reversing the two factors
NK and N⊥. We simply call such CR-submanifolds CR-warped products.
CR-warped products are characterized as follows.
Theorem 9.2.([19]) A proper CR-submanifold M of a Kahler manifold M
is locally a CR-warped product if and only if
AJZX = ((JX)µ)Z, X ∈ D, Z ∈ D⊥, (9.1)
for some function µ on M satisfying Wµ = 0, W ∈ D⊥.
We have the following solution to Problem 9.1.
Theorem 9.3.([19]) Let NK ×f N⊥ be a CR-warped product in a Kahler
manifold M . Then
||σ||2 ≥ 2p ||∇(ln f)||2, (9.2)
where ∇ ln f is the gradient of ln f on NK and p = dimN⊥.
If the equality sign of (9.2) holds identically, then NK is a totally geodesic
Kahler submanifold and N⊥ is a totally umbilical totally real submanifold of M .
Moreover, NK ×f N⊥ is minimal in M .
142 BANG-YEN CHEN
When M is anti-holomorphic, i.e., when JD⊥x = T⊥
x N , and p > 1. The
equality sign of (9.2) holds identically if and only if N⊥ is a totally umbilical
submanifold of M .
If M is anti-holomorphic and p = 1, then the equality sign of (9.2) holds
identically if and only if the characteristic vector field Jξ of M is a principal
vector field with zero as its principal curvature. (Notice that in this case, M
is a real hypersurface in M .) Also, in this case, the equality sign of (9.2) holds
identically if and only if M is a minimal hypersurface in M .
Remark 9.1. Theorem 9.1 and Theorem 9.3 have been extended to twisted
product CR-submanifolds in [16].
CR-warped products in complex space forms satisfying the equality case of
(9.2) have been completely classified. In fact, we have the following.
Theorem 9.4.([19]) A CR-warped product NK ×f N⊥ in Cm satisfies
||σ||2 = 2p||∇(ln f)||2, (9.3)
if and only if
(i) NK is an open portion of a complex Euclidean h-space Ch,
(ii) N⊥ is an open portion of the unit p-sphere Sp,
(iii) there is a = (a1, . . . , ah) ∈ Sh−1 ⊂ Eh such that f =
√
〈a, z〉2 + 〈ia, z〉2 for
z = (z1, . . . , zh) ∈ Ch, w = (w0, . . . , wp) ∈ Sp ⊂ E
p+1, and
(iv) up to rigid motions, the immersion is given by
x(z, w) =(
z1 + (w0 − 1)a1
h∑
j=1
ajzj , · · · zh + a(w0 − 1)ah
h∑
j=1
ajzj ,
w1
h∑
j=1
ajzj , · · · , wp
h∑
j=1
ajzj , 0, · · · , 0)
. (9.4)
A CR-warped product NT ×fN⊥ is said to be trivial if its warping function f
is constant. A trivial CR-warped product NT ×fN⊥ is nothing but a CR-product
NT ×Nf⊥, where N f
⊥ is the manifold with metric f 2gN⊥which is homothetic to
the original metric gN⊥on N⊥.
The following result classifies CR-warped products in complex projective
spaces satisfying the equality case of (9.2)
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 143
Theorem 9.5.([20]) A non-trivial CR-warped product NT ×fN⊥ in CPm(4)
satisfies the basic equality ||σ||2 = 2p||∇(ln f)||2 if and only if we have
(1) NT is an open portion of complex Euclidean h-space Ch,
(2) N⊥ is an open portion of a unit p-sphere Sp, and
(3) up to rigid motions, the immersion x of NT ×f N⊥ into CPm(4) is the com-
position π ◦ x, where
x(z, w) =(
z0 + (w0 − 1)a0
h∑
j=0
ajzj , · · · , zh + (w0 − 1)ah
h∑
j=0
ajzj ,
w1
h∑
j=0
ajzj , · · · , wp
h∑
j=0
ajzj, 0, . . . , 0)
, (9.5)
π is the projection π : Cm+1∗ → CPm(4), a0, . . . , ah are real numbers satisfy-
ing a20 +a2
1 + · · ·+a2h = 1, z = (z0, z1, . . . , zh) ∈ Ch+1 and w = (w0, . . . , wp) ∈
Sp ⊂ Ep+1.
Similarly, we have the following classification theorems for CR-warped prod-
ucts in complex hyperbolic spaces satisfying the equality case of (9.2).
Theorem 9.6.([20]) A CR-warped product NT ×f N⊥ in CHm(−4) satisfies
the basic equality ||σ||2 = 2p||∇(ln f)||2 if and only if one of the following two
cases occurs:
(1) NT is an open portion of complex Euclidean h-space Ch, N⊥ is an open por-
tion of a unit p-sphere Sp and, up to rigid motions, the immersion is the
composition π ◦ x, where π is the projection π : Cm+1∗1 → CHm(−4) and
x(z,w) =(
z0 + a0(1−w0)h
∑
j=0
ajzj , z1 + a1(w0 − 1)h
∑
j=0
ajzj , · · · ,
zh+ah(w0−1)h
∑
j=0
ajzj , w1
h∑
j=0
ajzj , . . . , wp
h∑
j=0
ajzj , 0, . . . , 0)
, (9.6)
for some real numbers a0, . . . , ah satisfying a20 − a2
1 − · · · − a2h = −1, where
z = (z0, . . . , zh) ∈ Ch+11 and w = (w0, . . . , wp) ∈ Sp ⊂ E
p+1.
(2) p = 1, NT is an open portion of Ch and, up to rigid motions, the immersion
144 BANG-YEN CHEN
is the composition π ◦ x, where
x(z, t) =(
z0 + a0(cosh t− 1)h
∑
j=0
ajzj , z1 + a1(1 − cosh t)h
∑
j=0
ajzj ,
. . . , zh + ah(1 − cosh t)h
∑
j=0
ajzj , sinh th
∑
j=0
ajzj , 0, . . . , 0)
, (9.7)
for some real numbers a0, a1 . . . , ah+1 satisfying a20 − a2
1 − · · · − a2h = 1.
10. Another Solution to Problem 9.1
In fact, CR-warped products in a complex space form also satisfy another
general inequality; which provides the second solution to Problem 9.1.
Theorem 10.1.([26]) Let N = NhK ×f N
p⊥ be a CR-warped product in a
complex space form M(4c) of constant holomorphic sectional curvature c. Then
we have
||σ||2 ≥ 2p{||∇(ln f)||2 + ∆(ln f) + 2hc}. (10.1)
If the equality sign of (10.1) holds identically, then NK is a totally geodesic
submanifold and N⊥ is a totally umbilical submanifold. Moreover, N is a minimal
submanifold in M(4c).
One may also classifies CR-warped products which satisfy the equality case
of (10.1) as given in the following theorems.
Theorem 10.2.([26]) Let φ : NhK ×f N
p⊥ → C
m be a CR-warped product in
Cm. Then we have
||σ||2 ≥ 2p{||∇(ln f)||2 + ∆(ln f)}. (10.2)
The equality case of (10.2) holds if and only if
(a) NK is an open portion of Ch∗ =: C
h − {0};(b) N⊥ is an open portion of Sp;
(c) There is α, 1 ≤ α ≤ h, and complex Euclidean coordinates {z1, . . . , zh} on Ch
such that f =√
∑αj=1 zj zj ;
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 145
(d) Up to rigid motions, the immersion φ is given by
φ = (w0z1, . . . , wpz1, . . . , w0zα, . . . , wpzα, zα+1, . . . , zh, 0, . . . , 0), (10.3)
for z = (z1, . . . , zh) ∈ Ch∗ and w = (w0, . . . , wp) ∈ Sp ⊂ E
p+1.
Theorem 10.3.([26]) Let φ : NT ×f N⊥ → CPm(4) be a CR-warped product
with dimCNT = h and dimRN⊥ = p. Then we have
||σ||2 ≥ 2p{||∇(ln f)||2 + ∆(ln f) + 2h}. (10.4)
The CR-warped product satisfies the equality case of (10.4) if and only if
(i) NT is an open portion of complex projective h-space CP h(4);
(ii) N⊥ is an open portion of unit p-sphere Sp; and
(iii) There exists a natural number α ≤ h such that, up to rigid motions, φ is the
composition π ◦ φ, where
φ(z, w) = (w0z0, . . . , wpz0, . . . , w0zα, . . . , wpzα, zα+1, . . . , zh, 0 . . . , 0),
(10.5)
for z = (z0, . . . , zh) ∈ Ch+1∗ and w = (w0, . . . , wp) ∈ Sp ⊂ E
p+1, where π is
the projection π : Cm+1∗ → CPm(4).
Theorem 10.4.([26]) Let φ : NT ×f N⊥ → CHm(−4) be a CR-warped
product with dimC NT = h and dimRN⊥ = p. Then we have
||σ||2 ≥ 2p{||∇(ln f)||2 + ∆(ln f) − 2h}. (10.6)
The CR-warped product satisfies the equality case of (10.6) if and only if
(a) NT is an open portion of complex hyperbolic h-space CHh(−4);
(b) N⊥ is an open portion of unit p-sphere Sp (or R, when p = 1); and
(c) up to rigid motions, φ is the composition π ◦ φ, where either φ is given by
φ(z, w) = (z0, · · · , zβ , w0zβ+1, · · · , wpzβ+1, · · · , w0zh, . . . , wpzh, 0 · · · , 0),(10.7)
for 0 < β ≤ h, z = (z0, . . . , zh) ∈ Ch+1∗1 and w = (w0, . . . , wp) ∈ Sp, or φ is
given by
φ(z, u) = (z0 cosh u, z0 sinhu, z1 cos u, z1 sinu, · · · , · · · , zα cos u,
zα sinu, zα+1, · · · , zh, 0 · · · , 0), (10.8)
146 BANG-YEN CHEN
for z = (z0, . . . , zh) ∈ Ch+1∗1 and u ∈ R, where π is the projection π : C
m+1∗1 →
CHm(−4).
11. Convolution of Riemannian Manifolds
Convolution of Riemannian manifolds introduced in [23, 29] is a natural
extension of direct and warped products. Convolution of Riemannian manifolds
is defined as follows:
Let (N1, g1) and (N2, g2) be two Riemannian manifolds and f and h be
positive functions on N1 and N2, respectively. Consider the symmetric tensor
field gf,h of type (0,2) on N1 ×N2 defined by
hg1 ∗f g2 = h2g1 + f2g2 + 2fhdf ⊗ dh, (11.1)
which is called the convolution of g1 and g2 via h and f . The product manifold
N1 × N2 equipped with hg1 ∗f g2 is called a convolution manifold, denoted by
hN1F fN2.
When the scale functions f, h are irrelevant, we simply denote hN1F fN2 and
hg1 ∗f g2 by N1FN2 and g1 ∗ g2, respectively.
If hg1 ∗f g2 is nondegenerate, it defines a pseudo-Riemannian metric on
N1×N2 with index ≤ 1. In this case, hg1 ∗f g2 is called a convolution metric and
hN1FfN2 is called a convolution pseudo-Riemannian manifold. If the index of the
pseudo-Riemannian metric is zero, hN1FfN2 is called a convolution Riemannian
manifold.
The notion of convolution arises naturally, since each tensor product immer-
sion gives rise to a convolution as shown in the next theorem.
Theorem 11.1.([23]) Let ψ1 : (N1, g1) → En∗ ⊂ E
n and ψ2 : (N2, g2) →E
m∗ ⊂ E
m be isometric immersions, where En∗ = E
n − {0}. Then
ψ = ψ1 ⊗ ψ2 : N1 ×N2 → En ⊗ E
m = Enm; (u, v) 7→ ψ1(u) ⊗ ψ2(v), (11.2)
gives rise to a convolution manifold N1FN2 equipped with
ρ2g1 ∗ρ1
g2 = ρ22g1 + ρ2
1g2 + 2ρ1ρ2dρ1 ⊗ dρ2, (11.3)
where ρ1 and ρ2 are the distance functions of ψ1 and ψ2, respectively.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 147
Therefore, we have the following induced metrics:
φ1 ⊕ φ2 : N1 ×N2 → Em1 ⊕ E
m2 =⇒ g1 + g2 (direct sum)
ψ1 ⊗ ψ2 : N1 ×N2 → En ⊗ E
m =⇒ ρ2g1 ∗ρ1
g2 (convolution)
Example 11.1. If ψ2 : (N2, g2) → Em is an isometric immersion such that
ψ2(N2) is contained in Sm−1(1) ⊂ Em which is centered at the origin. Then the
convolution ρ2g1 ∗ρ1
g2 on N1FN2 is the warped product metric: g1 + ρ21g2.
Definition 11.1. A convolution hg1 ∗ fg2 of two Riemannian metrics g1 and
g2 is called degenerate if det(hg1 ∗f g2) = 0 holds identically.
For X ∈ T (Nj), j = 1, 2, we denote by |X|j the length of X with respect to
metric gj on Nj .
The following theorem determines when a convolution is degenerate.
Theorem 11.2.([23]) Let hN1F fN2 be the convolution of Riemannian man-
ifolds (N1, g1) and (N2, g2) via h and f . Then hg1 ∗f g2 is degenerate if and only
if we have
(1) the length |grad f |1 of the gradient of f on (N1, g1) is a nonzero constant, say
c, and
(2) the length |grad h|2 of the gradient of h on (N2, g2) is the constant given by
1/c, i.e., the reciprocal of c.
The following result provides a necessary and sufficient condition for hg1∗f g2
to be Riemannian.
Theorem 11.3.([23]) Let hN1F fN2 be the convolution of Riemannian man-
ifolds (N1, g1) and (N2, g2) via h and f . Then hg1 ∗ fg2 is Riemannian if and
only if we have |grad f |1 |gradh|2 < 1.
In views of Theorems 11.2 and 11.3, it is natural to study Euclidean sub-
manifolds whose distance function ρ satisfies the condition: |grad ρ| = constant.
To do so, we introduce the notion of constant-ratio submanifolds.
Definition 11.2. Let x : M → En be an isometric immersion. We put
x = xN + xT , (11.4)
148 BANG-YEN CHEN
where xN and xT are the normal and tangential components of the position vector
field, respectively. The submanifold M is said to be of constant-ratio if the ratio
|xN | : |xT | is constant; or equivalently, |xT | : |x| = c = constant (cf. [21]).
The following results characterize constant-ratio submanifolds.
Theorem 11.4.([23]) Let x : M → En be an isometric immersion. Then
M is of constant-ratio with |xT | : |x| = c if and only if |grad ρ| = c which is
constant.
In particular, for submanifold of constant-ratio, we have |grad ρ| = c ≤ 1.
Theorem 11.3 and Theorem 11.4 imply
Corollary 11.1. Let ψ1 : (N1, g1) → En and ψ2 : (N2, g2) → E
m be isometric
immersions and ρ1 = |ψ1| and ρ2 = |ψ2| be their distance functions. Then
ρ2g1 ∗ρ1
g2 is degenerate if and only if |grad ρ1| = |grad ρ2| = 1.
Space-like constant-ratio submanifolds in pseudo-Euclidean space have been
completely classified in [25]. In particular, for constant-ratio submanifolds in
Euclidean space, we have the following.
Theorem 11.5.([25]) An isometric immersion x : M → Em is of constant-
ratio if and only if one of the following three cases occurs:
(1) x(M) is contained in hypercone with vertex at the origin.
(2) x(M) is contained in a hypersphere Sm−1(r2) of Em centered at the origin.
(3) There is a number b ∈ (0, 1) and local coordinate system {s, u2, . . . , un} on
M such that the immersion x is given by
x(s, u2, . . . , un) = bs Y (s, u2, . . . , un), (11.5)
where Y = Y (s, u2, . . . , un) satisfies conditions:
(3.a) Y = Y (s, u2, . . . , un) lies in Sm−1(1),
(3.b) Ys is perpendicular to Yu2, . . . , Yun , and
(3.c) |Ys| =√
1 − b2/(bs).
Remark 11.1. Submanifolds of constant-ratio also relate to a problem in
physics concerning the motion in a central force field which obeys the inverse-
cube law. For example, the trajectory of a particle subject to a central force of
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 149
attraction located at the origin which obeys the inverse-cube law is a curve of
constant-ratio.
Remark 11.2. The inverse-cube law was originated from Sir Isaac Newton
in his letter sent to Robert Hooke on December 13, 1679. This letter is of great
historical importance because it reveals the state of Newton’s development of
dynamics at that time (cf. [45, 46]).
12. Applications of Convolution
We define an isometric map from a convolution manifold into a Riemannian
manifold as follows.
Definition 12.1. A map ψ : (hN1FfN2, hg1 ∗ fg2) → (M, g) from a convo-
lution manifold into a Riemannian manifold is called isometric if ψ∗g = hg1 ∗ fg2.
For example, let z : Ch∗ → C
h and x : Ep∗ → E
p be inclusion maps. Consider
the map
ψz,x = z ⊗ x : Ch∗ × E
p∗ → (Chp, g0), (12.1)
defined by
ψz,x = z ⊗ x = (z1x1, . . . , z1xp, . . . , zhx1, . . . , zhxp), (12.2)
for z = (z1, . . . , zh) ∈ Ch∗ and x = (x1, . . . , xp) ∈ E
p∗. Then we have
ψ∗z,x(g0) = ρ2
g1 ∗ρ1g2 = ρ2
2g1 + ρ21g2 + 2ρ1ρ2dρ1 ⊗ dρ2, (12.3)
where g1, g2 and ρ1, ρ2 are the metrics and distance functions of Ch∗ ,E
p∗, respec-
tively. Thus
ψz,x : (Ch∗ × E
p∗, ρ2
g1 ∗ρ1g2) → (Chp, g0) (12.4)
is an isometric map.
An isometric map
φ : (Ch∗ × E
p∗, ρ2
g1 ∗ρ1g2) → (Cm, g0) (12.5)
is call a CR-map if φ carries each slice Ch∗ × {v} of C
h∗ × E
p∗ into a Kahler sub-
manifold and each slice {u} × Ep∗ of C
h∗ × E
p∗ into a totally real submanifold.
150 BANG-YEN CHEN
The notion of convolution is useful in the study of tensor product immersions.
For instance, we may applying the notion of convolution to obtain the following
two simple characterizations of the map ψz,x = z⊗ x : Ch∗ × E
p∗ → C
hp defined by
(12.2).
Theorem 12.1.([29]) Let φ : U → Cm be an isometric CR-map from an
open portion U of the convolution manifold (Ch∗ × E
p∗, ρ2
g1 ∗ρ1g2) into complex
Euclidean m-space. Then we have
(1) m ≥ hp.
(2) If m = hp, then, up to rigid motions of Cm, φ is the map ψz,x = z⊗x defined
by (12.2).
Theorem 12.2.([29]) Let φ : U → Cm be an isometric CR-map from an
open portion U of the convolution manifold (Ch∗ × E
p∗, ρ2
g1 ∗ρ1g2) into a complex
Euclidean space. Then the second fundamental form σ of φ satisfies
||σ||2 ≥ (2h − 1)(p− 1)
ρ21ρ
22
. (12.6)
The equality sign of (12.6) holds identically if and only if, up to rigid motions
of Cm, φ is given by φ(z, x) = (z ⊗ x, 0, . . . , 0).
These two theorems can be regarded as Euclidean version of Theorems 7.1
(with s = 2) and 7.3 related with the converse to Segre imbedding.
13. T -Constant and N-Constant Submanifolds
In views of constant-ratio submanifolds in Euclidean space, it is natural to
study submanifolds in Euclidean space such that either xT or xN has constant
length, where xT and xN are the tangential and the normal components of the
position function given in (11.4). For simplicity, we call a submanifold M in a
Euclidean space a T -constant (respectively, N -constant) submanifold if xT is of
constant length (respectively, |xN | is of constant length).
For T -constant submanifolds, we have the following.
Theorem 13.1.([30]) Let x : M → Em be an isometric immersion of a
Riemannian n-manifold into the Euclidean m-space. Then M is a T -constant
submanifold if and only if either
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 151
(i) x(M) is contained in a hypersphere Sm−1(r2) of Em, or
(ii) there exist a real numbers a, b and local coordinate systems {s, u2, . . . , un}on M such that the immersion x is given by
x(s, u2, . . . , un) =√
a2 + b+ 2as Y (s, u2, . . . , un), (13.1)
where Y = Y (s, u2, . . . , un) satisfies conditions:
(2.a) Y = Y (s, u2, . . . , un) lies in the unit hypersphere Sm−1(1),
(2.b) the coordinate vector field Ys is perpendicular to coordinate vector fields
Yu2, . . . , Yun , and
(2.c) Ys satisfies |Ys| =√b+ 2as/(a2 + b+ 2as).
For N -constant submanifolds, we have the following.
Theorem 13.2.([30]) Let x : M → Em be an isometric immersion of a
Riemannian n-manifold into the Euclidean m-space. Then M is N -constant if
and only if one of the following three cases occurs:
(1) x is a conic submanifold with the vertex at the origin.
(2) x(M) is contained in a hypersphere Sm−1(r2) of Em.
(3) There exist a positive number c and local coordinate systems {s, u2, . . . , un}on M such that the immersion x is given by
x(s, u2, . . . , un) =√
s2 + c2 Y (s, u2, . . . , un), (13.2)
where Y = Y (s, u2, . . . , un) satisfies conditions:
(3.a) Y = Y (s, u2, . . . , un) lies in the unit hypersphere Sm−1(1),
(3.b) the coordinate vector field Ys is perpendicular to coordinate vector fields
Yu2, . . . , Yun , and
(3.c) Ys satisfies |Ys| = c/(s2 + c2).
14. Rectifying curves and τ/κ
Consider a unit-speed space curve x : I → R3, I = (α, β), in E
3 that has at
least four continuous derivatives. Let t denote x′. It is possible that t′(s) = 0
for some s, however, we assume that this never happens. Then we can introduce
a unique vector field n and positive function κ so that t′ = κn. We call t′ the
152 BANG-YEN CHEN
curvature vector field, n the principal normal vector field, and κ the curvature.
Since t is a constant length vector field, n is orthogonal to t. The binormal vector
field is defined by b = t×n which is a unit vector field orthogonal to both t and
n. One defines the torsion τ by the equation b′ = −τn.
The famous Serret-Frenet equations are given by
t′ = κn,
n′ = −κt + τb, (14.1)
b′ = −τn.
At each point of the curve, the planes spanned by {t,n}, {t,b} and {n,b}are known as the osculating plane, the rectifying plane and the normal plane,
respectively. A curve in E3 is called a twisted curve if has nonzero curvature and
torsion.
In elementary differential geometry it is well-known that a curve in E3 is a
planar curve if its position vector lies in its osculating plane at each point. And
it is a spherical curve if its position vector lies in its normal plane at each point.
In views of these basic facts, it is natural to ask the following simple geometric
question:
Problem 14.1. When the position vector of a space curve x : I → R3 always
lies in its rectifying plane ?
For simplicity we call such a curve a rectifying curve. So, for a rectifying
curve x : I → E3, the position vector x satisfies
x(s) = λ(s)t(s) + µ(s)b(s), (14.2)
for some functions λ and µ.
The following result provides some simple characterizations of rectifying
curves. In particular, it shows that rectifying curves in E3 are N -constant curves.
Theorem 14.1.([31]) Let x : I → E3 be a rectifying curve in E
3 with κ > 0
and s be its arclength function. Then we have
(i) The curve is N -constant and the distance function ρ is non-constant.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 153
(ii) The distance function ρ = |x| satisfies ρ2 = s2 + c1s+ c2 for some constants
c1 and c2.
(iii) The tangential component of the position vector of the curve is given by
〈x, t〉 = s+ c for some constant c.
(iv) the torsion τ is non-zero and the binormal component of the position vector
is constant, i.e. 〈x,b〉 is constant.
Conversely, if x : I → E3 is a curve with κ > 0 and one of (i), (ii), (iii) or
(iv) holds, then x is a rectifying curve.
The fundamental existence and uniqueness theorem for curves in E3 states
that a curve is uniquely determined up to rigid motions by its curvature and
torsion, given as functions of its arclength. Furthermore, given continuous func-
tions κ(s) and τ(s), with κ(s) positive and continuously differentiable, there is
a differentiable curve (of class at least C3) with curvature κ and torsion τ . In
general, a result of Lie and Darboux shows that solving Serret-Frenet equations
is equivalent to solving a certain complex Riccati equation (see [53, p.36]). In
practice, space curves with prescribed curvature and torsion are often impossible
to find explicitly. Fortunately, we are able to determine completely all rectifying
curves in E3.
Theorem 14.2.([31]) Let x : I → E3 be a curve E
3 with κ > 0. Then it is a
rectifying curve if and only if, up to parameterization, it is given by
x(t) = a(sec t)y(t), (14.3)
where a is a positive number and y = y(t) is a unit speed curve in S2, where S2
denotes the unit sphere in E3 centered at the origin.
A generalized helix in Euclidean 3-space is a twisted curve whose unit tangent
vector field t makes constant angle with a fixed direction. It is well-known that
a twisted curve is a generalized helix if and only if the ratio τ/κ is a non-zero
constant. So, it is natural to ask the following.
Problem 14.2. When the ratio τ/κ of a twisted curve is a non-constant
linear function in arclength function ?
The following result provides an answer to this Problem.
154 BANG-YEN CHEN
Theorem 14.3.([31]) Let x : I → E3 be a twisted curve in E
3. Then x is
congruent to a rectifying curve if and only if the ratio of torsion and curvature
of the curve is a non-constant linear function in arclength function s, i.e., τ/κ =
c1s+ c2 for some constants c1, c2 with c1 6= 0.
Theorem 14.3 seems to be of independent interest by itself.
References
[1] J. Arroyo, M. Barros and O. J. Garay, Willmore-Chen tubes on homogeneous spaces in
warped product spaces, Pacific J. Math., 188 (1999), 201-207.
[2] J. K. Beem, P. E. Ehrlich and Th. G. Powell, Warped product manifolds in relativity,
Selected Studies: Physics-Astrophysics, Mathematics, History of Science, North-Holland,
New York, 1982.
[3] A. Bejancu, Geometry of CR-Submanifolds, D. Reidel Publ. Co., 1986.
[4] R. L. Bishop and B. O’Neill, Manifolds of negative curvature, Trans. Amer. Math. Soc.,
145 (1969), 1-49.
[5] I. Castro and C. R. Montealegre, Lagrangian submanifolds with zero scalar curvature in
complex Euclidean space, Geom. Dedicata, 86(2001), 179-183.
[6] I. Castro, C. R. Montealegre and F. Urbano, Closed conformal vector fields and Lagrangian
submanifolds in complex space forms, Pacific J. Math., 199(2001), 269-302.
[7] B. Y. Chen, Geometry of Submanifolds, M. Dekker, New York, 1973.
[8] B. Y. Chen, Some CR-submanifolds of a Kaehler manifold. I, J. Differential Geometry,
16(1981), 305-322.
[9] B. Y. Chen, Some CR-submanifolds of a Kaehler manifold. II, J. Differential Geometry,
16(1981), 493-509.
[10] B. Y. Chen, Geometry of Submanifolds and Its Applications, Science University of Tokyo,
1981.
[11] B. Y. Chen, Some pinching and classification theorems for minimal submanifolds, Arch.
Math., 60(1993), 568-578.
[12] B. Y. Chen, Jacobi’s elliptic functions and Lagrangian immersions, Proc. Royal Soc.
Edinburgh Sect. A, Math., 126(1996), 687-704.
[13] B. Y. Chen, Complex extensors and Lagrangian submanifolds in complex Euclidean spaces,
Tohoku Math. J., 49(1997), 277-297.
[14] B. Y. Chen, Interaction of Legendre curves and Lagrangian submanifolds, Israel J. Math.,
99(1997), 69-108.
[15] B. Y. Chen, Some new obstructions to minimal and Lagrangian isometric immersions,
Japan. J. Math., 26(2000), 105-127.
[16] B. Y. Chen, Twisted product CR-submanifolds in Kaehler manifolds, Tamsui Oxford J.
Math. Sci., 16(2000), 105-121.
[17] B. Y. Chen, Riemannian Submanifolds, Handbook of Differential Geometry, Vol. I (2000)
(eds. F. Dillen and L. Verstraelen), North Holland Publ., pp. (1)(2000),187-418.
[18] B. Y. Chen, Complex extensors, warped products and Lagrangian immersions, Soochow J.
Math., 26(2000), 1-17.
GEOMETRY OF WARPED PRODUCTS AS RIEMANNIAN SUBMANIFOLDS 155
[19] B. Y. Chen, Geometry of warped product CR-submanifolds in Kaehler manifolds, Monatsh.
Math., 133(2001), 177-195.
[20] B. Y. Chen, Geometry of warped product CR-submanifolds in Kaehler manifolds, II, Monatsh.
Math., 135(2001).
[21] B. Y. Chen, Constant-ratio hypersurfaces, Soochow J. Math., 21(2001), 353-361.
[22] B. Y. Chen, Riemannian geometry of Lagrangian submanifolds, Taiwanese J. Math., 5(2001),
681-720.
[23] B. Y. Chen, More on convolution of Riemannian manifolds, Beitrage Algebra Geom.,
43(2002).
[24] B. Y. Chen, Real hypersurfaces in complex space forms which are warped products, Hokkaido
Math. J., 31(2002).
[25] B. Y. Chen, Constant-ratio space-like submanifolds in pseudo-Euclidean space, Houston J.
Math., 28 (2002).
[26] B. Y. Chen, CR-warped products in complex space forms, (to appear).
[27] B. Y. Chen, Warped products in real space forms, (to appear).
[28] B. Y. Chen, On isometric minimal immersions from warped products into real space forms,
(to appear).
[29] B. Y. Chen, Convolution of Riemannian manifolds and its applications, (to appear).
[30] B. Y. Chen, Geometry of position functions of Riemannian submanifolds in pseudo-Euclidean
space, J. Geometry, (to appear).
[31] B. Y. Chen, When the position vector of a space curve always lies in its rectifying plane?,
Amer. Math. Monthly, (to appear).
[32] B. Y. Chen and W. E. Kuan, The Segre imbedding and its converse, Ann. Fac. Sci.
Toulouse, Math., 7(1981), 1-28.
[33] B. Y. Chen and S. Maeda, Real hypersurfaces in nonflat complex space forms are irreducible,
(to appear).
[34] B. Y. Chen and K. Ogiue, On totally real submanifolds, Trans. Amer. Math. Soc.,
193(1974), 257-266.
[35] B. Y. Chen and K. Ogiue, Two theorems on Kaehler manifolds, Michigan Math. J.,
21(1974), 225-229.
[36] B. Y. Chen and L. Vrancken, Lagrangian submanifolds satisfying a basic equality, Math.
Proc. Camb. Phil. Soc., 120(1996), 291-307.
[37] M. Dajczer and L. A. Florit, On Chen’s basic equality, Illinois J. Math., 42(1998), 97-106.
[38] F. Dillen and S. Nolker Semi-parallelity, multi-rotation surfaces and the helix-property, J.
Reine Angew. Math., 435(1993), 33-63.
[39] N. Ejiri, A generalization of minimal cones, Trans. Amer. Math. Soc., 276(1983), 347-360.
[40] N. Ejiri, Some compact hypersurfaces of constant scalar curvature in a sphere, J. Geom.,
19(1982), 197-199.
[41] M. Gromov, A topological technique for the construction of solutions of differential equations
and inequalities, Intern. Congr. Math., (Nice 1970), 2(1971), 221-225.
[42] M. Meumertzheim, H. Reckziegel and M. Schaaf, Decomposition of twisted and warped
product nets, Results Math., 36 (1999), 297-312.
[43] J. D. Moore, Isometric immersions of Riemannian products, J. Differential Geometry,
5(1971), 159-168.
[44] H. Nakagawa and R. Takagi, On locally symmetric Kaehler submanifolds in complex pro-
jective space, J. Math. Soc. Japan, 28(1976), 638-667.
156 BANG-YEN CHEN
[45] M. Nauenberg, Newton’s early computational method for dynamics, Arch. Hist. Exact Sci.,
46(1994), 221-252.
[46] I. Newton, Principia, Motte’s translation revised, University of California, Berkeley, 1947.
[47] S. Nolker, Isometric immersions of warped products, Differential Geom. Appl., 6(1996),
1-30.
[48] K. Ogiue, Differential geometry of Kaehler submanifold, Adv. in Math., 13(1974), 73-114
[49] B. O’Neill, Semi-Riemannian Geometry with Applications to Relativity, Academic Press,
New York, 1983.
[50] T. Sasahara, On Chen invariant of CR-submanifolds in a complex hyperbolic space, Tsukuba
Math. J., 25(2001).
[51] M. Schaaf, Decomposition of isometric immersions between warped products, Beitrage Al-
gebra Geom., 41(2000), 427-436.
[52] C. Segre, Sulle varieta che rappresentano le coppie di punti di due piani o spazi, Rend. Cir.
Mat. Palermo, 5(1891), 192-204.
[53] D. J. Struik, Differential Geometry, second ed., Addison-Wesley, Reading, Massachusetts,
1961.
[54] B. Suceava, The Chen invariants of warped products of hyperbolic planes and their applica-
tions to immersibility problems, Tsukuba J. Math., 25(2001), 311-320.
Department of Mathematics, Michigan State University, East Lansing, MI 48824-1027, USA.
E-mail: [email protected]