Professional Documents
Culture Documents
r
X
i
v
:
1
1
0
7
.
1
6
7
9
v
2
[
m
a
t
h
.
D
G
]
2
6
A
u
g
2
0
1
1
Umbilical submanifolds of S
n
R.
Bruno Mendonca & Ruy Tojeiro
Abstract
We give a complete classication of umbilical submanifolds of arbitrary dimen-
sion and codimension of S
n
R, extending the classication of umbilical surfaces
in S
2
R by Rabah-Souam and Toubiana as well as the local description of um-
bilical hypersurfaces in S
n
R by Van der Veken and Vrancken. We prove that,
besides small spheres in a slice, up to isometries of the ambient space they come in
a two-parameter family of rotational submanifolds whose substantial codimension
is either one or two and whose prole is a curve in a totally geodesic S
1
R or
S
2
R, respectively, the former case arising in a one-parameter family. All of them
are dieomorphic to a sphere, except for a single element that is dieomorphic
to Euclidean space. We obtain explicit parametrizations of all such submanifolds.
We also study more general classes of submanifolds of S
n
R and H
n
R. In par-
ticular, we give a complete description of all submanifolds in those product spaces
for which the tangent component of a unit vector eld spanning the factor R is
an eigenvector of all shape operators. We show that surfaces with parallel mean
curvature vector in S
n
R and H
n
R having this property are rotational surfaces.
We also prove a Dajczer-type reduction of codimension theorem for submanifolds
of S
n
R and H
n
R.
1 Introduction
Roughly speaking, a submanifold of a Riemannian manifold is totally umbilical, or simply
umbilical, if it is equally curved in all tangent directions. More precisely, an isometric
immersion f : M
m
M
n
between Riemannian manifolds is umbilical if there exists a
normal vector eld along f such that its second fundamental form
f
: TM TM
N
f
M with values in the normal bundle satises
f
(X, Y ) = X, Y for all X, Y TM.
Umbilical submanifolds are the simplest submanifolds after the totally geodesic ones
(for which the second fundamental form vanishes identically), and their knowledge sheds
light on the geometry of the ambient space.
Apart from space forms, however, there are few Riemannian manifolds for which
umbilical submanifolds are classied. Recently, this was accomplished for all three-
dimensional Thurston geometries of non-constant curvature as well as for the Berger
spheres in [14]. The richest case turned out to be that of the product spaces S
2
R and
1
H
2
R. For these manifolds, it was shown that, up to isometries of the ambient space,
umbilical nontotally geodesic surfaces come in a one-parameter family of rotational
surfaces, whose prole curves have been completely determined in terms of solutions of
a certain ODE.
A local description of umbilical hypersurfaces of S
n
R and H
n
R of any dimension n
was given in [16] and [5], respectively. Again, the nontotally geodesic ones are rotational
hypersurfaces over curves in totally geodesic products S
1
R and H
1
R, respectively.
In this paper we give a complete classication of umbilical submanifolds of arbitrary
dimension and codimension of S
n
R. To state our result, for a given integer m 2
let : S
m+1
R R
m+2
0 be the conformal dieomorphism given by (x, t) = e
t
x.
Choose a closed half-line := x [0, ) R
m+2
= R
m+1
R with x ,= 0. Let M
m
r,h
be the image by
1
of the m-dimensional sphere S
m
r,h
in R
m+2
of radius r centered on
that lies in the ane hyperplane through ( x, h) orthogonal to , with the origin removed
if h = 0 and r = d := [ x[. Then we prove:
Theorem 1. The submanifold M
m
r,h
is a complete umbilical submanifold of S
m+1
R
for every r > 0 and h 0. Moreover, it has the following properties:
(i) It is dieomorphic to S
m
if (r, h) ,= (d, 0) and to R
m
if (r, h) = (d, 0);
(ii) It lies in a totally geodesic hypersurface S
m
R S
m+1
R if and only if h = 0;
(iii) M
m
r,0
is homologous to zero in S
m
R if r < d and inhomologous to zero if r > d;
(iv) It is a rotational submanifold whose prole is a curve in a totally geodesic sub-
manifold S
2
R (respectively, S
1
R) if h ,= 0 (respectively, h = 0);
(v) M
m
r,h
is not congruent to M
m
r
,h
if (r, h) ,= (r
, h
).
Conversely, any umbilical nontotally geodesic submanifold of S
n
R with dimension
m 2 is, up to an isometry of the ambient space, an open subset of one of the following:
(i) a small sphere in S
n
0;
(ii) M
m
r,0
for some r > 0 if n = m;
(iii) M
m
r,h
for some r > 0 and h 0 if n = m + 1;
(iv) M
m
r,h
in a totally geodesic S
m+1
R for some r > 0 and h 0 if n > m + 1.
Moreover, we provide an explicit parametrization of all submanifolds M
m
r,h
, r > 0,
h 0 (see Proposition 14 below) in terms of elementary functions. The precise meaning
of M
m
r,h
being rotational is explained in Section 4.
In the process of proving Theorem 1, we have been led to study more general classes
of submanifolds with interest on their own.
2
Let Q
n
denote either S
n
, R
n
or H
n
, according as = 1, = 0 or = 1, respectively.
Given an isometric immersion f : M
m
Q
n
R, let
t
be a unit vector eld tangent
to the second factor. Thus, for = 0 we just choose a unit constant vector eld
t
in
R
n+1
. Then, a tangent vector eld T on M
m
and a normal vector eld along f are
dened by
t
= f
T + . (1)
We denote by / the class of isometric immersions f : M
m
Q
n
R with the
property that T is an eigenvector of all shape operators of f. Our next result is a
complete description of all isometric immersions in class /. First note that trivial
examples are products N
m1
R, where N
m1
is a submanifold of Q
n
, which correspond
to the case in which the normal vector eld in (1) vanishes identically. We call
these examples vertical cylinders. More interesting ones are constructed as follows. We
consider the case 1, 1, the case = 0 being similar.
Let g : N
m1
Q
n
Q
n
R and i : Q
n
R E
n+2
denote the canonical inclusions, and let
k = i j. Here E
n+2
denotes either Euclidean space R
n+2
if = 1 or Lorentzian space
L
n+2
if = 1. Set
i
= k
i
, 1 i k,
0
= g := k g and
k+1
= i
0
, . . . ,
k+1
is
also parallel and at, and we may dene a vector bundle isometry : N
m1
E
k+2
E
by
x
(y) := (x, y) =
k+1
i=0
y
i
i
, for y = (y
0
, . . . , y
k+1
) E
k+2
.
Now let
f : M
m
:= N
m1
I Q
n
R
be given by
f(x, s) := (i f)(x, s) =
x
((s)) =
k+1
i=0
i
(s)
i
(x), (2)
where : I Q
k
R E
k+2
, = (
0
, . . . ,
k
,
k+1
), is a smooth regular curve such
that
2
0
+
2
1
+. . . +
2
k
= and
k+1
has nonvanishing derivative. Notice that vertical
cylinders correspond to the case in which the curve is the generator of Q
k
R through
(1, 0, , 0) Q
k
.
Theorem 2. The map f denes, at regular points, an immersion in class /. Con-
versely, any isometric immersion f : M
m
Q
n
R,
1, 1, that are included in class / are hypersurfaces with constant sectional
curvature [13], rotational hypersurfaces [10] and constant angle hypersurfaces (see, e.g.,
[15]; see also Corollary 4 below and the comments before it).
Let f : M
m
Q
n
R E
n+2
is the canonical inclusion. It was shown in [13] that if m = n
then f is in class / if and only if the vector eld T in (1) is nowhere vanishing and
f
has at normal bundle. For submanifolds of higher codimension we have the following.
Corollary 3. The following assertions are equivalent:
(i) The vector eld T in (1) is nowhere vanishing and
f has at normal bundle;
(ii) f has at normal bundle and is in class /;
(iii)
f is locally given as in (2) in terms of an isometric immersion g : N
m1
Q
n
R E
k+2
,
= (
0
, . . . ,
k
,
k+1
), with
k+1
nowhere vanishing.
Observe that the vector eld T vanishes at some point if and only if f(M
m
) is tangent
to the slice Q
n
t of Q
n
R E
k+2
, = (
0
, . . . ,
k
,
k+1
),
with
k+1
nowhere vanishing. Clearly, in this case the existence of an orthonormal set
of parallel normal vector elds
1
, . . . ,
k
along g is automatic for any 1 k n 1.
In the case of a hypersurface f : M
n
Q
n
R a
geodesic of Q
k
R
with curves in totally geodesic submanifolds Q
R Q
n
R Q
n
R;
(ii) f is given as in (2) in terms of an umbilical isometric immersion g : N
m1
Q
n
f
(X, Y ) = X, Y for all X TM and Y T
, (3)
where T is the vector eld dened by (1), and is parallel in the normal connection
along T
if m = 2.
Moreover, if = 1 then the preceding assertions are equivalent to f being given as in
(2) in terms of a totally geodesic isometric immersion g : N
m1
Q
n
. This is also
the case if = 1 and f is assumed to be of hyperbolic type in (i) and g an equidistant
hypersurface in (ii).
Notice that in the hypersurface case, i.e., for n = m, the second fundamental form
satises (3) if and only if f has at most two principal curvatures, and if it has exactly
two then one of them is simple with T as an eigenvector.
A key step in the classication of umbilical submanifolds in S
n
R is the following re-
sult on reduction of codimension of isometric immersions into Q
n
R. That an isometric
immersion f : M
m
Q
n
R of
Q
n
R. We denote by
N
1
L. Then f reduces codimension to .
5
Lemma 6 should be compared with its well-known counterpart for submanifolds of
space forms (see, e.g., [11]), in which case the corresponding condition for a submanifold
to reduce codimension is that its rst normal spaces form a parallel subbundle of the
normal bundle. A necessary and sucient condition for parallelism of the rst normal
bundle of a submanifold of a space form in terms of its normal curvature tensor R
and
mean curvature vector eld H was obtained by Dajczer [8] (see also Chapter 4 of [9]).
The proof of Dajczers theorem can be easily adapted to yield the following result for
submanifolds of Q
n
R.
Theorem 7. Let f : M
m
Q
n
N
1
L if and only if the following two conditions
hold:
(i)
[
L
= 0;
(ii)
H L.
As an application of Theorem 7, in Subsection 51 we give a simple proof of The-
orem 1 in [3] on surfaces with parallel mean curvature vector in Q
n
R. By using this
result together with Corollary 5 we prove the following.
Corollary 8. Any surface f : M
2
Q
n
R,
m 4, over a curve in a totally geodesic submanifold Q
s
R, s 3.
In the case n = 2, the preceding corollary is a special case of Theorem 3 in [15], in
which hypersurfaces f : M
n
Q
n
2
(
u
+i
v
) and the metric on M
2
is extended to a C-bilinear
map.
6
The same authors observed in [3] that surfaces with parallel mean curvature vector
that are not contained in a slice of Q
n
R and either
(i) M
2
is homeomorphic to a sphere, or
(ii) = 1, M
2
is complete with K 0 and K is not identically zero.
Then f is a rotational surface (of spherical type in case (i)) in a totally geodesic sub-
manifold Q
m
R, s 3.
We observe that, if f(M
2
) is contained in a slice Q
n
t of Q
n
R, then by
Theorem 4 in [17] either f is a minimal surface of an umbilical hypersurface of Q
n
t
or it is a surface with constant mean curvature in a three-dimensional umbilical or totally
geodesic submanifold of Q
n
t. Moreover, if M
2
is homeomorphic to a sphere, then
by Hopfs Theorem it must be a totally umbilical 2-sphere of Q
n
t.
The paper is organized as follows. In the next section we recall the basic equations
of an isometric immersion into Q
n
R. Then
we apply the latter to give a simple proof of Theorem 1 in [3] on surfaces with parallel
mean curvature vector in Q
n
Q
n
R.
Using that
t
is a parallel vector eld in Q
n
X
T = A
f
X (4)
and
f
(X, T) =
X
, (5)
7
for all X TM. Here and in the sequel A
f
X, Y =
f
(X, Y ), for all X, Y TM.
Notice that the vector eld T is a gradient vector eld. Namely, if 1, 1 and
f = if, where i : Q
n
R E
n+2
denotes the canonical inclusion, then T is the gradient
of the height function h =
f, i
t
. If = 0 then T is the gradient of h = f,
t
.
The Gauss, Codazzi and Ricci equations for f are, respectively (see, e.g., [12]),
R(X, Y )W = (X Y Y, TX T +X, TY T)W + A
f
(Y,W)
X A
f
(X,W)
Y, (6)
_
_
(Y, W)
_
_
(X, W) = (X, WY, T Y, WX, T) (7)
and
R
(X, Y ) = (X, A
f
Y ) (A
f
X, Y ). (8)
Equation (7) can also be written as
(
Y
A
f
)(X, ) (
X
A
f
)(Y, ) = , (X Y )T, (9)
where (X Y )T = Y, TX X, TY .
Although this will not be used in the sequel, it is worth mentioning that equations (4)
(8) completely determine an isometric immersion f : M
m
Q
n
R up to isometries
of Q
n
R E
n+2
,
1, 1, where : E
n+1
R E
n+1
is the projection, and
Z
=
Z = i
Z i
Z, i
t
i
t
= i
(Z Z,
t
t
),
for every Z T(Q
n
R), where
is the derivative in E
n+2
. Hence
A
i
Z = Z +Z,
t
t
. (10)
The normal spaces of f and
f are related by
N
f
M = i
N
f
M span,
where = f =
f. Let
denote the Levi-Civita connection of Q
n
R. Given
N
f
M, we obtain from (10) that
X
i
= i
X
+
i
(f
X, )
=
A
f
X + i
X
+X, T, ,
8
hence
A
f
i
= A
f
and
X
i
= i
X
+X, T, (11)
for every N
f
M, where
X
=
X
f =
fX
=
f
(X X, TT) X, Ti
,
hence
A
X = X + X, TT,
or equivalently,
A
T = ||
2
T and A
X = X for X T
, (12)
and
X
= X, Ti
. (13)
3 Class /
In this section we study submanifolds in class /. In particular, we give the proofs of
Theorem 2 and of Corollaries 3 and 4. We start with the following observation.
Proposition 10. Assume that the vector eld T in (1) is nowhere vanishing. Then the
following assertions are equivalent:
(i) T is an eigenvector of A
f
for all N
f
M;
(ii) is parallel along T
;
(iii) A
commutes with A
f
for all N
f
M.
Proof: The equivalence between (i) and (ii) follows from (5), whereas (12) implies the
equivalence between (i) and (iii).
Before going into the proof of Theorem 2, we write down in the next proposition the
dierential, the normal space and the second fundamental form of an immersion
f = i f : M
m
:= N
m1
I Q
n
R E
n+2
, 1, 1,
which is given by (2) in terms of an isometric immersion g : N
m1
Q
n
and a smooth
regular curve : I Q
k
R E
k+2
, = (
0
, . . . ,
k
,
k+1
), with
2
0
+
2
1
+. . .+
2
k
= .
The case = 0 is similar. We use the notations before the statement of Theorem 2.
Given x N
m1
, X T
x
N and s I, we denote by X
H
the unique vector in T
(x,s)
M
such that
1
X
H
= X and
2
X
H
= 0, where
1
: M
m
N
m1
and
2
: M
m
I are
the canonical projections.
9
Proposition 11. The following holds:
(i) The dierential of
f is given by
(x, s)X
H
= g
(x)(
0
(s)I
k
i=1
i
(s)A
g
i
(x))X, for every X T
x
N, (14)
where I is the identity endomorphism of T
x
N, and
(x, s)
s
=
x
(
(s)). (15)
(ii) The map
f (and hence f) is an immersion at (x, s) if and only if
P
s
(x) :=
0
(s)I
k
i=1
i
(s)A
g
i
(x) = A
g
x( (s))
,
where (s) = (
0
(s), . . . ,
k
(s), 0), is an invertible endomorphism of T
x
N.
(iii) If
f is an immersion at (x, s) then
N
f
(x,s)
M = k
E(x)
x
(
(s)
) N
g
x
N,
where E(x)
N
f
(x,s)
M span(
f)(x, s) = i
N
f
(x,s)
M
x
( (s)). (16)
(iv) If
f is an immersion at (x, s) then
A
(x, s)X
H
= (P
s
(x)
1
A
g
(x)X)
H
(17)
for all N
f
(x,s)
M and X T
x
N,
A
(x, s)
s
= 0, if k
E(x)
, (18)
and
A
f
x()
(x, s)
s
=
(s),
(s),
(s)
s
, if E
k+2
, ,
(s) = 0. (19)
Moreover,
A
f
(x, s) = A
f
i
(x, s) (20)
for every N
f
(x,s)
M.
10
Proof: Given a smooth curve : J N
m1
with 0 J, (0) = x and
(0) = X, for
each s I let
s
: J M
m
be given by
s
(t) = ((t), s). Then
s
(0) = (x, s) and
s
(0) = X
H
. Hence
(x, s)X
H
=
d
dt
[
t=0
f(
s
(t)) =
d
dt
[
t=0
k+1
i=0
i
(s)
i
((t))
= g
(x)(
0
(s)I
k
i=1
i
(s)A
g
i
(x))X,
and (14) follows from the fact that A
g
i
= A
g
i
for any 1 i k.
The proof of (15) is straightforward, and the assertions in (ii) and (iii) follow im-
mediately from (i). To prove (17), given N
f
(x,s)
M and X T
x
N, let : J N
m1
and
s
: J M
m
be as in the beginning of the proof. Then, using (14) we obtain
(x, s)A
(x, s)X
H
= (
X
H)
T
= (
d
dt
[
t=0
(
s
(t)))
T
= g
(x)A
g
(x)X
= g
(x)P
s
(x)P
s
(x)
1
A
g
(x)X =
(x, s)(P
s
(x)
1
A
g
(x)X)
H
,
and (17) follows. Here, putting T as a superscript of a vector means taking its tangent
component.
Formula (18) is clear. As for (19), given E
k+2
with ,
(s) = 0, extend to a
parallel normal vector eld along , so that
(s) =
(s),
(s)
(s),
(s)
(s) =
(s), (s)
(s),
(s)
(s).
Then we have
(x, s)A
f
x()
(x, s)
s
=
s
x
() =
x
(
(s)) =
(s), (s)
(s),
(s)
x
(
(s))
=
(x, s)
(s), (s)
(s),
(s)
s
,
where we have used (15) in the last equality. This gives (19) and completes the proof,
for (20) is clear.
Proof of Theorem 2: It follows from (14) and (15) that X
H
,
s
= 0 for any X TN,
with respect to the metric induced by f. On the other hand, we also have from (14)
that X
H
, T =
X
H
, i
t
= 0 for any X TN. Hence T is in the direction of /s.
We have
T, /s =
T,
f
/s = i
/t,
x
(
(s)) =
k+1
(s),
hence
T =
T, /s
/s, /s
s
=
k+1
(s)
|
(s)|
2
s
.
11
In particular, T is nowhere vanishing by the assumption that
k+1
(s) ,= 0 for all s I.
That f belongs to class / now follows from (18), (19) and (20).
Let us prove the converse. Since f : M
m
Q
n
T
T, X =
X
T, T = A
f
T, X = 0 for any X T
. (21)
Hence, the one-dimensional distribution spanned by T is totally geodesic. Moreover,
since T is a gradient, then the orthogonal distribution T
is integrable. Therefore,
there exists locally a dieomorphism : N
m1
I M
m
, where I is an open interval
containing 0, such that (x, ): I M
n
are integral curves of T for any x N
m1
and
(, s): N
m1
M
m
are leaves of T
X, i
t
=
X, T = 0 (22)
for every X E
1
. Moreover,
f
(X,
s
) = 0 for every X E
1
, in view of (12) and the
fact that f belongs to class /. Hence, using that E
2
is totally geodesic we obtain that
X =
f
s
X +
f
_
X,
s
_
=
f
s
X
f
E
1
for all X E
1
, hence
f
E
1
is constant in E
n+2
along the leaves of E
2
. In view of (22)
we can assume that g :=
f(, 0) satises g(N
m1
) Q
n
0. Set
f =
f and
h =
f, i
/t, so that
f =
f + hi
t
.
Using (22) we obtain
X,
f =
X,
f = 0
for all X E
1
, since
f,
f = . Therefore, we have
f(x, s) (
f
(x, s)E
1
(x, s))
= (
f
(x, 0)E
1
(x, 0))
= ( g
(x)T
x
N)
,
where in the rst equality we have used that
f
E
1
is constant in E
n+2
along E
2
. Hence,
for xed s I, we have that
s
(x) :=
f(x, s) denes a normal vector eld along g.
Moreover,
s
f
(x, s)E
1
(x, s) = g
(x)T
x
N,
thus
s
is parallel along g in the normal connection. It follows that
x N span
f(x, s) : s I
12
is a parallel at subbundle of N
g
N and, for xed x N, the ber
f(x, s) : s I is
contained in a cylinder Q
I N
g
x
N, for
f(x, s),
f(x, s) = .
Let g : N
m1
Q
n
i
= i
i
, 1 i k,
0
= g and
k+1
= i
/t span
f,
i
=
X,
i
+
f,
i
= 0,
for
f is a normal vector eld,
f
(x, s)X g
(x)T
x
N and
i
is parallel in the normal
connection of g. Then we can write
f(x, s) =
k+1
i=0
i
, with
i
=
i
(s).
Moreover, from
f,
f = we obtain that
2
0
+
k
i=1
2
i
= .
Proof of Corollary 3: It follows from the Ricci equation (8) that f has at normal bundle
if and only if all shape operators A
f
, N
f
M, are simultaneously diagonalizable,
whereas
f has at normal bundle if and only if this holds for all shape operators A
,
N
f
M. Since the vector eld T in (1) is nowhere vanishing, the equivalence between (i)
and (ii) then follows from Proposition 10 and the fact that A
f
i
= A
f
for all N
f
M.
Let f : M
m
Q
n
. Since T is
an eigenvector of all shape operators of f, it follows from (17) and (20) that all shape
operators of f commute if and only if the same holds for the shape operators of g. By
the Ricci equation, we conclude that f has at normal bundle if and only if the same
holds for g. Hence (ii) and (iii) are equivalent.
Proof of Corollary 4: Let f : M
m
:= N
m1
I Q
n
R a geodesic of Q
k
f
(T, T) = 0 by (19). Thus is parallel in the normal connection of f by (5).
Conversely, let f : M
m
:= N
m1
I Q
n
R
In this section we dene rotational submanifolds in Q
n
= (x
0
, . . . , x
n
) E
n+1
: x
2
0
+ x
2
2
+ . . . + x
2
n
= ( x
0
> 0 if = 1).
Let P
nm+3
be a subspace of E
n+2
of dimension nm+3 containing the e
0
and the
e
n+1
directions, where e
0
, . . . , e
n+1
is the canonical basis. Then
(Q
n
R) P
nm+3
= Q
nm+1
R.
Denote by T the group of isometries of E
n+2
that x pointwise a subspace P
nm+2
P
nm+3
also containing the e
n+1
direction. Consider a curve in Q
nm+1
R P
nm+3
that lies in one of the two half-spaces of P
nm+3
determined by P
nm+2
.
Denition 12 An m-dimensional rotational submanifold in Q
n
i=m
im+1
(s)e
i
+ h(s)e
n+1
,
with
nm+1
i=0
2
i
= 1, the rotational submanifold in S
n
R with prole curve and axis
P
nm+2
can be parametrized by
f(s, t) = (
0
(s)
1
(t), . . . ,
0
(s)
m
(t),
1
(s), . . . ,
nm+1
(s), h(s)), (23)
where t = (t
1
, . . . , t
m1
) and = (
1
, . . . ,
m
) parametrizes S
m1
R
m
.
For = 1, one has three distinct possibilities, according as P
nm+2
is Lorentzian,
Riemannian or degenerate, respectively, and the rotational submanifold is called accord-
ingly of spherical , hyperbolic or parabolic type. In the rst case, we can assume that
P
nm+2
is spanned by e
0
, e
m+1
, . . . , e
n+1
and that
(s) =
0
(s)e
0
+
n
i=m
im+1
(s)e
i
+ h(s)e
n+1
, (24)
14
with
2
0
(s) +
nm+1
i=1
2
i
= 1. Then, the submanifold can be parametrized by
f(s, t) = (
0
(s),
1
(s)
1
(t), . . . ,
1
(s)
m
(t),
2
(s), . . . ,
nm+1
(s), h(s)),
where again t = (t
1
, . . . , t
m1
) and = (
1
, . . . ,
m
) parametrizes S
m1
R
m
.
In the second case, we can assume that P
nm+2
is spanned by e
m
, . . . , e
n+1
. Then,
with the curve also given as in (24), a parametrization is
f(s, t) = (
0
(s)
1
(t), . . . ,
0
(s)
m
(t),
1
(s), . . . ,
nm+1
(s), h(s)),
where t = (t
1
, . . . , t
m1
) and = (
1
, . . . ,
m
) parametrizes H
m1
L
m
.
Finally, when P
nm+2
is degenerate, we choose a pseudo-orthonormal basis
e
0
=
1
2
(e
0
+ e
n
), e
n
=
1
2
(e
0
+ e
n
), e
j
= e
j
,
for j 1, . . . , n1, n+1, and assume that P
nm+2
is spanned by e
m
, . . . , e
n+1
. Notice
that e
0
, e
0
= 0 = e
n
, e
n
and e
0
, e
n
= 1. Then, we can parametrize by
(s) =
0
(s) e
0
+
n
i=m
im+1
(s) e
i
+ h(s) e
n+1
,
with 2
0
(s)
nm+1
(s) +
nm
i=1
2
i
(s) = 1, and a parametrization of the rotational
submanifold is
f(s, t) = (
0
,
0
t
1
, . . . ,
0
t
m1
,
1
, . . . ,
nm
,
nm+1
0
2
m1
i=1
t
2
i
, h), (25)
where t = (t
1
, . . . , t
m1
) parametrizes R
m1
,
i
=
i
(s), 0 i nm+1, and h = h(s).
Proof of Corollary 5: We can write (23) as
f(s, t) =
0
(s) g(t) +
n
i=m
im+1
(s)e
i
+ h(s)e
n+1
,
where g(t) =
m
i=1
i
(t)e
i
for t = (t
1
, . . . , t
m1
). This shows that for = 1 a rota-
tional submanifold is given as in (2) in terms of a totally geodesic isometric immersion
g : S
m1
S
n
. The case of a rotational submanifold of hyperbolic type in H
n
R is
similar. In particular, this proves that (i) implies (ii) in these cases.
Equation (25) can be written as
f(s, t) =
0
g +
n
i=m
im+1
(s) e
i
+ h(s) e
n+1
, (26)
15
where g(t) = e
0
+
m1
i=1
t
i
e
i
1
2
(
m1
i=1
t
2
i
) e
n
. Notice that g denes an isometric im-
mersion of R
m1
into L
n+2
(in fact into the light-cone V
n+1
, for g, g = 0), and that
g, e
m
, . . . , e
n
, e
n+1
is a pseudo-othonormal basis of N
g
R
m1
, with g, g = 0 = e
n
, e
n
,
g, e
n
= 1 and e
m
, . . . , e
n1
, e
n+1
an orthonormal basis of span g, e
n
n
i=m
im+1
(s
0
) e
i
. Then g denes
an umbilical isometric immersion with the same normal space in L
n+2
as g at every
t R
m1
, i.e.,
span g, e
m
, . . . , e
n
, e
n+1
= spang,
1
, . . . ,
nm+1
, e
n+1
,
where
i
= i
i
, 1 i n m + 1, for a parallel orthonormal frame
1
, . . . ,
nm+1
of
N
g
R
m1
. Hence we can also write (26) as
f(s, t) =
0
g +
nm+1
i=1
i
(s)
i
+ h(s) e
n+1
,
where : I E
nm+3
is a regular curve satisfying
2
0
+
nm+1
i=1
2
i
= 1. Thus
condition (ii) holds for f. The case of a spherical rotational submanifold is similar and
easier.
Now suppose that f is given as in (2) in terms of an umbilical isometric immersion
g : N
m1
Q
n
1
, . . . ,
nm+1
, e
n+1
,
where
i
= i
i
, 1 i n m + 1, for a parallel orthonormal frame
1
, . . . ,
nm+1
of
N
g
S
m1
. Hence f can also be parametrized by
f(s, t) =
0
g +
n
i=m
im+1
(s) e
i
+ h(s) e
n+1
,
where : I E
nm+3
is a smooth regular curve satisfying
nm+1
i=0
2
i
= 1. Thus f is
a rotational submanifold with as prole.
If = 1, we argue for the parabolic case, the others being similar and easier. We
can assume that N
m1
= R
m1
and that the the ane hull of g(R
m1
) in L
n+1
is v +W,
where W is the subspace spanned by e
1
, . . . , e
m1
, e
n
and v W
. Then g = a g +v,
where a R and g(t) = e
0
+
m1
i=1
t
i
e
i
1
2
(
m1
i=1
t
2
i
) e
n
for t = (t
1
, . . . , t
m1
). As before,
by using the fact that g and g have the same normal spaces in L
n+1
for every t R
m1
,
we conclude that f is a rotational submanifold parametrized as in (26).
16
The second fundamental form of f being given by (3) is equivalent to the restriction
of each shape operator A
f
to T
to T
is a multiple of the
identity tensor, hence the second fundamental form of f is as in (3). To conclude the
proof that (ii) and (iii) are equivalent, it remains to show that for m = 2 the additional
assumption that the vector eld in (3) be parallel along T
is equivalent to the
unit-speed curve g : J := N
m1
Q
n
f
(X, X) = i
f
(X, X) ,
hence we obtain using (11) and (13) that
f
(X, X) =
f
(X,X)
X +
f
(X, X)
=
f
(X,X)
X + i
f
(X, X).
Thus,
f
(X, X) = 0 if and only if
f
(X, X) =
f
(X,X)
X. (27)
It follows from (17) that at the point (t, s) we have
f
(X, X) =
g
(t)
g
(t),
t
( (s))
, (28)
where g = k g. From (14) we obtain
X =
1
g
(t),
t
( (s))
d
dt
H
, (29)
where
d
dt
is a unit vector eld along J. Hence
f
(X, X) =
g
(t),
t
( (s)
g
(t),
t
( (s))
3
g
(t) +
1
g
(t),
t
( (s))
2
g
(t).
On the other hand, equations (14), (28) and (29) yield
f
(X,X)
X =
g
(t), g
(t)
g
(t),
t
( (s))
2
g
(t).
It follows easily that (27) holds if and only if
g
(t) = g
(t), g
(t) g
(t),
which is equivalent to g being a geodesic circle.
17
5 Reduction of codimension
In this section we prove Lemma 6 and Theorem 7 stated in the introduction.
Proof of Lemma 6: We have from (5) that
X
N
1
L for every X TM. Since
N
1
L by assumption, it follows that L is a parallel subbundle of N
f
M. Let
f = i f, where i : Q
n
R E
n+2
is the inclusion. Given L
= N
, from
(11) and the fact that L is a parallel subbundle of N
f
M we obtain
X
i
= i
X
i
,
hence i
is a parallel subbundle of N
f
M.
Since i
1
, where
N
1
(x) is the rst normal space of
f at x M
m
, it follows
that i
is a constant subspace of E
n+2
, which is orthogonal to
t
. Denote by K the
orthogonal complement of i
in E
n+2
. Then, for any xed x
0
M
m
we have
f(M
m
)
f(x
0
) + K.
But since K contains
t
and (x
0
), it also contains the position vector
f(x
0
). Thus
f(x
0
) + K = K. We conclude that
f(M) (Q
n
R) K = Q
m+1
R.
Proof of Theorem 7: Assume that
N
1
L. Then condition (ii) is trivially satised.
To prove (i), rst notice that for N
1
the Ricci equation gives
R
(X, Y ) = (X, A
Y ) (A
X, Y ) = 0.
Given L
, we have that N
1
and that
Z
N
1
by our assumption, hence
(
Z
R
)(X, Y, ) =
Z
R
(X, Y )R
(
Z
X, Y )R
(X,
Z
Y )R
(X, Y )
Z
= 0.
To prove the converse, let L
. Since R
(X, Y )
Z
= 0
for all X, Y, Z TM. Using the Ricci equation again, we obtain that
[A
, A
] = 0
for all Z, W TM. Hence, at any x M there exists an orthonormal basis Z
1
, . . . , Z
n
of T
x
M that diagonalizes simultaneously all shape operators A
, Z TM. We will
show that
Z
k
, (Z
i
, Z
j
) = 0
for all 1 i, j, k n, which implies that
X
N
1
for all X TM.
From the choice of the basis Z
1
, . . . , Z
n
, we have
(Z
i
, Z
j
),
Z
k
= A
Z
k
Z
i
, Z
j
= 0
18
if i ,= j. It follows from the Codazzi equation (9) and the fact that L
that
A
Z
i
Z
k
= A
Z
k
Z
i
,
hence the eigenvalue
ki
of A
Z
k
correspondent to X
i
vanishes unless k = i. Therefore,
(Z
i
, Z
i
),
Z
k
= A
Z
k
Z
i
, Z
i
= 0, if i ,= k.
Finally, the assumption
Z
i
= nH,
Z
i
= 0.
5.1 Alencardo CarmoTribuzzi Theorem
In this subsection we apply Theorem 7 to give a simple proof of the following theorem
due to Alencar, do Carmo and Tribuzzi [3].
Theorem 13. Let f : M
2
Q
n
t.
(ii) f is a surface with constant mean curvature in a three-dimensional umbilical or
totally geodesic submanifold of a slice Q
n
t.
(iii) f(M
2
) lies in a totally geodesic submanifold Q
m
R, m 4, of Q
n
R.
Proof: Since the mean curvature vector H is parallel and nonzero, the function :=
|H|
2
on M
2
is a nonzero constant. Suppose rst that A
H
= I everywhere on M
2
. We
claim that the vector eld T vanishes identically. Assuming otherwise, there exists an
open subset U where T ,= 0. Choose a unit vector eld X on U orthogonal to T. Then
H, (X, T) = X, T = 0. (30)
By the Codazzi equation (7) we have
(
T
)(X, X) (
X
)(T, X), H = |T|
2
, H.
It follows easily from (30) and the fact that is constant on M
2
that the left-hand-side
of the preceding equation is zero. Thus , H vanishes on U, and hence
0 = T, H =
T
, H = (T, T), H = |T|
2
,
where we have used (5) in the third equality. This is a contradiction and proves the
claim.
19
Therefore, if A
H
= I everywhere on M
2
then f(M
2
) is contained in a slice Q
n
t
of Q
n
] = 0 for any x M
2
and every normal
vector N
x
M. Then, the fact that A
H
has distinct eigenvalues on V implies that the
eigenvectors of A
H
are also eigenvectors of A
for any N
x
M, x V . Hence all shape
operators are simultaneously diagonalizable at any x V , which implies that f has at
normal bundle on V by the Ricci equation (8). In particular, the rst normal spaces N
1
of f have dimension at most two at any x V . Let W V be an open subset where
L = dimN
1
+ span has constant dimension 3. It follows from Theorem 7 that
f(W) lies in a totally geodesic submanifold Q
2+1
R of Q
n
R. By analyticity of f
(see Remark 1 of [3]), we conclude that f(M
2
) Q
2+1
R.
Proof of Corollary 8: Let X be a unit vector eld orthogonal to T. By Corollary 5, in
order to prove that f is a rotational surface it suces to show that
f
(X, X) = 0.
We follow essentially the proof of Proposition 2 in [3]. Since the mean curvature vector
eld
H =
1
2
(
f
(X, X) +|T|
2
f
(T, T))
is parallel in the normal connection, we have
f
(X, X) =
X
(|T|
2
f
(T, T)) = X(|T|
2
)
f
(T, T) |T|
2
f
(T, T).
Now, since f is in class /, we have from (21) that
T
T, X = 0 =
X
T, T.
In particular, X(|T|
2
) = 0. Moreover, using the Codazzi equation (7) we obtain
f
(T, T) = (
f
)(T, T) + 2(
X
T, T) = (
f
)(T, T) = (
f
)(X, T)
=
f
(X, T)
f
(
T
X, T)
f
(X,
T
T) = 0.
That f(M
2
) is contained in a totally geodesic submanifold Q
m
R, m 4, and hence
that its prole curve lies in a totally geodesic submanifold Q
s
R, s 3, follows from
Theorem 13.
6 Umbilical submanifolds of S
n
R
We are now in a position to prove Theorem 1 in the introduction.
Proof of Theorem 1: Since : S
m+1
R R
m+2
0 given by (x, t) = e
t
x is a
conformal dieomorphism, it follows that M
m
r,h
=
1
(S
m
r,h
) is an umbilical submanifold
of S
m+1
R, for a conformal dieomorphism preserves umbilical submanifolds. Assertion
20
(i) and completeness of M
m
r,h
are clear, for M
m
r,h
=
1
(S
m
r,h
) if (r, h) ,= (d, 0) and M
d,0
=
1
(S
m
d,0
0). It is easily seen that the totally geodesic hypersurfaces S
m
R of S
m+1
R
are the images by
1
of the hyperplanes through the origin in R
m+2
. Since S
m
r,h
lies
in such a hyperplane if and only if h = 0, the assertion in (ii) follows. Assertion (iii)
follows from the fact that S
m
r,0
is homologous to zero in R
m+1
if r < d and inhomologous
to zero in R
m+1
if r > d.
We now prove (iv). Since orthogonal transformations of R
m+2
correspond under
the dieomorphism to isometries of S
m+1
R xing pointwise the factor R, and
homotheties of R
m+2
correspond to translations along R, we can assume that x =
2
2
(0, . . . , 0, 1) R
m+1
. Let
T = (p, q) R : (p 1)
2
q < p
2
.
For each (p, q) T, set J
p,q
= (
_
p
q,
_
p
q) and dene h
p,q
:
J
p,q
R by
h
p,q
(s) =
_
p s
2
+
_
(p s
2
)
2
q .
Let Y
p,q
: S
m1
J
p,q
S
m+1
R and Z
p,q
: S
m1
J
p,q
S
m+1
R be given by
Y
p,q
(x, s) =
_
sx,
2
2
_
h
p,q
(s) +
1 p
h
p,q
(s)
_
,
_
q (p 1)
2
2 h
p,q
(s)
, log h
p,q
(s)
_
and, for q ,= 0,
Z
p,q
(x, s) =
_
sx,
2
2
_
1 p
q
h
p,q
(s) +
q
h
p,q
(s)
_
,
_
q (p 1)
2
2q
h
p,q
(s), log
q
h
p,q
(s)
_
.
Notice that Z
p,q
= Y
p,q
, where : S
m+1
R S
m+1
R is the isometry dened by
(x, s) = (Ax, s + log
q
_
1 p
_
q (1 p)
2
_
q (1 p)
2
(1 p)
_
.
Let : T (0, ) [0, ) be the dieomorphism given by
(p, q) =
2
2
_
_
p
2
q,
_
q (p 1)
2
_
.
Then (iv) is a consequence of the following fact.
Lemma 14. For (r, h) = (p, q) we have
M
m
r,h
=
_
Y
p,q
(S
m1
J
p,q
) Z
p,q
(S
m1
J
p,q
), if (r, h) ,= (d, 0),
Y
1,0
(S
m1
(1, 1)), if (r, h) = (d, 0).
21
Proof: We argue for (r, h) ,= (d, 0), the case (r, h) = (d, 0) ( i.e., (p, q) = (1, 0)) being
similar and easier. A straightforward computation shows that
( Y
p,q
)(x, s) =
_
sh
p,q
(s)x,
2
2
(h
2
p,q
(s) p), 0
_
+ ( x, h)
and
( Z
p,q
)(x, s) =
_
s
h
p,q
(s)x,
2
2
(
h
2
p,q
(s) p), 0
_
+ ( x, h),
where
h
p,q
(s) =
q/h
p,q
(s). Let :
J
p,q
R
2
and :
J
p,q
R
2
be given by
(s) = (sh
p,q
(s),
2
2
(h
2
p,q
(s) p)) and (s) = (s
h
p,q
(s),
2
2
(
h
2
p,q
(s) p)),
respectively. Then, the statement follows from the fact that (
J
p,q
) (
J
p,q
) is the circle
of radius r =
_
p
2
q
2
centered at the origin.
We now prove the converse. Let f : M
m
S
n
R, m 2, be an umbilical isometric
immersion. If the vector eld T in (1) vanishes identically, then f(M
n
) is contained in
a slice S
n
R, and this gives the rst possibility in the statement. Now assume that
T does not vanish at some point, and hence on some open subset U M
n
. It suces
to prove that there exist open subsets
U U and V S
m
, (p, q) T and an interval
I I
p,q
such that, up to an isometry of S
n
R, we have f(
U) Y
p,q
(V I). For this
implies that f(
U) M
m
r,h
with (r, h) = (p, q), and thus ( f)(
U) (M
m
r,h
) = S
m
r,h
.
Since f is an umbilical immersion into R
n+2
0, it follows that (f)(M
m
) S
m
r,h
,
and hence f(M
m
) M
m
r,h
.
From Codazzi equation (7) and
f
(X, Y ) = nX, Y H for all X, Y TM, where H
is the mean curvature vector of f, we obtain
n
X
H = X, T (31)
for every X TM. If H and are linearly dependent on U, it follows from Lemma 6
that f has substantial codimension one on U. Otherwise, there exists an open subset
U
is a rotational submanifold of substantial codimension at most two
over a curve in a totally geodesic submanifold S
s
R, s 2. Hence, we can assume
that n = m + 1 and s = 2. Equivalently, in view of the last assertion in Corollary 5,
we obtain that f[
U
is given by (2) in terms of a totally geodesic isometric immersion
22
g : V S
m1
S
m
and a regular curve : I S
2
R R
4
, = (
0
,
1
,
2
,
3
), with
2
0
+
2
1
+
2
2
= 1.
With notations as in Proposition 11, we have by (16) and the umbilicity of f that
A
f
x()
is a multiple of the identity tensor for every
(s)
(s)
. Using that
P
s
(x) =
0
(s)I, it follows from (17), (19) and (20) that
g,
x
()
0
(s)
=
(s),
(s),
(s)
for all
(s)
(s)
,
or equivalently,
0
(s)
(s) + (s)e
0
, = 0 for all
(s)
(s)
,
since g =
x
(e
0
). Here (s) =
(s),
+ e
0
= y
+ z . (32)
We write the preceding equation as
+
0
3
e
3
+ e
0
= y
+ y
3
e
3
+ z . (33)
Notice that
, = 1,
, = 0,
= (
3
)
2
=
, , and
=
1
2
(
3
).
(34)
On the other hand, taking the inner product of both sides of (33) with e
3
yields
3
= y
3
. (35)
Using (34) and (35), we obtain by taking the inner product of both sides of (33) with
and
, respectively, that
z =
0
(
3
)
2
and y =
0
+ 2
0
2
. (36)
Hence (32) becomes
2
0
+ 2
2
e
0
(
0
+ 2
0
)
2
0
(
3
)
2
= 0. (37)
Taking the inner product of both sides of (37) with e
3
yields
3
=
_
2
+
0
_
3
,
23
which easily implies that
3
= c
0
3
= ac by (38), thus z = a
3
c
2
and y = 0 by (36). Replacing into (37), the e
0
-component
gives c
2
a
4
= 1, whereas for 1 i 2 the e
i
-component then yields
i
= (1/a)
i
. We
obtain that
i
= a
i
exp(s/a) + b
i
exp(s/a) for some a
i
, b
i
R, 1 i 2. Replacing
into 1 =
2
0
+
2
1
+
2
2
= a
2
+
2
1
+
2
2
implies that a
i
= 0 = b
i
for 1 i 2, i.e.,
1
= 0 =
2
, and that a = 1. Therefore f[
U
is totally geodesic, contradicting our
assumption.
Hence, there must exist an open interval J I such that
0
(s) ,= 0 for all s J, thus
we can reparametrize on J so that
0
(s) = s for all s J. Then the e
0
-component of
(37) gives
s
+ 2(c
2
s
4
1)
2
+ 2 = 0.
This is easily seen to be equivalent to
1
(s) = c
2
s
4
+ c
2
s
2
+ 1 for some c
2
R. Hence
the right-hand-side of the preceding equation is nowhere vanishing, and we can write
b
1
(s) = s
4
+ as
2
+ b for a = c
2
/c
2
and b = 1/c
2
, or equivalently,
(s) =
p
2
q
(s
2
p)
2
q
, p
2
> q, (39)
for p =
a
2
and q =
a
2
4
b. Equation (38) becomes
3
(s) =
s
_
(s
2
p)
2
q
. (40)
Taking the inner product of both sides of (37) with e
i
, 1 i 2, and using (39) and
(40) yields
s((s
2
p)
2
q)
i
+ (s
4
p
2
+ q)
s
3
i
= 0, 1 i 2.
Lemma 15. Let
i
: I R, 1 i 2, be linearly independent solutions of the ODE
s((s
2
p)
2
q)
i
+ (s
4
p
2
+ q)
s
3
i
= 0, p
2
> q, (41)
on an open interval I R where (s
2
p)
2
q > 0. Assume that s
2
+
2
1
+
2
2
= 1 for
all s I. Then (p, q) T, I J
p,q
and there exists R such that
2(
1
(s),
2
(s)) =
_
h(s) +
1 p
h(s)
_
(cos , sin )
_
q (1 p)
2
h(s)
(sin , cos ), (42)
where h(s) =
_
p s
2
+
_
(p s
2
)
2
q.
24
Proof: Let F be a primitive of : I R given by
(s) =
s
_
(s
2
p)
2
q
.
Then, it is easily checked that the functions
+
:= exp F and
:= exp (F)
form a basis of the space of solutions of (41) on I. Thus, there exist a
i
, b
i
R, 1 i 2,
such that
i
= a
i
+
+ b
i
, 1 i 2. (43)
Replacing into s
2
+
2
1
+
2
2
= 1 gives
s
2
+ Aexp(2F(s)) + B + C exp(2F(s)) = 0, for all s I, (44)
where A = a
2
1
+ a
2
2
, B = 2(a
1
b
1
+ a
2
b
2
) 1 and C = b
2
1
+ b
2
2
.
Assume that either of the following conditions holds:
(i) q < 0; (ii) q > 0 and p 0; (iii) q > 0, p > 0 and I is not contained in J
p,q
.
Then, up to a constant,
F(s) =
1
2
log
_
s
2
p +
_
(s
2
p)
2
q
_
,
hence (44) gives
A(u +
_
u
2
q) + C(u +
_
u
2
q)
1
= u + E,
where u = s
2
p and E = B p. This yields
2(2A+1)(CAq) = (2A+1)
2
q, (CAq)E = (2A+1)Eq and (CAq)
2
= E
2
q.
Since 2A+1 > 0, the rst and the third of the preceding equations give q = 0 if E = 0,
whereas the same conclusion follows from the rst and second equations if E ,= 0. This
is a contradiction and shows that either q = 0 or q > 0, p > 0 and I J
p,q
.
Let us consider rst the case q = 0. Suppose either that p < 0 or that p > 0 and I
is not contained in J
p,0
= (
p,
1 s
2
,
2
= sin
1 s
2
, and hence the statement is true in this case.
Now suppose that q > 0, p > 0 and I J
p,q
. Then
F(s) =
1
2
log
_
_
(p s
2
) q s
2
+ p
_
= log h(s), (45)
where h(s) is as in the statement. We obtain from (44) that
A(
_
u
2
q u)
1
+ C(
_
u
2
q u) = u + E,
with u = s
2
p and A, C, E as before. This is equivalent to
2(2C1)(AqC) = q(2C1)
2
, (AqC)E = q(2C1)E and (AqC)
2
= qE
2
,
and hence to
E = 0, A =
q
2
and C =
1
2
.
This gives
a
1
b
1
+ a
2
b
2
=
1
2
(1 p), a
2
1
+ a
2
2
=
q
2
and b
2
1
+ b
2
2
=
1
2
. (46)
By the last equation in (46), there exists R such that (b
1
, b
2
) =
2
2
(cos , sin ). Set
u := (cos , sin ) and v := (sin , cos ). Then the rst equation can be written as
(a
1
, a
2
), u =
2
2
(1 p). (47)
Using this and the second equation we obtain
q
2
= a
2
1
+ a
2
2
=
1
2
(p 1)
2
+(a
1
, a
2
), v
2
,
hence
(a
1
, a
2
), v
2
=
1
2
(q (p 1)
2
).
In particular, this shows that q (p 1)
2
, thus (p, q) T. Moreover, together with
(47) it implies that
a
1
=
2
2
((1p) cos
_
q (1 p)
2
sin ), a
2
=
2
2
((1p) cos
_
q (1 p)
2
sin ),
and (42) follows.
We obtain from Lemma 15 and (40) that (s) is given by
_
s,
2
2
_
h(s) +
1 p
h(s)
_
(cos , sin )
2
2
_
q (1 p)
2
h(s)
(sin , cos ), h log h(s)
_
,
26
and f[
U
can be parametrized by Y : S
m1
I S
m+1
R given by
Y (X, s) = (sX,
1
(s),
2
(s),
3
(s)).
Let A be the linear isometry of R
m+2
S
m+1
R dened by
Ae
m
= cos e
m
+ sin e
m+1
, Ae
m+1
= sin e
m
+ cos e
m+1
,
Ae
i
= e
i
for i 1, . . . , m and Ae
m+2
= e
m+2
. Then A
1
Y (X, s)he
m+2
= Y
p,q
(X, s).
It remains to prove assertion (v) in the direct statement. This is equivalent to
showing that Y
p,q
and Y
p
,q
do not parametrize congruent submanifolds for distinct
pairs (p, q) and (p
, q
p,q
()|d =
_
t
0
_
p,q
()d,
with
p,q
given by (39). If Y
p,q
and Y
p
,q
parametrize congruent submanifolds, then
the induced metrics, and hence the corresponding warping functions, must coincide. It
follows that
p,q
=
p
,q
, which easily implies that (p, q) = (p
, q
).
Acknowledgement: A rst draft of Lemma 6 was derived in a conversation of the second
author with M. Dajczer. We thank him for allowing us to include it here.
References
[1] U. Abresch and H. Rosenberg, A Hopf dierential for constant mean curvature
surfaces in S
2
R and H
2
R. Acta Math. 193 (2004) 141174.
[2] H. Alencar, M. do Carmo and R. Tribuzzi, A theorem of Hopf and the Cauchy-
Riemann inequality. Communications in Analysis and Geometry 15 (2007) 283298.
[3] H. Alencar, M. do Carmo and R. Tribuzzi, A Hopf theorem for ambient spaces of
dimensions higher than three. J. Di. Geom. 84 (2010), 1-17.
[4] H. Alencar, M. do Carmo and R. Tribuzzi, Surfaces in product spaces invariant
under ambient isometries and a characterization of the cylinder. Preprint.
[5] G. Calvaruso, D. Kowalczyk and J. Van der Veken, On extrinsically symmetric
hypersurfaces in H
n
R. Bull. Aust. Math. Soc. 82 (2010), 390-400.
[6] S. Carter and A. West, Partial tubes about immersed manifolds. Geom. Dedicata 54
(1995), 145169.
27
[7] S. Carter and U. Dursun, Partial tubes and Chen submanifolds, J. Geom. 63 (1998),
30-38.
[8] M. Dajczer, Reduction of codimension of regular isometric immersions, Math.Z.
179 (1982), 263-286.
[9] M. Dajczer et al., Submanifolds and Isometric Immersions, Math. Lecture Ser.
13, Publish or Perish Inc. Houston, 1990.
[10] F. Dillen, J. Fastenakels, J. Van der Veken, Rotation hypersurfaces in S
n
R and
H
n
R, Note di Matematica 29 (2008), 41-54.
[11] J. Erbacher, Reduction of the codimension of an isometric immersion, J. Di.
Geom. 5 (1971), 333-340.
[12] J. H. Lira, R. Tojeiro and F. Vitorio, A Bonnet theorem for isometric immersions
into products of space forms, Archiv der Math. 95 (2010), 469479.
[13] F. Mano and R. Tojeiro, Hypersurfaces with constant sectional curvature of S
n
R
and H
n
R, to appear in Illinois J. Math.
[14] R. Souam and E. Toubiana, Totally umbilic surfaces in homogeneous 3-manifolds.
Comment. Math. Helv. 84 (3) (2009), 673-704.
[15] R. Tojeiro, On a class of hypersurfaces in S
n
R and H
n
R. Bull. Braz. Math.
Soc. 41 (2) (2010), 199-209.
[16] J. Van der Veken and L. Vrancken, Parallel and semi-parallel hypersurfaces of
S
n
R. Bull. Braz. Math. Soc. 39 (2008), 355-370.
[17] S. T. Yau, Submanifolds with constant mean curvature I. Amer. J. Math. 96 (1974)
346-366.
Universidade Federal de Sao Carlos
Via Washington Luiz km 235
13565-905 Sao Carlos Brazil
e-mail: bruno@dm.ufscar.br
tojeiro@dm.ufscar.br
28