An Estimate for the Curvature of Bounded Submanifolds Author(s): L. Jorge and D. Koutroufiotis
Reviewed work(s):
Source: American Journal of Mathematics, Vol. 103, No. 4 (Aug., 1981), pp. 711-725 Published by: The Johns Hopkins University Press
Stable URL: http://www.jstor.org/stable/2374048 .
Accessed: 29/01/2013 09:43
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Mathematics.
OF BOUNDED SUBMANIFOLDS
By L. JORGE and D. KOUTROUFIOTIS*
1. Statement of the result. All manifolds considered in this paper shall be connected, of class C"' (smooth), and dimension at least 2. Im- mersions will also be smooth, and of codimension at least 1.
If M and M are Riemannian manifolds and (p: M - M is an isometric immersion with the property that sp(M) lies in a ball, we intend to estimate the supremum of the sectional curvature K of M in terms of the sectional curvature K of M and the radius of the ball. This is our result:
THEOREM A. Let Ml' be a complete Riemannian manifold whose scalar curvature is bounded below; let M" +q be a Riemannian manifold with q c n -1, and Bx a closed normal ball in Mt4+q, of radius X. Sup- pose Sp:M" - M"+q j is an isometric immersion with the property that ep(M"l) is contained in Bx. Take a positive number 6.
(i) If maxBx K = 62 and X < ir/26, then minBx K+ 62 [cotan(Xb6)12 < supm K,
(ii) If maxBx K = 0, then minBx K + 1/X2 < supM K,
(iii) If maxB, K_ =62, then minBx K+ 62[cotanh(X6)j2 ' supM K.
Note that (ii) is the limit case of (iii), as 6 tends to zero.
These estimates are sharp in the sense that, taking for M,,+q each of the three standard space forms, and for M"l = yO(Mn) the boundary of the corresponding normal ball Bx, we obtain equality.
Theorem A implies and extends several older non-immersibility theorems [14, 16, 7, 11] as well as some recent ones [2, 5]; we shall discuss them in Part 4 of this paper. The restriction on the codimension of the im- mersion is, of course, due to the use of a variant of Otsuki's lemma who also obtained results in this direction [15]. The well-known Chern-Kuiper extension [4] of a result of Topkins' [17] is the origin of this long series of
Manuscript received May 6, 1980.
*Work done under support of C.N.Pq., Brazil, at the Federal University of Ceara.
American Journal of Mathematics, Vol. 103, No. 4, pp. 711-725 711 0002-9327/81/1034-0711 $1.50
712 L. JORGE AND D. KOUTROUFIOTIS
papers on isometric immersibility of a compact manifold; we are able to consider here complete but not necessarily compact manifolds by making use of a powerful theorem of Omori's [13], as in [2] and [5]. From the proof of our theorem we also deduce, for hypersurfaces under appropriate assumptions, the existence of points where the second fundamental form is definite, thus generalizing known results [12] of this type.
Finally, the inequalities obtained in the proof of Theorem A permit us to give, modulo Omori's theorem, a simple proof of a recent result by Jorge and Xavier [8] concerning bounded isometric immersions of bounded mean curvature. We shall consider this in Part 5.
2. Preliminaries. Here we assemble all the formulae and proposi- tions that we shall use in the proof of Theorem A.
Let M and M be Riemannian manifolds, (p:M - M an isometric im- mersion. Henceforth, we shall tacitly make the usual identification of X E TpM with dop(X). In particular, if g:M - R is smooth and we consider the compositionf = g o s?, then we have at p E M for every X E TpM:
(gradf, X> = df(X) = dg(X) = <gradg, X >,
so that
gradg = gradf + (gradg)', (2.1)
where (grad g) l is perpendicular to TpM.
Take now a point x E M and consider in particular the function g:M - R given by
g(x) [d(xo, x)]2, (2.2) 2
where d is the Riemannian distance function on M. This g is smooth at x if x lies inside a normal ball around xo, and we can connect xo with x by a unique minimizing geodesic 'y:'y(O) = xo, oy(s) = x, I y'(0)I = 1. We have d(xo, x) = s and grad g at x is perpendicular to the hypersphere d(xo,) s; we obtain from this
Denote by V, respectively V, the Riemannian connection of M, respectively M. We wish to compute the Hessian form V 2f of the function
f
= g o p, where g is the function (2.2). By definition, V2f(X, Y) ( Vx gradf, Y>, at a point p E M and for X, Y E TpM. Denote by cx the second fundamental form of the immersion <p. Using the Gauss equation and (2.1), we have:(Vx gradf, Y> Vx gradf - a(X, grad f), Y> < Vx gradf, Y>
= X<gradf, Y> - <gradf, VxY> (< Vxgradg, Y>
+ <(gradg)', VxY>
= V2g(X, Y) + <(gradg)', o(X, Y)> = V2g(X, Y)
+ < grad g, c (X, Y) >.
Thus,
V2fp(X, Y) = V2g,,O(p)(X, Y) + (gradg, (x(X, Y)>s,(p). (2.4)
We wish to estimate the Hessian of the function (2.2) in terms of the curvature of the ambient manifold. The estimate below, known in various formulations, is obtained using well-known comparison methods. We give its proof for the sake of completeness of exposition.
(2.5) LEMMA. Let N be an m-dimensional Riemannian manifold with sectional curvature K, xo and x points of N so that x does not lie in the cut locus of xo. Let -y: [0, 1] - N be the minimizing geodesic segment connecting xo with x, parametrized by arc-length. Take a positive number 6. For any unit vector X E TxN, perpendicular to y' (1), the Hessian of the function (2.2) at x satisfies V 2g(X, X) -, u(l), where
h5.cotan(16) if max K = 62 and l < r/6
ydl) t I 1 if max K 0
16. cotanh(16) if max K = 62
-Y
Proof. Take a vector X E T,N as in the lemma, and a geodesic
,B:(-E, +E)-N so that O(O)-x, S'(O) = X. For e small, we can lift f to
714 L. JORGE AND D. KOUTROUFIOTIS
expj0(lv(s)) = 3(s), s E (-E, +E). Consider the smooth variation of the geodesic -y(t)
f:(-E, +E) X [0, l1 N, f(s, t) = expj0(tv(s)).
The vector field along -y(t)
J(t)
= (0, t) = (d expX0)t,Y(O)(tv'(0))as
is a Jacobi field which satisfies
J(0) = 0, J(I) = X, (J(t), -y'(t)>3 0
The third equality is a consequence of the assumption that X is perpen- dicular to -y'(1).
Consider now the function g along the geodesic f. We compute easily (g o 3)"(0) = V2g(X, X). On the other hand,
1
af
-2 i/ a2 1 (go
f3)(s)
=Ij[j
V| | dt = 1 af dt = IE(s),2 0 at
2 0 at
2
where E stands for the energy of the geodesic -ys t -f(s, t). Thus, (g 0)"(0) = (1/2)lE"(0). Now using the fact that the variation has fixed left endpoint, and that 3 is a geodesic, we have (1/2)E"(0) = I,(J, J), where II is the index form of -y = -yo; that is, for a vector field V along -y,
I,(V, V) = {<V', V'> - <R(y', V)'y', V>}dt,
where V' is the covariant derivative of V along -y and R is the curvature tensor of N. Consult [10], p. 75, for the above second variation formula for the energy of -y. We use here the sign convention for R adopted in [10], p. 51.
It follows now from the above that V 2g(X, X) = II,(J, J).
gSi? if max K = 62
-Y
N= Rni if maxK = 0
-Y
Hn" if maxK =- 62
-Y
where S", is the sphere of curvature 62 and H"1 the hyperbolic space of cur- vature -62. Pick a point E E N and a geodesic j(t), t arc-length, with _y(0) = - ; set y(l) = xi. We have I < 7r/6 in case max) K = 62, SO y con- tains no points conjugate to x-0 on (0, 1]. Set e,,,(t) = -y'(t) and let {eI(t),
en,(t)} be an orthonormal basis field, parallel along -y. Set e,(t) = y'(t) and let {Ie(t), ..., e#(t)} be an orthonormal basis field, parallel along y. If J(t) = 7I gi(t)ei(t), define (4J)(t) = eI gi(t)e(t). It is easy to see, using (J, -y'> = 0 =(J, ( y'> and the curvature assump- tions, that the index forms I and I of -y and y satisfy I,(J, J) > I,(4J, 4J).
Define now a vector field W along y by setting W(l) = (4J)(l) and let- ting W(t) be the parallel-translate of W(l) along y, from xl to xiT. Clearly,
W is a unit vector field, perpendicular to -. Consider the Jacobi field
sin(6t)
W(t) if maxK = 62andl < - sin(61) 'Y6
J(t)= W(t) if maxK=0
sinh(lt) W(t) if max K =- 62. sinh(61) '
JsatisfiesJ(0) = 0 = (bJ)(0), J(l) = (4J)(l), (J(t),y '(t)> 0 O. We con- clude from the Index Lemma (see for example [3], p. 24), since y contains no points conjugate to xi0 on (0, 11, that l,(J, J) < 11(4J, (J). Therefore,
V2g(X, X) > lI,(J, J). We compute now I,(J, J) and obtain the an- nounced estimate. This ends the proof of Lemma (2.5).
The main tool in the proof of our result will be the following theorem due to Omori [13].
716 L. JORGE AND D. KOUTROUFIOTIS
propertiesf(p) >f(p(), IgradfIp < -, V2f(X,X) < EtXt2foreveryXE
TpM.
Finally we mention a variant of what is generally known as Otsuki's lemma, which we shall use further on in the manner of [7]. Its proof is contained in the proof of the lemma on p. 28 of [9].
(2.6) Let a::Rn X Rn - Rq, q < n - 1, be a bilinear symmetric mapping satisfying ox(X, X) ? Ofor X ? 0. There exist linearly indepen- dent vectors X, Y so that a(X, X) = a(X, Y) and ox(X, Y) = 0.
3. The proof. Denote by xo the center of the closed normal ball Bx. Let d be the distance in M and g the function (2.2); it is smooth on BX since the cut locus of xo lies outside Bx. Setf = g o (p; nowf is a smooth function on M because (p(M) C BA, andf ' X2/2. We may assume that
infM K > -oo since, otherwise, the hypothesis that the scalar curvature of
M is bounded below would entail supM K = +oo, and therefore the estimates in Theorem A would be satisfied trivially. Thus, we may apply Omori's theorem to the functionf. Pick a pointp- E M so that (p(j-) = x- ? x0. For every positive integer m, there exists a point Pn, E M satisfying
1 -)
,1
f(Pm) 2f(p) = - [d(xo,x)]2> 0, tgradfp,,, <-9
2 Pin m
v2f(X, X) <-1X12 for allX Tp",M. M (3.1)
m Ir
Set xn, = (P(P,,1). Let -yM,1 be the geodesic segment of length Sn,9
emanating from xo and ending at x,,1, which realizes the distance between
xo and xM :d(xo, x,,,) = sni. Put d(xo, xJ) = s; we have 0 < s < S_ < .
From (2.3) we know that
I
grad gI
= s n; also-Igrad g * I oe(X, X)I < (grad g, oe(X, X)>,
so, using (2.4), we deduce from the third inequality in (3.1) that, at Pn, and for all X E TP,, M,
1X12.
v
2g(X, X) -Sn,1I U(X, X) I <-
From now on, for the sake of economy of notation, we omit the bar over V and rewrite the above inequality for all X ? 0 at pill in the form
1 [V2g(X,X) I 1 IaU(X,X)I(
s,,,
L
lx12
mJ| < X2(3.2)We wish to estimate the Hessian of g at X E TpM. To this purpose, we decompose X as
X = XT + XN, (3.3)
where XT is normal to grad g and XN is collinear with grad g. Lemma (2.5) supplies an estimate for V2g(XT, XT); we show, using the second ine-
quality in (3.1), that the remaining terms in
V2g(X, X) _ V2g(XT, XT) V2g(XT, XN) V 2g(XN, XN)
=
~
~
+ +IX12 IXxI2 IXI 2
(3.4)
are negligible for n large. This is a consequence of the fact that, by con-
struction, the tangent space to so(M) at x,, tends to become perpendicular
to the radial geodesics as m tends to infinity. More precisely, we have at
Xnz
<X, grad g>
XN='n Si
Snm
and so, by (2.1) and the second inequality in (3.1),
I XN I < 1 . (3.5)
It follows from this and (3.3) that
s
- I < (3.6)
718 L. JORGE AND D. KOUTROUFIOTIS
Now, at every point x E BX, the symmetric bilinear form V 2g on T,,M can
be estimated:
IV2g(X,Y)j = <LXX, Y>I ' IILx IXI X IYI,
where Lx is the linear transformation X - Vx grad g on TXM. By con- tinuity and compactness of Bx, we have sup{ I Lx I , x E Bx} = c < oo.
Thus, I V2g(X, Y) I < c IXI X I YI on Bx. Using this, (3.5) and I XT I < IXI, we obtain from (3.4)
V2g(X,X) V2g(XT,XT)_ 3c for all non-zero
XETp ,M.
IX12X2 msni "
Now we have at x,,,, according to Lemma (2.5), (3.7)
V 2g(XT, XT) IXrTI (3.8)
> IA s,,,).8
Recall that 0 < s c s,,i. Thus, for m large enough, the left-hand side of (3.6) is positive. This enables us to combine (3.6), (3.7), (3.8) and obtain
V2g(X, X) ( _)(1 __ _2 3c
t ( )
IX12 JmSm Msm a Xm~op
(3.9)
We restrict the radius of Bx further in the first case of Lemma (2.5) to ensure that u(sm) is bounded below by a positive number:
(6s cotan(Gsm) bs-cotan(6X) > 0 if 26
A(SM) = I
16sm cotanh(6sm) 2 bicotanh(6X) > 0.
From (3.9) we deduce now
for all non-zero vectors X E TpmM and m large enough. It follows from (3.2) that a(X, X) ? 0 for all non-zero X E TpmM, and Lemma (2.6) is ap- plicable: We pick vectors X, Y as in (2.6), write down the corresponding inequalities (3.2), and multiply them term-by-term. Applying (3.9), we obtain
1 [2(sI) (1 m-) - m
Q2<
|x(X, X) |a(Y,Y)|
sm2 mm msm m |X12|
y12
(a(X, X), a(Y, Y)> - Ia(X, y)1 2
IX12. I y12
<
(C!(X, X), c(Y, Y)) - (I(X, y) 1 2 IXI2. 1 yI2 - <X, y>2
Thus, by the Gauss equation, we have at .p(p.), for m large, a plane a = X A Y satisfying
0 < Srn2 [,sm)(1 - m ) _ncm
112
< K(a)-K(r).sm
~
~~~i mm msm m(3.11)
It follows that
minK+ I[.]2 <supK. (3.12)
BX Sm2 M
We take now a convergent subsequence of the sequence {sm}, which we denote again by {sm }:
O<Ts? limsm=XcX.
m -00
From (3.12) we obtain
min K + [y(X)]2 C sup K.
720 L. JORGE AND D. KOUTROUFIOTIS
However, in the three cases under consideration,
(62[cotan(bX)12
X\ yX1 2 X 21
[(2[cotanh(bX)12
since the functions involved are monotonically decreasing, and the proof of Theorem A is complete.
4. Applications. Theorem A applied to the unit sphere Sn+q C
Rn +q+l yields immediately:
THEOREM 1. Let Mn be a complete Riemannian manifold with scalar curvature bounded below, and p:Mn - Sn+q, q ' n - 1, an
isometric immersion with the property that #p(Mn) lies inside a geodesic ball of radius X < Ir/2. Then 1 + (cotanX)2 < supm K.
This implies, under the above restriction on codimension, that a complete Mn with sectional curvature K c 1 and scalar curvature bounded below, which is isometrically immersed in Sn+q, must have accumulation points on every great hypersphere; and must in fact intersect every great hypersphere if it is compact. This was proved by Hasanis [5] using an estimate for the supremum of K which is somewhat weaker than ours.
An Hadamard manifold M is a simply-connected complete Rieman- nian manifold with non-positive sectional curvature. Recall that every point p of an Hadamard manifold is a pole, that is, expp: TpM - M is a diffeomorphism.
THEOREM 2. Let Mn be a complete Riemannian manifold with
scalar curvature bounded below, Mn+q an Hadamard manifold with q c
n - 1, and sp: Mn _ Mn +q an isometric immersion satisfying K(U) ' K(U) at every point of Mn and for every plane a through it. Then #p(Mn) is un- bounded.
Proof. We argue indirectly. If #p(Mn) were bounded, it would be contained in a normal ball. But then, as we have shown in the proof of Theorem A, there would exist a point in Mn and a plane a through it so that inequalities (3.11) hold; this is a contradiction.
isometric immersion of Mn into an Hadamard manifold Mn+q so that K(a) ' K(or) everywhere; this result is due to O'Neill [14].
Note that results similar to Theorem 2 can be obtained by assuming merely that Mn+q possesses a pole.
The case Mn+q = Rn+q of Theorem 2 is due to Baikousis and Koufogiorgos [2], and Jacobowitz [7] if Ml' is assumed compact. Here is an interesting implication in this case: a complete M2 with Gaussian cur- vature satisfying -00 < a < K < 0, isometrically immersed in R3, is ex-
trinsically unbounded.
THEOREM 3. Let Mn+q, q ' n - 1, be an Hadamard manifold with sectional curvature satisfying a < K < b < 0; let Mn be a complete Riemannian manifold with scalar curvature bounded below and sectional curvature satisfying K c a - b. If Sc : _ Mn +q is an isometric immer- sion, then #o(Mn) is unbounded.
Proof. Say sp(Mn) C B, and
a ' minK c maxK = -62 < b < 0, 6 > 0.
BX BX
Then,
a - b < a - b[cotanh(GX)]2 c min K + 62[cotanh(6X)]2 c sup K.
BX M
Thus, there exists a point in Mn and a plane a through it so that a - b <
K(o), which is a contradiction.
The case maxBx K = b = 0 is dealt with similarly, using inequality (ii) of Theorem A. This completes the proof.
In particular, if Mn is compact and all the hypotheses of Theorem 3 hold, there exists no isometric immersion of Mn into M' +q; this is due to J. D. Moore [11]. The special case K = const. < 0 is Stiel's theorem [16].
THEOREM 4. Let M be a Riemannian manifold and let M be a com- plete hypersurface of M, whose sectional curvature is bounded below. Suppose that M lies in a closed normal ball Bx of M. If either maxBx K = 62 with 6 > 0 and X < r/(26) or maxBx K c 0, then M has points where the second fundamental form is definite.
722 L. JORGE AND D. KOUTROUFIOTIS
hand side is positive at some point of M for all tangent vectors X, a fact we proved later on (inequality (3.10)).
Theorem 4 was proven by S. B. Myers ([12], p. 214) in case M is assumed compact. The case where M is the euclidean space is due to Baikousis and Koufogiorgos [2]. Clearly, inequality (3.2) contains infor- mation about the second fundamental form even for codimension larger than one; we shall not pursue this however.
5. Immersions of bounded mean curvature. The estimates derived in the proof of Theorem A permit us to give a succinct proof of the follow- ing result, due to Jorge and Xavier [8].
THEOREM H. Let M be a complete Riemannian manifold whose scalar curvature is bounded below, and <p:M - M an isometric immer- sion into the Riemannian manifold M. Suppose <p(M) is contained in a normal ball Bx, and the mean curvature vector H of <p satisfies
I
HI <H,where Ho is a positive constant. Take a positive number 6.
(i) If maxJ3 K
=
62 and X < 7r/26, then X 2 (1/6) arctan(6/H0). (ii) If maxBX K = 0, then X> 12/H,(iii) If maxBx K =-62, then 6 < Ho and X 2 (1/6) arctanh(6/H0).
Note that no codimension assumption is necessary.
We need the lemma below, which enables us to apply Omori's theorem.
LEMMA. Let M and M be Riemannian manifolds with dim M < dim M; assume the scalar curvature of M is bounded below. If there exists an isometric immersion p: M - M satisfying
I
HI s Ho and <p(M) C B with B compact, then the sectional curvature of M is bounded in absolute value.Proof of the Lemma. Suppose the scalar curvature R of M satisfies inf R = a > -xo. Take a point p E M and an orthonormal basis {X1}1=I of TpM, where n = dim M. We have the following identity, obtained from the Gauss equation by contraction:
R = E K(Xi A Xj) + n2IH12-S, i?j
S s n(n - 1) maxK + n2HO 2+ a = const., B
so S is bounded on M. Now, for an arbitrary plane X A Y at so(p),
IK(X A Y) l I K(X A Y) I + I (c(X, X), a(Y, Y)>-I a(X, Y) 1 21
s max K + 2S < const.
B
Proof of Theorent H. The notation is the same as in Part 3. By vir- tue of the previous lemma and the assumptions, we may again apply Omori's theorem to the functionf = g o sp, and construct the sequence of points pm which satisfy (3.1). Taking an orthonormal basis {Xi}1=, of TpmM, we have from (2.4) for the Laplacian of f at Pm:
n
~~~~~~~n
Af = V2g(Xi, X) + (grad g, nH) <-.
From I HI < Ho_ we obtain by virtue of the Cauchy-Schwarz inequality:
-nSmHo < (grad g, nH>,
and from (3.9),
n[!t(sm) (1 - m:) - < E V2g(X1,X,).
Combining the last three inequalities, we have at pm:
IM(S ) I1- )- _ i, - < <SI H0
M M msm m
Now we let m go to infinity, pick a convergent subsequence of the se- quence {sn }, and obtain, as in the proof of Theorem A, a positive number X satisfying X s X and A(X) s XHO. This last inequality, together with the definition (2.5) of 14, imply immediately the assertion in case (ii). In case (i), we obtain for X < ir/a:
724 L. JORGE AND D. KOUTROUFIOTIS
If X < Or/(26), so cotan(X6) > 0, hence tan(X6) 2 6/Ho and the assertion follows from the monotonicity of the tangent function. In case (iii) we have
6 < 6 cotanh(X6) c 6 cotanh(X6) c Ho,
so 6 < Ho and tanh(X6) 2 6/H_. Since 6/H0 < 1, we can take inverses, and the assertion follows from the monotonicity of the hyperbolic tangent function.
The case M = RN and dim M = 2 of Theorem H is due to Aminov [1], the case M = RN to Hasanis and Koutroufiotis [6], the case M = SN
to Hasanis [5]. Theorem H implies in particular that a complete manifold with scalar curvature bounded below, isometrically and minimally im- mersed in an Hadamard manifold M, must be unbounded in M.
A consequence of case (iii) may be noteworthy: a compact sub- manifold of the hyperbolic space of sectional curvature -1, must satisfy max IHI > 1.
UNIVERSIDADE FEDERAL DO CEARA UNIVERSITY OF IOANNINA
REFERENCES
[11 Ju. A. Aminov, "The exterior diameter of an immersed Riemannian manifold," Math. USSR Sbornik 21 (1973), pp. 449-454 (AMS translation).
[21 Ch. Baikousis and Th. Koufogiorgos, "Isometric immersions of complete Riemannian manifolds into euclidean space," to appear in the Proc. Amer. Math. Soc. [3] J. Cheeger and D. G. Ebin, "Comparison theorems in Riemannian geometry," North-
Holland Mathematical Library, vol. 9, 1975.
[41 S. S. Chern and N. H. Kuiper, "Some theorems on the isometric embedding of com- pact Riemann manifolds in euclidean space," Ann. of Math. (2) 56 (1952), pp. 442-430.
[5] Th. Hasanis, "Isometric immersions into spheres," to appear in J. Math. Soc. Japan. [6] Th. Hasanis and D. Koutroufiotis, "Immersions of bounded mean curvature," Archiv
der Math. 33 (1979), pp. 170-171, and "Addendum" to appear.
[7] H. Jacobowitz, "Isometric embedding of a compact Riemannian manifold into euclid- ean space," Proc. Amer. Math. Soc. 40 (1973), pp. 245-246.
[9] S. Kobayashi and K. Nomizu, "Foundations of differential geometry," vol. II, Inter- science Tracts in Pure and Appl. Math., no. 15, New York, 1969.
[10] J. Milnor, "Morse theory," Annals of Math. Studies, no. 51, Princeton University Press 1969.
[11] J. D. Moore, "An application of second variation to submanifold theory," Duke Math. J. 42 (1975), pp. 191-193.
[12] S. B. Myers, "Curvature of closed hypersurfaces and non-existence of closed minimal hypersurfaces," Trans. Amer. Math. Soc. 71 (1951), pp. 211-217.
[13] H. Omori, "Isometric immersions of Riemannian manifolds," J. Math. Soc. Japan 19 (1967), pp. 205-214.
[14] B. O'Neill, "Immersions of manifolds of non-positive curvature," Proc. Amer. Math. Soc. 11 (1960), pp. 132-134.
[15] T. Otsuki, "Isometric imbedding of Riemannian manifolds in a Riemannian mani- fold," J. Math. Soc. Japan 6 (1954), pp. 221-234.
t16] E. Stiel, "Immersions into manifolds of constant negative curvature," Proc. Amer. Math. Soc. 18 (1967), pp. 713-715.
[17] C. Tompkins, "Isometric embedding of flat manifolds in euclidean space," Duke Math. J. 5 (1939), pp. 58-61.