A THEOREM OF PICK{ BERWALD TYPE
FOR A TOTALLY UMBILICAL AFFINE
IMMERSION OF GENERAL CODIMENSION
Naoyuki Koike and Yoko Torisawa
(Received October 26, 1998)
Abstract. It is known that a totally umbilical ane immersion of general
codimension into an ane space is anely congruent to a graph immersion or a centro-ane immersion. In this paper, we shall investigate a more detailed property of such an immersion.
AMS1991 Mathematics Subject Classication. 53A15.
Key words and phrases. totally umbilical ane immersion, ane shape tensor, ane fundamental form, hyperquadric.
Introduction
Throughout this paper, unless otherwise mentioned, we assume that all objects are of classC
1 and all manifolds are connected ones without
bound-ary. Also, denote by ;(E) the space of all cross sections of a vector bundle E. An ane (or equiane) immersion of codimension one into the (n+
1)-dimensional ane space (
R
n+1r~ !~) with the natural equiane structure has
been studied by some geometricians. In particular, if an equiane immersion into (
R
n+1r~ !~) satises the volume condition, then it is called a Blaschke
immersion (see 1], 5]). For a Blaschke immersion, G. Pick and L. Berwald proved the following result (see 5]):
If f is a Blaschke immersion of an n(2)-dimensional manifold (Mr)
with equiane structure into (
R
n+1r~ !~) and its cubic form (i.e., the
co-variant dierentiation of its ane fundamental form) vanishes, then f(M) is
contained in a hyperquadric in
R
n+1.Here we note that such an immersion is totally umbilic (see 5]). On the other hand, in 3], K. Nomizu and U. Pinkall proved the following characterization theorem for a totally umbilical ane immersion of general codimension into an ane space with the natural torsion-free ane connection:
Letf be a totally umbilical ane immersion of ann-dimensional manifold
(Mr) with torsion-free ane connection into the (n+r)-dimensional ane
space (
R
n+rr~) with the natural torsion-free ane connection, where n2
andr 1. Thenf is anely congruent to a graph immersion or a centro-ane
immersion.
Here a graph immersion is dened as follows. LetF be an
R
r-valued function
on then-dimensional ane space
R
nand f an immersion ofR
ninto (R
n+r r~) dened by f(x) = (xF(x)) 2R
nR
r =R
n+r ( x 2R
). Let N be thetransversal bundle along f such that N x (
x 2
R
n) are parallel to the ane
subspace
R
rofR
n+r. Denote byrthe induced connection on
R
nfor N. Then f is an ane immersion of (R
n r) into (R
n+rr~). Such an ane immersion is
called a graph immersion. Note that its ane shape tensor vanishes identically. Also, a centro-ane immersion is dened as follows. Letf be an immersion of
ann-dimensional manifoldM into (
R
n+r r~) admitting an (r;1)-dimensional vector subspace VofR
n+r such that f ( T x M)Spanff(x)gV =R
n+rholds for everyx2M, whereT x
M is the tangent space ofM at x, we identify T
f(x)
R
n+r with
R
n+r andf(x) is its position vector. Dene a tansversal
bundle N along f by N
x = Span
ff(x)gV (x 2 M). Denote by r the
induced connection onM forN. Thenf is an ane immersion of (Mr) into
(
R
n+rr~). Such an ane immersion is called the centro-ane immersion.
Note that its ane shape tensor A does not vanish and that V = Ker x
(x2M) holds, where is a cross section ofN with
A=I.
In this paper, we shall prove the following result similar to the Pick-Berald theorem for a totally umbilical ane immersion of general codimension.
Theorem.
Letf be a totally umbilical ane immersion of an n-dimensionalmanifold (Mr) with torsion-free ane connection into the (n+r)-dimensional
ane space (
R
n+rr~) with the natural torsion-free ane connection, where n2 andr 1. If its ane shape tensorAdoes not vanish and the covariant
dieretiationrof its ane fundamental form vanishes identically, thenf(M)
is contained in a cylinder over a hyperquadric in an (n+1)-dimensional ane
subspace of
R
n+r.Here we note thatf(M) is not necessarily contained in an (n+1)-dimensional
ane subspace of
R
n+r in spite of being totally umbilic andr= 0. In fact,
according to the reduction theorem for an ane immersion of K. Nomizu and U. Pinkall (see 3]), the conditionr= 0 implies that the dimension of its rst
normal space N 1
x at
x(i.e., the linear span of the image of
x) is independent
of the choice of x 2M and that f(M) is contained in an (n+s)-dimension
ane subspace of
R
n+r, wheres = dimN 1
x, but the totally umbilicity of f
does not necessarily imply dim N 1
x = 1. In
x3, we shall give an example of a
totally umbilical ane immersion as in the statement of Theorem such that the dimension of its rst normal space is more than one.
x
1. Fundamental formulas and denitions
In this section, we shall recall the fundamental formulas and denitions for an ane immersion. Let (Mr) (resp. ( ~Mr~)) be an n (resp. (n+ r))-dimensional manifold with torsion-free ane connection. An immersion f : (Mr) ,! ( ~Mr~) is called an ane immersion if there is a transversal
bundle N along f such that for every tangent vector elds X and Y on M,
~ r X f Y ;f ( r X
Y) is a cross section of N. Note that the choice of such a
transversal bundle N in general is not unique. In the sequel, we x such a
bundleN. Set (XY) := ~r X f Y ;f ( r X Y):
This quantity becomes an N-valued symmetric tensor eld of type (02)
on M. This tensor eld is called the ane fundamental form of f. For a
transversal vector eld along f (i.e., 2;(N)), we write
~ r X =;f ( A X) +r ? X where A X 2 ;(TM) and r ? X
2 ;(N). This quantities A becomes a cross
section of the tensor product bundle N T M TM and r ? becomes a connection onN, whereN (resp. T
M) is the dual bundle ofN (resp. TM).
This tensor eldA is called the ane shape tensor of f and r
? is called the
transversal connection of f. The covariant dierentiationr of is dened
by (r X )(YZ) :=r ? X( (YZ));(r X YZ);(Yr X Z)
for XYZ 2 ;(TM). The ane immersion f is said to be totally umbilic if
there is 2;(N
) with
A=I, whereI is the identity transformation of TM.
x
2. Proof of Theorem
In this section, we shall prove Theorem stated in Introduction. Proof of Theorem. Let be a cross section ofN
with
A=I. Sincef is
totally umbilic andA6= 0,f is anely congruent to a centro-ane immersion
from the Nomizu-Pinkall theorem. Hence, the transversal space N x of f at x2M is decomposed as follows: N x = Span ff(x)gKer x :
Now we dene a hypersurfaceF x ( x2M) in
R
n+r as F x:= ff(x) +f ( U) + f(x) +jU 2T x M 2R
2Ker x x( UU);( 2+ 2 )f(x) 0 (mod Ker x) g:Fig. 2.
We show that Fx is a cylinder over a hyperquadric in an (
n+ 1)-dimensional
ane subspace of
R
n+r. Let p 1 (resp. p 2) be the projection of N x onto Ker x(resp. Spanff(x)g). Then it is easy to show that x(
UU);( 2+2
)f(x) 0
(mod Ker
x) holds if and only if
(2.1) p 2( x( UU)) = ( 2 + 2 )f(x)
holds. We dene a symmetric bilinear form h onT x M by p 2( x( XY)) =h(XY)f(x) forX Y 2T x M. Then (2:1) is equivalent to (2.2) h(UU) = 2 + 2 :
Let (e 1 e n) be a basis of T x M. We putU = n P i=1 U i e i and h ij = h(e i e j) (ij = 1n). Then (2:2) is rewritten as (2.3) n X ij=1 h ij U i U j = 2 + 2 : Let= (y 1 y
n+1) be the ane coordinate system of the (
n
+1)-dimension-al ane subspacef (
T
x
M)Spanff(x)gassociated with the basis (f e 1 f e n
f(x)), where the origin is the pointf(x). Setv:=f(x)+f ( U)+ f(x). Then we have (v) =(f(x) + n X i=1 U i f e i+ f(x)) = (U 1 U n ) that is, y i( v) =U i ( i= 1n) y n+1( v) = : Hence (2:3) is rewritten as n X ij=1 h ij y i( v)y j( v) =y n+1( v) 2 + 2y n+1( v):
Therefore, by noticing that Ker is parallel with respect to ~r, we see thatF x
is a cylinder over a hyperquadric n P ij=1 h ij y i y j = y 2 n+1+ 2 y n+1 in the ( n+
1)-dimensional ane subspacef (
T
x
M)Spanff(x)g of
R
n+r.Now we shall show thatf(M) is contained inF x. Fix x 0 2M andz 0 2F x 0.
We dene a tangent vector eld ~U onM, a function ~ onM and a transversal
vector eld ~ on M satisfying ~ x 2Ker x ( x2M) by z 0 = f(x) +f (~ U x) + ~ x f(x) + ~ x ( x2M):
Fig. 3.
Then we have ~ r X z 0 = (~ + 1)f ( X) +f ( r X~ U) +(XU~) + (X ~)f+r ? X~ = 0
for every tangent vector eldX on M. By taking notice of the tangent
com-ponent and the transversal comcom-ponent of this equation, we have
(2.4) r X~ U =;(~ + 1)X and (2.5) (XU~) =;(X ~)f ;r ? X~ :
Now we dene the transversal vector eld onM by
=( ~UU~);(~ 2+ 2~
)f:
From r= 0, (2:4) and (2:5), we have r ? X = r ? X( ( ~UU~));r ? X((~ 2+ 2~ )f) = 2(r X~ UU~);2(~ + 1)(X ~)f ;(~ 2+ 2~ )r ? X f =;2(~ + 1)(XU~);2(~ + 1)(X ~)f = 2(~ + 1)(X ~)f+ 2(~ + 1)r ? X~ ;2(~ + 1)(X ~)f = 2(~ + 1)r ? X~ :
Since Ker is parallel with respect to r
?, we have r ? X~ 2 Ker. Hence we have r ? X 0 (mod Ker): It follows fromz 0 2F x
0 and the denitions of ~
U and ~ that x 0 = (~U x 0 U~ x 0) ;(~ 2 x0+ 2~ x 0) f(x 0) 0 (mod Ker x 0) : Hence we have x= (~U x U~ x) ;(~ 2 x+ 2~ x) f(x) 0 (mod Ker x)
for everyx2M. Therefore, we can obtain z 0 = f(x) +f (~ U x) + ~ x f(x) + ~ x 2F x ( x2M):
This together with the arbitrariness of z
0 deduces F x 0 F x ( x 2 M).
Fur-thermore, from the arbitrariness ofx 0 and
x, we haveF x0 =
F
f(x) 2 F
x0 because of
f(x) 2 F
x. After all, from the arbitrariness of x, we
can obtainf(M)F x
0. This has completed the proof.
x
3. An example
In this section, we shall give an example of a totally umbilical ane im-mersion as in the statement of Theorem such that the dimension of its rst normal space is more than one.
Example. Dene a map f from
R
n to
R
n+r by f(x 1 x n) := ( x 1 x n a 1 x 2 1+ b 1 x 1+ c 1 a r 0x 2 r 0+ b r 0x r 0+c r 000c) ((x 1 x n) 2R
n) wheren2 r3 r 0:= min fnr;1g a i( i= 1r 0) and care non-zero constants, and b i and c i ( i = 1 r0) are constants. Also, dene a map N
from
R
n to the Grassmann manifold G rn+r of all r-dimensional subspaces ofR
n+r by N (x1xn) := Span f @ @y n+1 @ @y n+r;1 f(x 1 x n) g ((x 1 x n) 2R
n) where (y 1 yn+r) is the natural coordinate system of
R
n+r and f(x
1 x
n) is its position vector. Easily we have f ( @ @x i ) = @ @y i + (2a i x i+ b i) @ @y n+i (i= 1 r 0) f ( @ @x j ) = @ @y j (j=r 0+ 1 n) f(x 1 x n) = n X j=1 x j @ @y j + r 0 X j=1 (a j x 2 j + b j x j+ c j) @ @y n+j +c @ @y n+r :
From these relations, we can show the linearly independence off ( @ @x 1) f ( @ @xn) @ @yn+1 @ @yn+r;1 f(x 1 x n). That is, f is an immersion andN
is regarded as a transversal bundle alongf. Let ~rbe the natural torsion-free
ane connection of
R
n+r andrthe induced connection on
R
n forSinceN = Spanf @ @yn+1 @ @yn+r;1 gSpanff(x 1 x n) gand Spanf @ @yn+1 @ @y n+r;1
gis parallel with respect to ~r, the immersionf is a centro-ane
immersion of (
R
nr) into (
R
n+rr~). That is, the immersion f is totally
umbilic and its ane shape tensor does not vanish. Concretely its ane shape tensorA is given byA=I, where is the cross section of N
dened by (f(x 1 x n)) = ;1 and ( @ @y n+i) = 0 ( i= 1r;1). Also, we have ~ r @ @x i f ( @ @x i ) = 2a i @ @y n+i (i= 1 r 0) ~ r @ @x i f ( @ @x i ) = 0 (i=r 0+ 1 n) r~ @ @x i f ( @ @x j ) = 0 (1i6=jn) and hence ( @ @x i @ @x i ) =2a i @ @y n+i (i= 1r 0) ( @ @x i @ @x i ) = 0 (i=r 0+ 1 n) ( @ @x i @ @x j ) = 0 (1i6=j n) r @ @x i @ @x j = 0 (ij= 1n):
Thus its rst normal space is spanned by @ @yn+1 @ @y n+r 0 at each point of M, that is, its dimension is equal tor
0 ( 2). Also, we have (r @ @x i )( @ @x j @ @x k ) =r ? @ @x i (( @ @x j @ @x k ));(r @ @x i @ @x j @ @x k ) ;( @ @x j r @ @x i @ @x k ) = 0 (ijk = 1 n)
which impliesr = 0. Thus this ane immersionf : (
R
nr),! (
R
n+rr~)
is a desired totally umbilical ane immersion.
Acknowledgement
The authors would like to express their sincere gratitude to Professor S. Yamaguchi for his helpful advice and to Professor N. Abe for his constant encouragement.
References
1] Blaschke, W., Vorlesungen uber Dierentialgeometrie II, Ane Dierentialgeometrie,
Springer, Berlin, 1923.
2] Erbacher, J., Reduction of the codimension of an isometric immersion, J. Dierential Geometry5(1971), 333-340.
3] Nomizu, K. and Pinkall, U., Cubic form theorem for ane immersions, Results in Math.
13(1988), 338-362.
4] Nomizu, K. and Sasaki, T., Centroane immersions of codimension two and projective hypersurface theory, Nagoya Math. J.132(1993), 63-90.
5] Nomizu, K. and Sasaki, T., Ane dierential geometry, Cambridge University Press, 1994.
6] Simon, U., Schmenk-Schellschmidt, A. and Viesel, H., Introduction to the ane dier-ential geometry of hypersurfaces, Lecture Notes, Science university of Tokyo, 1991. Naoyuki Koike
Department of Mathematics Faculty of Science Science University of Tokyo Wakamiya-cho 26, Shinjuku-ku Tokyo 162-0827, Japan
Yoko Torisawa
Department of Mathematics Faculty of Science Science University of Tokyo Wakamiya-cho 26, Shinjuku-ku Tokyo 162-0827, Japan