On the Moebius deformable hypersurfaces

Let f_2\colon M^n\to\mathbb{R}^{n+1} be a Moebius deformation of f_1:=f. By Proposition 6, the Moebius shape operators S_1 and S_2 of f_1 and f_2, respectively, share a common kernel \Delta of dimension n-2. Let S_i, i=1,2, denote also the restriction S_i|_{\Delta^\perp} and define D\in\Gamma(\mbox{End}(\Delta^\perp)) by

It follows from Proposition 6 that \det D=1 at any point of M^n, while Proposition 3 implies that D cannot be the identity endomorphism up to sign on any open subset U\subset M^n, for otherwise f_1|_U and f_2|_U would be Moebius congruent by Lemma 7. Therefore, we can write D=aI+bJ, where b does not vanish on any open subset of M^n and J\in\Gamma(\mbox{End}(\Delta^\perp)) satisfies J^2=\epsilon I, with \epsilon\in \{1,0,-1\}, J\neq I if \epsilon=1 and J\neq 0 if \epsilon=0.

From the symmetry of S_2 and the fact that b does not vanish on any open subset of M^n, we see that S_1J must be symmetric. Moreover, given that \mbox{tr\,} S_1=0=\mbox{tr\,} S_2, also \mbox{tr\,} S_1J=0.

Assume first that J^2=0. Let X,Y\in\Gamma(\Delta^\perp) be orthogonal vector fields, with Y of unit length (with respect to the Moebius metric g^*), such that JX=Y and JY=0. Replacing J by \|X\|^*J, if necessary, we can assume that also X has unit length. Let \alpha, \beta, \gamma\in C^{\infty}(M) be such that S_1X=\alpha X+\beta Y and S_1Y=\beta X+\gamma Y, so that S_1JX=\beta X+\gamma Y and S_1JY=0. From the symmetry of S_1J and the fact that \mbox{tr\,} S_1J=0 we obtain \beta=0=\gamma, and hence \alpha=\mbox{tr\,} S_1=0. Thus S_1=0, which is a contradiction.

Now assume that J^2=I, J\neq I. Let X,Y be a frame of unit vector fields (with respect to g^*) satisfying JX=X and JY=-Y. Write S_1X=\alpha X+\beta Y and S_1 Y=\gamma X+\delta Y for some \alpha, \beta, \gamma, \delta\in C^{\infty}(M), so that S_1JX=\alpha X+\beta Y and S_1JY=-\gamma X-\delta Y. Since \mbox{tr\,} S_1J=0=\mbox{tr\,} S_1, then \alpha=0=\delta. The symmetry of S_1 and S_2J implies that \beta=0=\gamma, which is again a contradiction.

Therefore, the only possible case is that J^2=-I. Let X,Y\in\Gamma(\Delta^\perp) be a frame of unit vector fields such that JX=Y and JY=-X. Write as before S_1X=\alpha X+\beta Y and S_1 Y=\gamma X+\delta Y for some \alpha, \beta, \gamma, \delta\in C^{\infty}(M). Then S_1JX=\gamma X+\delta Y and S_1JY=-\alpha X-\beta Y, hence \beta=\gamma, for \mbox{tr\,} S_1J=0. From the symmetry of S_1 we obtain

and similarly,

Comparing the two preceding equations, taking into account the symmetry of S_1J and the fact that \mbox{tr\,} S_1=0, yields \beta\<X,Y\>=0. If \beta is nonzero, then X and Y are orthogonal to each other. This is also the case if \beta, hence also \gamma, is zero, for in this case X and Y are eigenvectors of S_1. Thus, in any case, we conclude that J acts as a rotation of angle \pi/2 on \Delta^\perp.

Eq. (?) and the fact that \omega_i|_\Delta=0 imply that the splitting tensor of \Delta with respect to the Moebius metric satisfies

for all T\in\Gamma(\Delta) and 1\leq i\leq 2, where

for all X\in\Gamma(\Delta^\perp) and T\in\Gamma(\Delta). Here \nabla^{*h}_T X=(\nabla^*_TX)_{\Delta^\perp}. In particular,

Therefore

and hence

This implies that C^*_T commutes with J, and hence C^*_T\in\mbox{span}\{I,J\} for any T\in\Gamma(\Delta). It follows from (?) that also the splitting tensor C of \Delta corresponding to the metric induced by f satisfies C_T\in\mbox{span}\{I,J\} for any T\in\Gamma(\Delta). Moreover, by Lemma 7 and the assumption that f is not surface-like on any open subset, we see that C(\Gamma(\Delta))\not\subset \mbox{span}\{I\} on any open subset. Now, since J acts as a rotation of angle \pi/2 on \Delta^\perp, then \nabla_T^hJ=0. We conclude that f is elliptic with respect to J.

By Proposition 6, the central sphere congruence S\colon M^n\to \mathbb{S}_{1,1}^{n+2} of f is a two-parameter congruence of hyperspheres, which therefore gives rise to a surface s\colon L^2\to\mathbb{S}_{1,1}^{n+2} such that s=S\circ \pi, where \pi\colon M^n\to L^2 is the (local) quotient map onto the space of leaves of \Delta. Since \nabla_T^hJ=0=[C_T,J] for any T\in\Gamma(\Delta), it follows from Corollary 11.7 in [DT] that J is projectable with respect to \pi, that is, there exists \bar{J} \in \mbox{End}(TL) such that \bar{J}\circ \pi_*=\pi_*\circ J. In particular, the fact that J^2=-I implies that \bar{J}^2=-I, where we denote also by I the identity endomorphism of TL.

Now observe that, since f_2 shares with f_1 the same Moebius metric, its induced metric is conformal to the metric induced by f_1. Moreover, f_2 is not Moebius congruent to f_1 on any open subset of M^n and f_1 has a principal curvature of constant multiplicity (n-2). Thus f_1 is a so-called Cartan hypersurface. By the proof of the classification of Cartan hypersurfaces given in Chapter 17 of [DT] (see Lemma 17.4 therein), the surface s is elliptic with respect to \bar{J}, that is, for all \bar{X},\bar{Y}\in \mathfrak{X}(L) we have

We claim that \bar{J} is an orthogonal tensor, that is, it acts as a rotation of angle \pi/2 on each tangent space of L^2. The minimality of s will then follow from this, the fact that \bar{J}^2=-I and (?).

In order to show the orthogonality of \bar{J}, we use the fact that the metric \<\cdot,\cdot\>' on L^2 induced by s is related to the metric of M^n by

for all \bar{Z},\bar{W}\in\mathfrak{X}(L), where A is the shape operator of f, \lambda is the principal curvature of f having \Delta as its eigenbundle, which coincides with the mean curvature H of f by Proposition 6, and Z, W are the horizontal lifts of \bar{Z} and \bar{W}, respectively. Notice that (A-\lambda I) is a multiple of S_1. Since S_1J is symmetric, then also (A-\lambda I)J is symmetric. Therefore, given any \bar{X}\in\mathfrak{X}(L) and denoting by X\in\Gamma(\Delta^\perp) its horizontal lift, we have

where in the last step we have used that J acts as a rotation of angle \pi/2 on \Delta^\perp. Using again the symmetry of (A-\lambda I)J, the proof of the orthogonality of \bar{J} is completed by noticing that

Conversely, assume that the central sphere congruence of f\colon M^n\to\mathbb{R}^{n+1}, with M^n simply connected, is determined by a space-like minimal surface s\colon L^2\to \mathbb{S}_{1,1}^{n+2}. Let \bar{J}\in \Gamma(\mbox{End}(TL)) represent a rotation of angle \pi/2 on each tangent space. Then {\bar J}^2=-I and the second fundamental form of s satisfies (?) by the minimality of s. In particular, s is elliptic with respect to \bar{J}. By Lemma 17.4 in [DT], the hypersurface f is elliptic with respect to the lift J\in\Gamma(\mbox{End} (\Delta^\perp)) of \bar{J}, where \Delta is the eigenbundle correspondent to the principal curvature \lambda of f with multiplicity n-2, which coincides with its mean curvature. Therefore, the splitting tensor of \Delta satisfies C_T\in\mbox{span}\{I,J\} for any T\in\Gamma(\Delta). Since (A-\lambda I)C_T is symmetric for any T\in \Gamma(\Delta), as follows from the Codazzi equation, and C(\Gamma(\Delta))\not \subset \mbox{span}\{I\} on any open subset, for f is not conformally surface-like on any open subset, then (A-\lambda I)J is also symmetric.

By Theorem 17.5 in [DT], the set of conformal deformations of f is in one-to-one correspondence with the set of tensors \bar{D}\in \Gamma(\mbox{End}(TL)) with \det \bar{D}=1 that satisfy the Codazzi equation

for all \bar{X},\bar{Y}\in\mathfrak{X}(L), where \nabla' is the Levi-Civita connection of the metric induced by s. For a general elliptic hypersurface, this set either consists of a one-parameter family (continuous class) or of a single element (discrete class; see Section 11.2 and Exercise 11.3 in [DT]). The surface s is then said to be of the complex type of first or second species, respectively. For a minimal surface s\colon L^2\to \mathbb{S}_{1,1}^{n+2}, each tensor \bar{J}_\theta=\cos\theta I+\sin \theta \bar{J}, \theta\in [0,2\pi), satisfies both the condition \det \bar{J}_\theta=1 and the Codazzi equation, since it is a parallel tensor in L^2. Thus \{\bar{J}_\theta\}_{\theta\in [0,2\pi)} is the one-parameter family of tensors in L^2 having determinant one and satisfying the Codazzi equation. In particular, the surface s is of the complex type of first species. Therefore, the hypersurface f admits a one-parameter family of conformal deformations, each of which determined by one of the tensors \bar{J}_\theta\in \mbox{End}(TL), \theta\in [0,2\pi). The proof of Theorem (1) will be completed once we prove that any of such conformal deformations shares with f the same Moebius metric.

Let f_\theta\colon M^n\to\mathbb{R}^{n+1} be the conformal deformation of f determined by \bar{J}_\theta. Let F_\theta\colon M^n\to \mathbb{V}^{n+2} be the isometric light-cone representative of f_\theta, that is, F_\theta is the isometric immersion of M^n into the light-cone \mathbb{V}^{n+2}\subset \mathbb{L}^{n+3} given by F_\theta=\varphi_\theta^{-1}(\Psi\circ f_\theta), where \varphi_\theta is the conformal factor of the metric \<\cdot,\cdot\>_\theta induced by f_\theta with respect to the metric \<\cdot,\cdot\> of M^n, that is, \<\cdot,\cdot\>_\theta=\varphi_\theta^2\<\cdot,\cdot\>, and \Psi\colon \mathbb{R}^n\to \mathbb{V}^{n+2} is the isometric embedding of \mathbb{R}^n into \mathbb{V}^{n+2} given by (?). As shown in the proof of Lemma 17.2 in [DT], as part of the proof of the classification of Cartan hypersurfaces of dimension n\geq 5 given in Chapter 17 therein, the second fundamental form of F_\theta is given by

for all X,Y\in\mathfrak{X}(M), where \{\mu,\zeta,\bar{\zeta}\} is an orthonormal frame of the normal bundle of F_\theta in \mathbb{L}^{n+3} with \mu space-like, \lambda=-\<\mu, F_\theta\>^{-1} and \zeta=\lambda F_\theta+\mu (hence \<\zeta,\zeta\>=-1). Here J_\theta is the horizontal lift of \bar{J}_\theta, which has been extended to TM by setting {J_\theta}|_\Delta=I.

Let \bar{X},\bar{Y}\in\mathfrak{X}(L) be an orthonormal frame such that \bar{J}\bar{X}=\bar{Y} and \bar{J}\bar{Y}=-\bar{X}, and let X,Y\in\Gamma(\Delta^\perp) be the respective horizontal lifts. It follows from (?) that \{(A-\lambda I)X,(A-\lambda I)Y\} is an orthonormal frame of \Delta^\perp. From the symmetry of (A-\lambda I)J and (A-\lambda I) we have

In a similar way one verifies that \<J(A-\lambda I)Y,(A-\lambda I)Y\>=0 and

Thus J acts on \Delta^\perp as a rotation of angle \pi/2. The symmetry of both (A-\lambda I)J and (A-\lambda I) implies that \mbox{tr\,}(A-\lambda I)=0=\mbox{tr\,} (A-\lambda I)J, hence

for all \theta\in [0,2\pi).

Now we use the relation between the second fundamental forms of f_\theta and F_\theta, given by Eq. 9.32 in [DT], namely,

where \<\cdot,\cdot\>_\theta=\varphi_\theta^2\<\cdot,\cdot\> is the metric induced by f_\theta, A_\theta and H_\theta are its shape operator and mean curvature, respectively, \psi is a certain symmetric bilinear form, \tilde{N}\in \Gamma(N_FM), with \<\tilde{N}, F_\theta\>=0, is a unit space-like vector field, and \zeta_2\in \Gamma(N_FM) satisfies \<\tilde{N},\zeta_2\>=0=\<\zeta_2,\zeta_2\> and \<F_\theta,\zeta_2\>=1. Eqs. (?) and (?) give

for all X,Y\in\mathfrak{X}(M), or equivalently,

Using that

we obtain from the preceding equation that \<\zeta_2,\bar{\zeta}\>=0. Thus \bar{\zeta}\in \mbox{span}\{F_\theta,\zeta_2\}^\perp, and hence \bar{\zeta}=\pm\tilde{N}. Therefore, (?) reduces to

In particular, (A_\theta-H_\theta I)|_\Delta=0, hence also S_\theta|_\Delta=0, where S_\theta=\phi_\theta^{-1}(A_\theta-H_\theta I) is the Moebius shape operator of f_\theta, with \phi_\theta given by (?) for f_\theta. Since the Moebius shape operator of an umbilic-free immersion is traceless and has constant norm \sqrt{(n-1)/n}, then S_\theta must have constant eigenvalues \sqrt{(n-1)/2n}, -\sqrt{(n-1)/2n} and 0. The same holds for the Moebius second fundamental form S_1 of f, which has also \Delta as its kernel. We conclude that the eigenvalues of (A_\theta-H_\theta I)|_{\Delta^\perp} are

and, similarly, the eigenvalues of (A-\lambda I)|_{\Delta^\perp} are

where \phi_1 is given by (?) with respect to f. On the other hand, since

for \det J_\theta=1, and both (A-\lambda I) and (A-\lambda I)J_\theta are traceless (see (?)), it follows that (A-\lambda I) and (A-\lambda I)J_\theta have the same eigenvalues. This and (?) imply that

hence the Moebius metrics of f and f_\theta coincide. 5 pt 5 pt width4 pt height6 pt depth1.5 pt 1pt