Appendix A
Geometry of Hypersurfaces
In this appendix we list standard definitions and well-known geometric facts for hypersurfaces in Rn+1. We consider only smooth embedded hypersurfaces M contained in some open set U ⊂ Rn+1. Let the embedding map be denoted by F : → Rn+1 with F() = M where ⊂ Rn is open. M is called a properly embedded hypersurface if F−1(K ) ⊂ is compact whenever K ⊂ U is compact. ∂ F The coordinate tangent vectors ∂i F(p) ≡ (p), 1 ≤ i ≤ n form a basis ∂pi of the tangent space Tx M at x = F(p) at every p ∈ . The metric on M is given by
gij = ∂i F · ∂ j F for 1 ≤ i, j ≤ n, the inverse metric by
ij −1 (g ) = (gij) and the area element of M by √ g = detgij. We are able to integrate compactly supported functions h : M → R over a properly embedded hypersurface. The integral is defined by h ≡ hdHn ≡ h(x) dHn(x) = h(F(p)) g(p) dp. M M M Here Hn denotes the n-dimensional Hausdorff measure on M (see Appendix C). Note in particular that Hn(M ∩ K )<∞
109
K. Ecker, Regularity Theory for Mean Curvature Flow © Springer Science+Business Media New York 2004 110 Regularity Theory for Mean Curvature Flow for any compact K ⊂ U. The tangential gradient of a function h : M → R (which we may think of as a function on via the embedding map) is defined by
M ij ∇ h = g ∂ j h ∂i F where we sum over repeated indices. For a smooth tangent vector field
i ij X = X ∂i F = g X j ∂i F on M (note that Xi = X · ∂i F) we define the covariant derivative tensor by ∇ M j = ∂ j + j k = jl ∂ − k i X i X ik X g i Xl il Xk k where the Christoffel symbols ij are given by (∂ ∂ )T = k ∂ i j F ij k F (the superscript T denotes the tangential component of a vector) or in terms of the metric by 1 k = gkl ∂ g + ∂ g − ∂ g . ij 2 i jl j il l ij The tangential divergence of X on M is defined by 1 √ div X =∇M X i = gij∇ M X = √ ∂ ggijX M i i j g i j and the Laplace–Beltrami operator of h on M by 1 √ h = div ∇ M h = gij ∂ ∂ h − k ∂ h = √ ∂ ggij∂ h . M M i j ij k g i j
For a smooth vector field X : M → Rn+1 which is not necessarily tangent to M we can also define the divergence with respect to M by
ij divM X = g ∂i X · ∂ j F.
This reduces to the previous expression for tangent vector fields. Let ν be a choice of unit normal field to M. In particular, this satisfies
ν · ∂i F = 0 Appendix A. Geometry of Hypersurfaces 111
on M for 1 ≤ i ≤ n. Note also that ∂i ν is a tangent vector field to M for 1 ≤ i ≤ n since ν has unit length. The second fundamental form of M is defined by
Aij = ∂i ν · ∂ j F =−ν · ∂i ∂ j F.
The eigenvalues κ1,...,κn of the Weingarten map (which is a map from the tangent space to itself) given by i = ik A j g Akj are called the principal curvatures of M. The mean curvature can then be expressed in various forms by