Documentation

TauCeti.Analysis.InnerProductSpace.ShapeOperator

The shape operator of a map into a real inner product space #

Let i : V → E be a map from a finite-dimensional real normed space into a real inner product space whose derivative L = Di(x) is injective, so that g(v, w) = ⟪L v, L w⟫ is a metric on V, the metric induced by i. For a vector ν : E, the shape operator S_ν is the endomorphism of V determined by

g(S_ν v, w) = ⟪ν, B(v, w)⟫,

where B is the second fundamental form TauCeti.secondFundamentalForm ℝ i x. It depends linearly on ν, only through the normal component of ν, and it is g-self-adjoint for a C² map.

The shape operator controls the squared-distance function ρ_a(y) = ‖i y - a‖². At a critical point x, where a - i x is normal, the Hessian of ρ_a is 2 g((1 - S_{a - i x}) v, w), so its null space is the 1-eigenspace of S_{a - i x}. Along the normal line a = i x + t ν with t ≠ 0 this is the t⁻¹-eigenspace of S_ν: such a point i x + t ν is a focal point of i at x exactly when t⁻¹ is an eigenvalue of S_ν, with multiplicity the dimension of that eigenspace. A zero eigenvalue of S_ν gives no focal point, and i x itself (t = 0) is never one.

For a holomorphic immersion into a complex inner product space, with the real inner product re ⟪x, y⟫, the complex identities of the second fundamental form give

S_ν J = -J S_ν and S_{Jν} = J S_ν,

where J is multiplication by I. Hence J identifies the κ- and -κ-eigenspaces of S_ν and preserves its generalized 0-eigenspaces: the principal curvatures of a complex submanifold come in pairs ±κ with equal multiplicities, and on every normal line the focal points at distances t and -t have the same multiplicity.

Main definitions #

Main results #

References #

noncomputable def TauCeti.shapeOperator {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] (i : V → E) (x : V) (ν : E) :

The shape operator S_ν of i : V → E at x in the direction ν: the endomorphism of V with ⟪Di(x) (S_ν v), Di(x) w⟫ = ⟪ν, B(v, w)⟫ for all v and w, where B is the second fundamental form. It is defined when Di(x) is injective, so that the induced metric ⟪Di(x) v, Di(x) w⟫ is nondegenerate, and is 0 otherwise.

Instances For
    theorem TauCeti.inner_fderiv_shapeOperator_apply {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (h : Function.Injective ⇑(fderiv ℝ i x)) (ν : E) (v w : V) :
    inner ℝ ((fderiv ℝ i x) ((shapeOperator i x ν) v)) ((fderiv ℝ i x) w) = inner ℝ ν (((secondFundamentalForm ℝ i x) v) w)

    The defining identity of the shape operator. For an immersion, the induced metric pairs S_ν v with w as ν pairs with the second fundamental form B(v, w).

    The shape operator of a map whose derivative is not injective is 0 by convention.

    theorem TauCeti.eq_shapeOperator_iff {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (h : Function.Injective ⇑(fderiv ℝ i x)) {ν : E} {S : Module.End ℝ V} :
    S = shapeOperator i x ν ↔ ∀ (v w : V), inner ℝ ((fderiv ℝ i x) (S v)) ((fderiv ℝ i x) w) = inner ℝ ν (((secondFundamentalForm ℝ i x) v) w)

    Uniqueness of the shape operator. For an immersion, S_ν is the only endomorphism S with ⟪Di(x) (S v), Di(x) w⟫ = ⟪ν, B(v, w)⟫ for all v and w.

    @[simp]
    theorem TauCeti.shapeOperator_add {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (ν ν' : E) :
    shapeOperator i x (ν + ν') = shapeOperator i x ν + shapeOperator i x ν'

    The shape operator is additive in the normal direction.

    @[simp]
    theorem TauCeti.shapeOperator_smul {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (c : ℝ) (ν : E) :
    shapeOperator i x (c • ν) = c • shapeOperator i x ν

    The shape operator is homogeneous in the normal direction.

    @[simp]
    @[simp]
    theorem TauCeti.shapeOperator_neg {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (ν : E) :
    @[simp]
    theorem TauCeti.shapeOperator_sub {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (ν ν' : E) :
    shapeOperator i x (ν - ν') = shapeOperator i x ν - shapeOperator i x ν'
    theorem TauCeti.shapeOperator_of_mem_range {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} {ν : E} (hν : ν ∈ (↑(fderiv ℝ i x)).range) :
    shapeOperator i x ν = 0

    The shape operator depends only on the normal component of ν: tangent vectors have shape operator 0.

    theorem ContDiffAt.inner_fderiv_shapeOperator_apply_comm {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (hi : ContDiffAt ℝ 2 i x) (ν : E) (v w : V) :
    Inner.inner ℝ ((fderiv ℝ i x) ((TauCeti.shapeOperator i x ν) v)) ((fderiv ℝ i x) w) = Inner.inner ℝ ((fderiv ℝ i x) v) ((fderiv ℝ i x) ((TauCeti.shapeOperator i x ν) w))

    The shape operator of a C² map is self-adjoint for the induced metric: g(S_ν v, w) = g(v, S_ν w).

    Squared distance and focal points #

    theorem TauCeti.fderiv_fderiv_norm_sq_sub_comp_apply_eq_inner_shapeOperator {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} {a : E} (hi : ContDiffAt ℝ 2 i x) (h : Function.Injective ⇑(fderiv ℝ i x)) (ha : i x - a ∈ (↑(fderiv ℝ i x)).rangeᗮ) (v w : V) :
    ((fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - a‖ ^ 2) x) v) w = 2 * inner ℝ ((fderiv ℝ i x) (v - (shapeOperator i x (a - i x)) v)) ((fderiv ℝ i x) w)

    The Hessian of squared distance at a critical point. If i x - a is normal, the Hessian of y ↦ ‖i y - a‖² at x is 2 g((1 - S_{a - i x}) v, w), for the induced metric g.

    theorem TauCeti.ker_fderiv_fderiv_norm_sq_sub_comp {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} {a : E} (hi : ContDiffAt ℝ 2 i x) (h : Function.Injective ⇑(fderiv ℝ i x)) (ha : i x - a ∈ (↑(fderiv ℝ i x)).rangeᗮ) :
    (↑(fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - a‖ ^ 2) x)).ker = (shapeOperator i x (a - i x)).eigenspace 1

    The null space of the squared-distance Hessian. If i x - a is normal, the kernel of the Hessian of y ↦ ‖i y - a‖² at x is the 1-eigenspace of the shape operator S_{a - i x}. Its dimension is the multiplicity of a as a focal point of i at x.

    theorem TauCeti.ker_fderiv_fderiv_norm_sq_sub_add_smul_comp {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (hi : ContDiffAt ℝ 2 i x) (h : Function.Injective ⇑(fderiv ℝ i x)) {ν : E} (hν : ν ∈ (↑(fderiv ℝ i x)).rangeᗮ) {t : ℝ} (ht : t ≠ 0) :
    (↑(fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - (i x + t • ν)‖ ^ 2) x)).ker = (shapeOperator i x ν).eigenspace t⁻¹

    Focal points on a normal line. For a normal vector ν and t ≠ 0, the kernel of the Hessian at x of squared distance from i x + t • ν is the t⁻¹-eigenspace of S_ν. Thus i x + t • ν is a focal point of i at x exactly when t⁻¹ is an eigenvalue of S_ν, with multiplicity the dimension of that eigenspace.

    theorem TauCeti.ker_fderiv_fderiv_norm_sq_sub_self_comp {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℝ V] [FiniteDimensional ℝ V] [NormedAddCommGroup E] [InnerProductSpace ℝ E] {i : V → E} {x : V} (hi : ContDiffAt ℝ 2 i x) (h : Function.Injective ⇑(fderiv ℝ i x)) :
    (↑(fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - i x‖ ^ 2) x)).ker = ⊥

    The base point is not a focal point: squared distance from i x has nondegenerate Hessian at x along an immersion.

    Holomorphic immersions #

    The real inner product re ⟪x, y⟫ on a complex inner product space is enabled locally; the shape operators below are taken with respect to it.

    theorem ContDiffAt.shapeOperator_apply_I_smul {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℂ V] [FiniteDimensional ℂ V] [NormedAddCommGroup E] [InnerProductSpace ℂ E] {i : V → E} {x : V} (hi : ContDiffAt ℂ 2 i x) (ν : E) (v : V) :

    The shape operator anticommutes with J. For a C² holomorphic map, S_ν (I • v) = -(I • S_ν v).

    theorem ContDiffAt.shapeOperator_I_smul_apply {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℂ V] [FiniteDimensional ℂ V] [NormedAddCommGroup E] [InnerProductSpace ℂ E] {i : V → E} {x : V} (hi : ContDiffAt ℂ 2 i x) (ν : E) (v : V) :

    Rotating the normal vector by J. For a C² holomorphic map, S_{I • ν} v = I • S_ν v.

    Paired principal curvatures. For a C² holomorphic map, multiplication by I identifies the generalized κ- and -κ-eigenspaces of each shape operator, so they have the same dimension. For k = 1 these are the geometric multiplicities of the principal curvatures κ and -κ.

    For a C² holomorphic map, -κ is a principal curvature in the direction ν exactly when κ is.

    The zero principal curvature. For a C² holomorphic map, the generalized 0-eigenspaces of a shape operator, in particular its kernel, are complex subspaces.

    theorem ContDiffAt.finrank_ker_fderiv_fderiv_norm_sq_sub_sub_comp {V : Type u_1} {E : Type u_2} [NormedAddCommGroup V] [NormedSpace ℂ V] [FiniteDimensional ℂ V] [NormedAddCommGroup E] [InnerProductSpace ℂ E] {i : V → E} {x : V} (hi : ContDiffAt ℂ 2 i x) (h : Function.Injective ⇑(fderiv ℂ i x)) {ν : E} (hν : ν ∈ (↑(fderiv ℝ i x)).rangeᗮ) :
    Module.finrank ℝ ↥(↑(fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - (i x - ν)‖ ^ 2) x)).ker = Module.finrank ℝ ↥(↑(fderiv ℝ (fderiv ℝ fun (y : V) => ‖i y - (i x + ν)‖ ^ 2) x)).ker

    Symmetric focal points. For a holomorphic immersion and a normal vector ν, the points i x + ν and i x - ν have the same focal multiplicity, possibly zero: one is a focal point of i at x exactly when the other is, and the Hessians at x of squared distance from them have null spaces of the same dimension.