File size: 6,822 Bytes
4a56cd6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
/-
Copyright 2026 The Formal Conjectures Authors.
Released under Apache 2.0 license as described in the file LICENSE.
This file has been modified.
-/
import Mathlib.Analysis.Calculus.InverseFunctionTheorem.FDeriv
import Mathlib.Analysis.Calculus.FDeriv.Add
import Mathlib.Analysis.Calculus.FDeriv.Prod
import Mathlib.Analysis.Normed.Module.FiniteDimension

/-! # Split Derivative Normal Chart -/

namespace OAI.Analysis.Foundations.SplitDerivativeNormalChart

noncomputable section

open _root_.ContinuousLinearMap
open _root_.HasStrictFDerivAt

/-!
# Normal coordinates from a split derivative

A map with a split injective strict derivative extends to a local coordinate map
by adding vectors in the kernel of a left inverse. This is an application of the inverse
function theorem, not an assumed flattening equivalence. The zero-normal slice is exactly
the original parametrization on the coordinate domain. Identifying that slice
with an entire geometric support additionally requires a local embedding assertion.
-/

open _root_.Topology

namespace ContinuousLinearMap

variable {π•œ E F : Type*} [NontriviallyNormedField π•œ] [CompleteSpace π•œ]
  [NormedAddCommGroup E] [NormedSpace π•œ E]
  [NormedAddCommGroup F] [NormedSpace π•œ F]
  [FiniteDimensional π•œ E] [FiniteDimensional π•œ F]

/-- Let `E,F` be finite-dimensional normed vector spaces over a complete nontrivially normed field
`π•œ`. If continuous linear maps `A: E β†’ F` and `P: F β†’ E` satisfy `P ∘ A = id`, this continuous
linear equivalence `E Γ— ker(P) β‰… F` sends `(v,n)` to `A(v)+n`. Its inverse sends `y` to
`(P(y),y-A(P(y)))`. -/
def splitKernelEquiv (A : E β†’L[π•œ] F) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) : (E Γ— P.ker) ≃L[π•œ] F :=
  LinearEquiv.toContinuousLinearEquiv
    { toFun := fun z => A z.1 + z.2
      invFun := fun y => (P y, ⟨y - A (P y), by
        have hPA : P (A (P y)) = P y := DFunLike.congr_fun h (P y)
        change P (y - A (P y)) = 0
        rw [map_sub, hPA, sub_self]⟩)
      left_inv := by
        intro z
        have hPA : P (A z.1) = z.1 := DFunLike.congr_fun h z.1
        have hz : P (z.2 : F) = 0 := z.2.property
        ext <;> simp [hPA, hz]
      right_inv := by intro y; dsimp; abel
      map_add' := by intros; simp only [Prod.fst_add, Prod.snd_add, map_add,
        Submodule.coe_add]; abel
      map_smul' := by intros; simp }

@[simp] theorem splitKernelEquiv_apply (A : E β†’L[π•œ] F) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) (z : E Γ— P.ker) :
    (splitKernelEquiv A) P h z = A z.1 + z.2 := rfl

@[simp] theorem splitKernelEquiv_symm_fst (A : E β†’L[π•œ] F) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) (y : F) :
    (((splitKernelEquiv A) P h).symm y).1 = P y := rfl

end ContinuousLinearMap

namespace HasStrictFDerivAt

variable {π•œ E F : Type*} [NontriviallyNormedField π•œ] [CompleteSpace π•œ]
  [NormedAddCommGroup E] [NormedSpace π•œ E] [CompleteSpace E]
  [NormedAddCommGroup F] [NormedSpace π•œ F]
  [FiniteDimensional π•œ E] [FiniteDimensional π•œ F]
  {f : E β†’ F} {A : E β†’L[π•œ] F} {a : E}

local instance splitNormalKernelComplete (P : F β†’L[π•œ] E) : CompleteSpace P.ker :=
  FiniteDimensional.complete π•œ P.ker

omit [CompleteSpace E] in
/-- Adding normal vectors makes the split injective derivative invertible. -/
theorem add_kernel (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) :
    HasStrictFDerivAt (fun z : E Γ— P.ker => f z.1 + z.2)
      ((SplitDerivativeNormalChart.ContinuousLinearMap.splitKernelEquiv A) P h).toContinuousLinearMap (a, 0) := by
  have h1 := hf.comp (a, (0 : P.ker))
    (ContinuousLinearMap.fst π•œ E P.ker).hasStrictFDerivAt
  have h2 := P.ker.subtypeL.hasStrictFDerivAt.comp
    (a, (0 : P.ker)) (ContinuousLinearMap.snd π•œ E P.ker).hasStrictFDerivAt
  convert h1.add h2 using 1 <;> rfl

/-- Let `E,F` be finite-dimensional normed vector spaces over a complete nontrivially normed field
`π•œ`. Suppose `f: E β†’ F` has strict derivative `A` at `a`, with continuous linear left inverse
`P`. This local homeomorphism from `E Γ— ker(P)` to `F` is given near `(a,0)` by `(v,n) ↦
f(v)+n`. Its local inverse is supplied by the inverse function theorem. -/
def normalChart (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) :
    OpenPartialHomeomorph (E Γ— P.ker) F :=
  ((add_kernel hf) P h).toOpenPartialHomeomorph _

@[simp] theorem normalChart_apply (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) (z : E Γ— P.ker) :
    (normalChart hf) P h z = f z.1 + z.2 := rfl

/-- The selected point belongs to the coordinate domain. -/
theorem normalChart_mem_source (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) :
    (a, 0) ∈ ((normalChart hf) P h).source :=
  ((add_kernel hf) P h).mem_toOpenPartialHomeomorph_source

/-- The selected image point belongs to the ambient coordinate neighborhood. -/
theorem normalChart_mem_target (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) :
    f a ∈ ((normalChart hf) P h).target := by
  simpa using ((normalChart hf) P h).map_source ((normalChart_mem_source hf) P h)

/-- Inverse normal coordinates send every parametrized point in the coordinate domain
to precisely its tangent parameter and zero normal coordinate. -/
theorem normalChart_symm_apply (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) (v : E)
    (hv : (v, 0) ∈ ((normalChart hf) P h).source) :
    ((normalChart hf) P h).symm (f v) = (v, 0) := by
  simpa using ((normalChart hf) P h).left_inv hv

/-- On its ambient target, zero normal coordinate is equivalent to membership in the
image of the zero-normal part of the coordinate domain. -/
theorem normalChart_normal_eq_zero_iff (hf : HasStrictFDerivAt f A a) (P : F β†’L[π•œ] E)
    (h : P.comp A = ContinuousLinearMap.id π•œ E) (y : F)
    (hy : y ∈ ((normalChart hf) P h).target) :
    (((normalChart hf) P h).symm y).2 = 0 ↔
      βˆƒ v : E, (v, 0) ∈ ((normalChart hf) P h).source ∧ f v = y := by
  let e := (normalChart hf) P h
  constructor
  Β· intro hn
    change (e.symm y).2 = 0 at hn
    refine ⟨(e.symm y).1, ?_, ?_⟩
    Β· simpa only [← hn] using e.map_target hy
    Β· have hright := e.right_inv hy
      change f (e.symm y).1 + (e.symm y).2 = y at hright
      rw [hn] at hright
      simpa only [Submodule.coe_zero, add_zero] using hright
  · rintro ⟨v, hv, rfl⟩
    rw [(normalChart_symm_apply hf) P h v hv]

end HasStrictFDerivAt

end

end OAI.Analysis.Foundations.SplitDerivativeNormalChart