-
Notifications
You must be signed in to change notification settings - Fork 1.7k
Expand file tree
/
Copy pathGelfandFormula.lean
More file actions
226 lines (185 loc) · 11.1 KB
/
Copy pathGelfandFormula.lean
File metadata and controls
226 lines (185 loc) · 11.1 KB
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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
/-
Copyright (c) 2021 Jireh Loreaux. All rights reserved.
Released under Apache 2.0 license as described in the file LICENSE.
Authors: Jireh Loreaux
-/
module
public import Mathlib.Analysis.Normed.Algebra.SpectralRadiusLimit
public import Mathlib.Analysis.Normed.Algebra.Spectrum
public import Mathlib.Analysis.Calculus.Deriv.Basic
public import Mathlib.Analysis.Normed.Operator.Mul
import Mathlib.Analysis.Complex.Polynomial.Basic
import Mathlib.Analysis.Analytic.RadiusLiminf
/-!
# Gelfand's formula and other results on the spectrum in complex Banach algebras
This file contains results on the spectrum of elements in a complex Banach algebra, including
**Gelfand's formula** and the **Gelfand-Mazur theorem** and the fact that every element in a
complex Banach algebra has nonempty spectrum.
## Main results
* `spectrum.hasDerivAt_resolvent_const_left`: the resolvent function is differentiable on the
resolvent set.
* `spectrum.pow_nnnorm_pow_one_div_tendsto_nhds_spectralRadius`: Gelfand's formula for the
spectral radius in Banach algebras over `ℂ`.
* `spectrum.nonempty`: the spectrum of any element in a complex Banach algebra is nonempty.
* `NormedRing.algEquivComplexOfComplete`: **Gelfand-Mazur theorem** For a complex
Banach division algebra, the natural `algebraMap ℂ A` is an algebra isomorphism whose inverse
is given by selecting the (unique) element of `spectrum ℂ a`
## Implementation notes
Note that it is important here that the complex analysis files are privately imported, since the
material proven here gets used in contexts that have nothing to do with complex analysis
(i.e. C⋆-algebras, etc).
-/
@[expose] public section
variable {𝕜 A : Type*}
open scoped NNReal Topology Ring
open Filter ENNReal
namespace spectrum
section NonTriviallyNormedField
variable [NontriviallyNormedField 𝕜] [NormedRing A] [NormedAlgebra 𝕜 A] [CompleteSpace A]
theorem hasDerivAt_resolvent_const_left {a : A} {k : 𝕜} (hk : k ∈ resolventSet 𝕜 a) :
HasDerivAt (resolvent a) (-resolvent a k ^ 2) k := by
have H₁ : HasFDerivAt Ring.inverse _ (algebraMap 𝕜 A k - a) :=
hasFDerivAt_ringInverse (𝕜 := 𝕜) hk.unit
have H₂ : HasDerivAt (fun k => algebraMap 𝕜 A k - a) 1 k := by
simpa using! (Algebra.linearMap 𝕜 A).hasDerivAt.sub_const a
simpa [resolvent, sq, hk.unit_spec, ← Ring.inverse_unit hk.unit] using! H₁.comp_hasDerivAt k H₂
@[deprecated (since := "2026-03-26")]
alias hasDerivAt_resolvent := hasDerivAt_resolvent_const_left
theorem hasFDerivAt_resolvent {a : A} {k : 𝕜} (hk : k ∈ resolventSet 𝕜 a) :
HasFDerivAt (resolvent · k)
(((ContinuousLinearMap.mulLeftRight 𝕜 A) (resolvent a k)) (resolvent a k)) a := by
have H₁ : HasFDerivAt Ring.inverse _ (algebraMap 𝕜 A k - a) :=
hasFDerivAt_ringInverse (𝕜 := 𝕜) hk.unit
have H₂ : HasFDerivAt (fun a => algebraMap 𝕜 A k - a) (- .id 𝕜 A) a := by
simpa using! (hasFDerivAt_const _ a).sub (hasFDerivAt_id a)
simpa [resolvent_eq hk] using! H₁.comp a H₂
end NonTriviallyNormedField
theorem hasDerivAt_resolvent_const_right [NontriviallyNormedField 𝕜] [NontriviallyNormedField A]
[NormedAlgebra 𝕜 A] [CompleteSpace A] {a : A} {k : 𝕜} (hk : k ∈ resolventSet 𝕜 a) :
HasDerivAt (resolvent · k) (resolvent a k ^ 2) a := by
convert! hasFDerivAt_resolvent (𝕜 := A) hk |>.hasDerivAt
simp [resolvent, pow_two]
open ENNReal in
/-- In a Banach algebra `A` over `𝕜`, for `a : A` the function `fun z ↦ (1 - z • a)⁻¹` is
differentiable on any closed ball centered at zero of radius `r < (spectralRadius 𝕜 a)⁻¹`. -/
theorem differentiableOn_inverse_one_sub_smul [NontriviallyNormedField 𝕜] [NormedRing A]
[NormedAlgebra 𝕜 A] [CompleteSpace A] {a : A} {r : ℝ≥0}
(hr : (r : ℝ≥0∞) < (spectralRadius 𝕜 a)⁻¹) :
DifferentiableOn 𝕜 (fun z : 𝕜 => (1 - z • a)⁻¹ʳ) (Metric.closedBall 0 r) := by
intro z z_mem
apply DifferentiableAt.differentiableWithinAt
have hu : IsUnit (1 - z • a) := by
refine isUnit_one_sub_smul_of_lt_inv_radius (lt_of_le_of_lt (coe_mono ?_) hr)
simpa only [norm_toNNReal, Real.toNNReal_coe] using
Real.toNNReal_mono (mem_closedBall_zero_iff.mp z_mem)
have H₁ : Differentiable 𝕜 fun w : 𝕜 => 1 - w • a := (differentiable_id.smul_const a).const_sub 1
exact DifferentiableAt.comp z (differentiableAt_inverse hu) H₁.differentiableAt
section Complex
variable [NormedRing A] [NormedAlgebra ℂ A] [CompleteSpace A]
open ContinuousMultilinearMap in
/-- The `limsup` relationship for the spectral radius used to prove `spectrum.gelfand_formula`. -/
theorem limsup_pow_nnnorm_pow_one_div_le_spectralRadius (a : A) :
limsup (fun n : ℕ => (‖a ^ n‖₊ : ℝ≥0∞) ^ (1 / n : ℝ)) atTop ≤ spectralRadius ℂ a := by
refine ENNReal.inv_le_inv.mp (le_of_forall_pos_nnreal_lt fun r r_pos r_lt => ?_)
simp_rw [inv_limsup, ← one_div]
let p : FormalMultilinearSeries ℂ ℂ A := fun n =>
ContinuousMultilinearMap.mkPiRing ℂ (Fin n) (a ^ n)
suffices h : (r : ℝ≥0∞) ≤ p.radius by
convert! h
simp only [p, p.radius_eq_liminf, ← norm_toNNReal, norm_mkPiRing]
congr
ext n
rw [norm_toNNReal, ENNReal.coe_rpow_def ‖a ^ n‖₊ (1 / n : ℝ), ite_eq_right]
exact fun ha => (lt_self_iff_false _).mp
(ha.2.trans_le (one_div_nonneg.mpr n.cast_nonneg : 0 ≤ (1 / n : ℝ)))
have H₁ := (differentiableOn_inverse_one_sub_smul r_lt).hasFPowerSeriesOnBall r_pos
exact ((hasFPowerSeriesOnBall_inverse_one_sub_smul ℂ a).exchange_radius H₁).r_le
/-- **Gelfand's formula**: Given an element `a : A` of a complex Banach algebra, the
`spectralRadius` of `a` is the limit of the sequence `‖a ^ n‖₊ ^ (1 / n)`. -/
theorem pow_nnnorm_pow_one_div_tendsto_nhds_spectralRadius (a : A) :
Tendsto (fun n : ℕ => (‖a ^ n‖₊ : ℝ≥0∞) ^ (1 / n : ℝ)) atTop (𝓝 (spectralRadius ℂ a)) :=
tendsto_of_le_liminf_of_limsup_le (spectralRadius_le_liminf_pow_nnnorm_pow_one_div ℂ a)
(limsup_pow_nnnorm_pow_one_div_le_spectralRadius a)
alias gelfand_formula := pow_nnnorm_pow_one_div_tendsto_nhds_spectralRadius
/- This is the same as `pow_nnnorm_pow_one_div_tendsto_nhds_spectralRadius` but for `norm`
instead of `nnnorm`. -/
/-- **Gelfand's formula**: Given an element `a : A` of a complex Banach algebra, the
`spectralRadius` of `a` is the limit of the sequence `‖a ^ n‖ ^ (1 / n)`. -/
theorem pow_norm_pow_one_div_tendsto_nhds_spectralRadius (a : A) :
Tendsto (fun n : ℕ => ENNReal.ofReal (‖a ^ n‖ ^ (1 / n : ℝ))) atTop
(𝓝 (spectralRadius ℂ a)) := by
convert! pow_nnnorm_pow_one_div_tendsto_nhds_spectralRadius a using 1
ext1
rw [← ofReal_rpow_of_nonneg (norm_nonneg _) _, ← coe_nnnorm, coe_nnreal_eq]
simp
/-- **Gelfand's formula**: Given an element `a : A` of a complex Banach algebra, the
`spectralRadius` of `a` is the limit of the sequence `‖a ^ n‖ ^ (1 / n)`. -/
theorem spectralRadius_eq_spectralRadiusLimit (a : A) :
spectralRadius ℂ a = ENNReal.ofReal (spectralRadiusLim a) := by
have h1 := pow_norm_pow_one_div_tendsto_nhds_spectralRadius a
simp_rw [one_div] at h1
have h2 := continuous_ofReal.continuousAt.tendsto.comp (tendsto_spectralRadiusLim a)
exact tendsto_nhds_unique h1 h2
section Nontrivial
variable [Nontrivial A]
/-- In a (nontrivial) complex Banach algebra, every element has nonempty spectrum. -/
protected theorem nonempty (a : A) : (spectrum ℂ a).Nonempty := by
/- Suppose `σ a = ∅`, then resolvent set is `ℂ`, any `(z • 1 - a)` is a unit, and `resolvent a`
is differentiable on `ℂ`. -/
by_contra! h
have H₀ : resolventSet ℂ a = Set.univ := by rwa [spectrum, Set.compl_empty_iff] at h
have H₁ : Differentiable ℂ fun z : ℂ => resolvent a z := fun z =>
hasDerivAt_resolvent_const_left (H₀.symm ▸ Set.mem_univ z : z ∈ resolventSet ℂ a)
|>.differentiableAt
/- Since `resolvent a` tends to zero at infinity, by Liouville's theorem `resolvent a = 0`,
which contradicts that `resolvent a z` is invertible. -/
have H₃ := H₁.apply_eq_of_tendsto_cocompact 0 <| by
simpa [Metric.cobounded_eq_cocompact] using resolvent_tendsto_cobounded a (𝕜 := ℂ)
exact not_isUnit_zero <| H₃ ▸ (isUnit_resolvent.mp <| H₀.symm ▸ Set.mem_univ 0)
/-- In a complex Banach algebra, the spectral radius is always attained by some element of the
spectrum. -/
theorem exists_nnnorm_eq_spectralRadius (a : A) :
∃ z ∈ spectrum ℂ a, (‖z‖₊ : ℝ≥0∞) = spectralRadius ℂ a :=
exists_nnnorm_eq_spectralRadius_of_nonempty (spectrum.nonempty a)
/-- In a complex Banach algebra, if every element of the spectrum has norm strictly less than
`r : ℝ≥0`, then the spectral radius is also strictly less than `r`. -/
theorem spectralRadius_lt_of_forall_lt (a : A) {r : ℝ≥0}
(hr : ∀ z ∈ spectrum ℂ a, ‖z‖₊ < r) : spectralRadius ℂ a < r :=
spectralRadius_lt_of_forall_lt_of_nonempty (spectrum.nonempty a) hr
open Polynomial in
/-- The **spectral mapping theorem** for polynomials in a Banach algebra over `ℂ`. -/
theorem map_polynomial_aeval (a : A) (p : ℂ[X]) :
spectrum ℂ (aeval a p) = (fun k => eval k p) '' spectrum ℂ a :=
map_polynomial_aeval_of_nonempty a p (spectrum.nonempty a)
open Polynomial in
/-- A specialization of the spectral mapping theorem for polynomials in a Banach algebra over `ℂ`
to monic monomials. -/
protected theorem map_pow (a : A) (n : ℕ) :
spectrum ℂ (a ^ n) = (· ^ n) '' spectrum ℂ a := by
simpa only [aeval_X_pow, eval_X_pow] using map_polynomial_aeval a (X ^ n)
end Nontrivial
omit [CompleteSpace A] in
theorem algebraMap_eq_of_mem (hA : ∀ {a : A}, IsUnit a ↔ a ≠ 0) {a : A} {z : ℂ}
(h : z ∈ spectrum ℂ a) : algebraMap ℂ A z = a := by
rwa [mem_iff, hA, Classical.not_not, sub_eq_zero] at h
/-- **Gelfand-Mazur theorem**: For a complex Banach division algebra, the natural `algebraMap ℂ A`
is an algebra isomorphism whose inverse is given by selecting the (unique) element of
`spectrum ℂ a`. In addition, `algebraMap_isometry` guarantees this map is an isometry.
Note: because `NormedDivisionRing` requires the field `norm_mul : ∀ a b, ‖a * b‖ = ‖a‖ * ‖b‖`, we
don't use this type class and instead opt for a `NormedRing` in which the nonzero elements are
precisely the units. This allows for the application of this isomorphism in broader contexts, e.g.,
to the quotient of a complex Banach algebra by a maximal ideal. In the case when `A` is actually a
`NormedDivisionRing`, one may fill in the argument `hA` with the lemma `isUnit_iff_ne_zero`. -/
@[simps]
noncomputable def _root_.NormedRing.algEquivComplexOfComplete (hA : ∀ {a : A}, IsUnit a ↔ a ≠ 0) :
ℂ ≃ₐ[ℂ] A :=
let nt : Nontrivial A := ⟨⟨1, 0, hA.mp ⟨⟨1, 1, mul_one _, mul_one _⟩, rfl⟩⟩⟩
{ Algebra.ofId ℂ A with
toFun := algebraMap ℂ A
invFun := fun a => (@spectrum.nonempty _ _ _ _ nt a).some
left_inv := fun z => by
simpa only [@scalar_eq _ _ _ _ _ nt _] using!
(@spectrum.nonempty _ _ _ _ nt <| algebraMap ℂ A z).some_mem
right_inv := fun a => algebraMap_eq_of_mem (@hA) (@spectrum.nonempty _ _ _ _ nt a).some_mem }
end Complex
end spectrum