Chapter 05 — Fourier Series
Proofs omitted. Hover over highlighted terms to see their Lean types.
Reading sections
5.1 — Periodic Functions
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Definition 5.1.1: Let L > 0. A function f : ℝ → ℂ is periodic with period L if for every x : ℝ, one has f (x + L) = f x.
def IsPeriodicWithPeriod (L : ℝ) (f : ℝ → ℂ) : Prop :=
0 < L ∧ Function.Periodic f L
Definition 5.1.2 (ℤ-periodic and Lℤ-periodic): A function f : ℝ → ℂ is ℤ-periodic iff it is 1-periodic, i.e. f (x + 1) = f x for all x : ℝ.
def IsZPeriodic (f : ℝ → ℂ) : Prop :=
Function.Periodic f 1
A function is Lℤ-periodic when it is periodic with positive period L.
abbrev IsLZPeriodic (L : ℝ) (f : ℝ → ℂ) : Prop :=
IsPeriodicWithPeriod L f
Definition 5.1.3 (Integer part and fractional part): for x : ℝ, [x] is the unique integer k : ℤ such that k ≤ x < k + 1.
noncomputable def integerPart (x : ℝ) : ℤ :=
Int.floor x
The fractional part of x : ℝ, defined by {x} = x - [x].
noncomputable def fractionalPart (x : ℝ) : ℝ :=
x - (integerPart x : ℝ)
Example 5.1.2: sin x, cos x, and exp(ix) are 2π-periodic, hence periodic with period 2πk for every integer k; the identity function is not periodic with positive period, while the constant function 1 is L-periodic for every L > 0.
theorem periodicity_examples_basic :
IsPeriodicWithPeriod (2 * Real.pi) (fun x : ℝ => (Real.sin x : ℂ)) ∧
IsPeriodicWithPeriod (2 * Real.pi) (fun x : ℝ => (Real.cos x : ℂ)) ∧
IsPeriodicWithPeriod (2 * Real.pi) (fun x : ℝ => Complex.exp (x * Complex.I)) ∧
(∀ k : ℤ, Function.Periodic (fun x : ℝ => (Real.sin x : ℂ)) ((2 * Real.pi) * (k : ℝ))) ∧
(∀ k : ℤ, Function.Periodic (fun x : ℝ => (Real.cos x : ℂ)) ((2 * Real.pi) * (k : ℝ))) ∧
(∀ k : ℤ, Function.Periodic (fun x : ℝ => Complex.exp (x * Complex.I)) ((2 * Real.pi) * (k : ℝ))) ∧
(¬ ∃ L : ℝ, IsPeriodicWithPeriod L (fun x : ℝ => (x : ℂ))) ∧
(∀ L : ℝ, 0 < L → IsPeriodicWithPeriod L (fun _ : ℝ => (1 : ℂ)))
Corollary 5.1.1 (kL-shift invariance): If f : ℝ → ℂ is L-periodic, then f (x + kL) = f x for all x : ℝ and k : ℤ. In particular, if f is 1-periodic, then f (x + k) = f x for all k : ℤ.
theorem periodic_integer_shift_invariance :
(∀ {L : ℝ} {f : ℝ → ℂ}, IsPeriodicWithPeriod L f →
∀ x : ℝ, ∀ k : ℤ, f (x + (k : ℝ) * L) = f x) ∧
(∀ {f : ℝ → ℂ}, IsPeriodicWithPeriod 1 f →
∀ x : ℝ, ∀ k : ℤ, f (x + (k : ℝ)) = f x)
The square-wave function on ℝ, defined by its fractional part: value 1 on the first half of each unit interval and 0 on the second half.
noncomputable def squareWave : ℝ → ℂ :=
fun x => if x - (Int.floor x : ℝ) < (1 / 2 : ℝ) then (1 : ℂ) else 0
Example 5.1.4: For n : ℤ, the functions x ↦ cos(2πnx), x ↦ sin(2πnx), and x ↦ exp(2πinx) are all ℤ-periodic (equivalently 1-periodic). Another ℤ-periodic function is the square wave.
theorem zPeriodic_trig_modes_and_squareWave (n : ℤ) :
IsZPeriodic (fun x : ℝ => (Real.cos (2 * Real.pi * (n : ℝ) * x) : ℂ)) ∧
IsZPeriodic (fun x : ℝ => (Real.sin (2 * Real.pi * (n : ℝ) * x) : ℂ)) ∧
IsZPeriodic (fun x : ℝ => Complex.exp (2 * Real.pi * (n : ℝ) * x * Complex.I)) ∧
IsZPeriodic squareWave
Lemma 5.1.5a (Basic properties of C(ℝ/ℤ;ℂ) (a), Boundedness): if f : ℝ → ℂ is continuous and ℤ-periodic, then f is bounded.
lemma isZPeriodic_continuous_bounded
{f : ℝ → ℂ} (hf_cont : Continuous f) (hf_per : IsZPeriodic f) :
∃ C : ℝ, ∀ x : ℝ, ‖f x‖ ≤ C
Lemma 5.1.5b (Basic properties of C(ℝ/ℤ;ℂ) (b), vector space and algebra properties): if f, g : ℝ → ℂ are continuous and ℤ-periodic, then f + g, f - g, and f * g are continuous and ℤ-periodic; and for any scalar c : ℂ, c • f is continuous and ℤ-periodic.
lemma isZPeriodic_continuous_add_sub_mul_smul
{f g : ℝ → ℂ} (hf : Continuous f ∧ IsZPeriodic f) (hg : Continuous g ∧ IsZPeriodic g)
(c : ℂ) :
(Continuous (f + g) ∧ IsZPeriodic (f + g)) ∧
(Continuous (f - g) ∧ IsZPeriodic (f - g)) ∧
(Continuous (f * g) ∧ IsZPeriodic (f * g)) ∧
(Continuous (c • f) ∧ IsZPeriodic (c • f))
Lemma 5.1.5c (Basic properties of C(ℝ/ℤ;ℂ) (c), closure under uniform limits): if (fₙ) is a sequence of continuous ℤ-periodic functions converging uniformly to f, then f is continuous and ℤ-periodic.
lemma isZPeriodic_continuous_of_tendstoUniformly
{f : ℝ → ℂ} {u : ℕ → ℝ → ℂ}
(hcont : ∀ n : ℕ, Continuous (u n))
(hper : ∀ n : ℕ, IsZPeriodic (u n))
(hlim : TendstoUniformly u f Filter.atTop) :
Continuous f ∧ IsZPeriodic f
Helper for Example 5.1.2: periodicity is preserved when viewing a real-valued periodic function as complex-valued via coercion.
lemma helperForExample_5_1_2_realToComplex_periodic
{g : ℝ → ℝ} {T : ℝ} (hg : Function.Periodic g T) :
Function.Periodic (fun x : ℝ => (g x : ℂ)) T
Helper for Example 5.1.2: base 2π-periodicity for sin, cos, and x ↦ exp(ix) in the ℝ → ℂ setting.
lemma helperForExample_5_1_2_base_periodic_trig_exp :
Function.Periodic (fun x : ℝ => (Real.sin x : ℂ)) (2 * Real.pi) ∧
Function.Periodic (fun x : ℝ => (Real.cos x : ℂ)) (2 * Real.pi) ∧
Function.Periodic (fun x : ℝ => Complex.exp (x * Complex.I)) (2 * Real.pi)
Helper for Example 5.1.2: if f is L-periodic, then it is periodic with period Lk for every integer k.
lemma helperForExample_5_1_2_periodic_mul_int_right
{f : ℝ → ℂ} {L : ℝ} (hf : Function.Periodic f L) :
∀ k : ℤ, Function.Periodic f (L * (k : ℝ))
Helper for Example 5.1.2: the identity map x ↦ x (viewed as ℝ → ℂ) is not periodic with any positive period.
lemma helperForExample_5_1_2_identity_not_periodic_with_positive_period :
¬ ∃ L : ℝ, IsPeriodicWithPeriod L (fun x : ℝ => (x : ℂ))
Helper for Example 5.1.2: the constant function 1 is periodic with every positive period.
lemma helperForExample_5_1_2_constant_one_isPeriodicWithPeriod
(L : ℝ) (hL : 0 < L) : IsPeriodicWithPeriod L (fun _ : ℝ => (1 : ℂ))
Helper for Corollary 5.1.1: integer multiples of a period are also periods.
lemma helperForCorollary_5_1_1_intShift_ofPeriodic
{L : ℝ} {f : ℝ → ℂ} (hLper : Function.Periodic f L) :
∀ x : ℝ, ∀ k : ℤ, f (x + (k : ℝ) * L) = f x
Helper for Corollary 5.1.1: IsPeriodicWithPeriod implies integer-shift invariance by L.
lemma helperForCorollary_5_1_1_intShift_ofIsPeriodicWithPeriod
{L : ℝ} {f : ℝ → ℂ} (hIsPer : IsPeriodicWithPeriod L f) :
∀ x : ℝ, ∀ k : ℤ, f (x + (k : ℝ) * L) = f x
Helper for Corollary 5.1.1: specialization to unit period gives k-shift invariance.
lemma helperForCorollary_5_1_1_unitPeriodSpecialization
{f : ℝ → ℂ} (hIsPer1 : IsPeriodicWithPeriod 1 f) :
∀ x : ℝ, ∀ k : ℤ, f (x + (k : ℝ)) = f x
Helper for Lemma 5.1.5a: a continuous ℤ-periodic function has bounded range.
lemma helperForLemma_5_1_5a_boundedRange_of_continuous_periodic
{f : ℝ → ℂ} (hf_cont : Continuous f) (hf_per : IsZPeriodic f) :
Bornology.IsBounded (Set.range f)
Helper for Lemma 5.1.5a: bounded range gives a uniform bound on values.
lemma helperForLemma_5_1_5a_exists_uniform_norm_bound_of_boundedRange
{f : ℝ → ℂ} (hbounded : Bornology.IsBounded (Set.range f)) :
∃ C : ℝ, ∀ x : ℝ, ‖f x‖ ≤ C
Helper for Lemma 5.1.5b: closure of continuous ℤ-periodic functions under addition.
lemma helperForLemma_5_1_5b_add_closed
{f g : ℝ → ℂ} (hf : Continuous f ∧ IsZPeriodic f) (hg : Continuous g ∧ IsZPeriodic g) :
Continuous (f + g) ∧ IsZPeriodic (f + g)
Helper for Lemma 5.1.5b: closure of continuous ℤ-periodic functions under subtraction.
lemma helperForLemma_5_1_5b_sub_closed
{f g : ℝ → ℂ} (hf : Continuous f ∧ IsZPeriodic f) (hg : Continuous g ∧ IsZPeriodic g) :
Continuous (f - g) ∧ IsZPeriodic (f - g)
Helper for Lemma 5.1.5b: closure of continuous ℤ-periodic functions under multiplication.
lemma helperForLemma_5_1_5b_mul_closed
{f g : ℝ → ℂ} (hf : Continuous f ∧ IsZPeriodic f) (hg : Continuous g ∧ IsZPeriodic g) :
Continuous (f * g) ∧ IsZPeriodic (f * g)
Helper for Lemma 5.1.5b: closure of continuous ℤ-periodic functions under scalar multiplication.
lemma helperForLemma_5_1_5b_smul_closed
{f : ℝ → ℂ} (c : ℂ) (hf : Continuous f ∧ IsZPeriodic f) :
Continuous (c • f) ∧ IsZPeriodic (c • f)
Helper for Lemma 5.1.5c: a uniform limit of eventually continuous maps is continuous.
lemma helperForLemma_5_1_5c_continuous_of_uniform_limit
{f : ℝ → ℂ} {u : ℕ → ℝ → ℂ}
(hlim : TendstoUniformly u f Filter.atTop)
(hcont : ∀ n : ℕ, Continuous (u n)) :
Continuous f
Helper for Lemma 5.1.5c: periodicity of each approximant gives eventual equality of shifted evaluation sequences.
lemma helperForLemma_5_1_5c_eventuallyEq_shifted_sequence
{u : ℕ → ℝ → ℂ}
(hper : ∀ n : ℕ, IsZPeriodic (u n))
(x : ℝ) :
(fun n : ℕ => u n (x + 1)) =ᶠ[Filter.atTop] (fun n : ℕ => u n x)
Helper for Lemma 5.1.5c: the uniform limit inherits the unit-shift identity.
lemma helperForLemma_5_1_5c_shift_eq_of_uniform_limit
{f : ℝ → ℂ} {u : ℕ → ℝ → ℂ}
(hlim : TendstoUniformly u f Filter.atTop)
(hper : ∀ n : ℕ, IsZPeriodic (u n))
(x : ℝ) :
f (x + 1) = f x
Helper for Lemma 5.1.5c: a uniform limit of ℤ-periodic functions is ℤ-periodic.
lemma helperForLemma_5_1_5c_isZPeriodic_of_uniform_limit
{f : ℝ → ℂ} {u : ℕ → ℝ → ℂ}
(hlim : TendstoUniformly u f Filter.atTop)
(hper : ∀ n : ℕ, IsZPeriodic (u n)) :
IsZPeriodic f5.2 — Inner Products on Periodic Functions — Part 1
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Definition 5.2.1 (Inner product): for f, g ∈ C(ℝ/ℤ;ℂ) (modeled as continuous, ℤ-periodic functions ℝ → ℂ), define ⟪f, g⟫ = ∫ x in [0,1], f x * conj (g x) dx.
noncomputable def periodicInnerProduct
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : ℂ :=
∫ x in Set.Icc (0 : ℝ) 1, f.1 x * star (g.1 x)
Orthogonality for continuous, 1-periodic complex-valued functions.
def periodicOrthogonal
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : Prop :=
periodicInnerProduct f g = 0
The L^2 metric on continuous, 1-periodic complex-valued functions.
noncomputable def periodicL2Metric
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : ℝ :=
Real.sqrt (∫ x in Set.Icc (0 : ℝ) 1, ‖f.1 x - g.1 x‖ ^ (2 : ℕ))
Definition 5.2.2 (Orthogonality and the L^2 metric): for f, g ∈ C(ℝ/ℤ;ℂ) (modeled as continuous, 1-periodic functions ℝ → ℂ), (i) f and g are orthogonal iff ⟪f, g⟫ = 0; and (ii) d_{L^2}(f,g) = (∫_[0,1] ‖f(x) - g(x)‖^2 dx)^(1/2).
noncomputable def periodicOrthogonalityAndL2Metric :
({h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} →
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} → Prop) ×
({h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} →
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} → ℝ) :=
(periodicOrthogonal, periodicL2Metric)
Definition 5.2.3 (L^2-convergence): for a sequence (f_n) in C(ℝ/ℤ;ℂ) and f ∈ C(ℝ/ℤ;ℂ), (f_n) converges to f in L^2 when d_{L^2}(f_n, f) → 0 as n → ∞.
def periodicL2ConvergesTo
(fSeq : ℕ → {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : Prop :=
Filter.Tendsto (fun n => periodicL2Metric (fSeq n) f) Filter.atTop (nhds (0 : ℝ))
The constant function x ↦ 1 is continuous and 1-periodic.
lemma continuous_periodic_constOne :
Continuous (fun _ : ℝ => (1 : ℂ)) ∧ Function.Periodic (fun _ : ℝ => (1 : ℂ)) 1
The function x ↦ exp(2πix) is continuous and 1-periodic.
lemma continuous_periodic_expTwoPiI :
Continuous (fun x : ℝ => Complex.exp (2 * Real.pi * Complex.I * x)) ∧
Function.Periodic (fun x : ℝ => Complex.exp (2 * Real.pi * Complex.I * x)) 1
The 1-periodic function induced by the constant function x ↦ 1.
noncomputable def constOnePeriodicFunction :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
⟨fun _ : ℝ => (1 : ℂ), continuous_periodic_constOne⟩
The 1-periodic function induced by x ↦ exp(2πix).
noncomputable def expTwoPiIPeriodicFunction :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
⟨fun x : ℝ => Complex.exp (2 * Real.pi * Complex.I * x), continuous_periodic_expTwoPiI⟩
Example 5.2.3: let f(x) = 1 and g(x) = exp(2πix). Then ⟪f, g⟫ = 0.
theorem periodicInnerProduct_constOne_expTwoPiI :
periodicInnerProduct constOnePeriodicFunction expTwoPiIPeriodicFunction = 0
The L^2 norm on continuous, 1-periodic complex-valued functions induced by the inner product ⟪f, g⟫ = ∫_[0,1] f \overline{g}.
noncomputable def periodicL2Norm
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : ℝ :=
Real.sqrt (periodicInnerProduct f f).re
Example 5.2.6: for f(x) = exp(2πix), one has ‖f‖_2 = 1.
theorem expTwoPiIPeriodicFunction_l2Norm_eq_one :
periodicL2Norm expTwoPiIPeriodicFunction = 1
Lemma 5.2.7a (Basic properties of ‖·‖_2 (a), non-degeneracy): for f ∈ C(ℝ/ℤ; ℂ) (modeled as a continuous, 1-periodic function ℝ → ℂ), ‖f‖_2 = 0 iff f is the zero function.
lemma periodicL2Norm_eq_zero_iff
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicL2Norm f = 0 ↔ f.1 = 0
Lemma 5.2.7b (Basic properties of ‖·‖_2 (b), Cauchy--Schwarz): for f, g ∈ C(ℝ/ℤ; ℂ) (modeled as continuous, 1-periodic functions ℝ → ℂ), |⟪f, g⟫| ≤ ‖f‖_2 ‖g‖_2.
lemma periodicInnerProduct_abs_le_periodicL2Norm_mul
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
‖periodicInnerProduct f g‖ ≤ periodicL2Norm f * periodicL2Norm g
Lemma 5.2.5a (Properties of the inner product (a), Hermitian): if f, g ∈ C(ℝ/ℤ; ℂ) (modeled as continuous, 1-periodic functions ℝ → ℂ), then ⟪g, f⟫ = \overline{⟪f, g⟫}.
lemma periodicInnerProduct_hermitian
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicInnerProduct g f = star (periodicInnerProduct f g)
Lemma 5.2.5b (Properties of the inner product (b), Positivity): if f ∈ C(ℝ/ℤ; ℂ) (modeled as a continuous, 1-periodic function ℝ → ℂ), then ⟪f, f⟫ is real and nonnegative, and ⟪f, f⟫ = 0 iff f is the zero function.
lemma periodicInnerProduct_positivity
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(0 : ℝ) ≤ (periodicInnerProduct f f).re ∧
(periodicInnerProduct f f).im = 0 ∧
(periodicInnerProduct f f = 0 ↔ f.1 = 0)
The sum of two continuous, 1-periodic complex-valued functions is continuous and 1-periodic.
lemma continuous_periodic_add
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
Continuous (f.1 + g.1) ∧ Function.Periodic (f.1 + g.1) 1
A scalar multiple of a continuous, 1-periodic complex-valued function is continuous and 1-periodic.
lemma continuous_periodic_smul
(c : ℂ) (f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
Continuous (c • f.1) ∧ Function.Periodic (c • f.1) 1
The pointwise sum in C(ℝ/ℤ; ℂ) modeled as continuous, 1-periodic functions ℝ → ℂ.
noncomputable def addPeriodicFunction
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
⟨f.1 + g.1, continuous_periodic_add f g⟩
Scalar multiplication in C(ℝ/ℤ; ℂ) modeled as continuous, 1-periodic functions ℝ → ℂ.
noncomputable def smulPeriodicFunction
(c : ℂ) (f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
⟨c • f.1, continuous_periodic_smul c f⟩
Lemma 5.2.7c (Basic properties of ‖·‖_2 (c), triangle inequality): for f, g ∈ C(ℝ/ℤ; ℂ) (modeled as continuous, 1-periodic functions ℝ → ℂ), ‖f + g‖_2 ≤ ‖f‖_2 + ‖g‖_2.
lemma periodicL2Norm_add_le
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicL2Norm (addPeriodicFunction f g) ≤ periodicL2Norm f + periodicL2Norm g
Lemma 5.2.7d (Basic properties of ‖·‖_2 (d), Pythagoras): for f, g ∈ C(ℝ/ℤ; ℂ) (modeled as continuous, 1-periodic functions ℝ → ℂ), if ⟪f, g⟫ = 0, then ‖f + g‖_2^2 = ‖f‖_2^2 + ‖g‖_2^2.
lemma periodicL2Norm_pythagoras
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hfg : periodicInnerProduct f g = 0) :
periodicL2Norm (addPeriodicFunction f g) ^ (2 : ℕ) =
periodicL2Norm f ^ (2 : ℕ) + periodicL2Norm g ^ (2 : ℕ)
Lemma 5.2.7e (Basic properties of ‖·‖_2 (e), homogeneity): for f ∈ C(ℝ/ℤ; ℂ) (modeled as a continuous, 1-periodic function ℝ → ℂ) and c ∈ ℂ, ‖cf‖_2 = ‖c‖ ‖f‖_2.
lemma periodicL2Norm_smul
(c : ℂ) (f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicL2Norm (smulPeriodicFunction c f) = ‖c‖ * periodicL2Norm f
Helper for Example 5.2.3: rewrite the integrand as an exponential with negative coefficient.
lemma helperForExample_5_2_3_integrand_rewrite (x : ℝ) :
(1 : ℂ) * star (Complex.exp (2 * Real.pi * Complex.I * x)) =
Complex.exp ((-(2 * Real.pi * Complex.I)) * x)
Helper for Example 5.2.3: the exponential coefficient is nonzero.
lemma helperForExample_5_2_3_coeff_ne_zero :
((-(2 * Real.pi * Complex.I)) : ℂ) ≠ 0
Helper for Example 5.2.3: the endpoint value exp(-2πi) equals 1.
lemma helperForExample_5_2_3_exp_neg_twoPiI_eq_one :
Complex.exp (-(2 * Real.pi * Complex.I)) = 1
Helper for Example 5.2.3: the relevant exponential integral on [0,1] vanishes.
lemma helperForExample_5_2_3_interval_integral_exp_neg_twoPiI :
∫ x in Set.Icc (0 : ℝ) 1, Complex.exp ((-(2 * Real.pi * Complex.I)) * x) = 0
Helper for Example 5.2.6: the self-inner-product integrand is identically 1.
lemma helperForExample_5_2_6_pointwise_selfInner_integrand_eq_one (x : ℝ) :
Complex.exp (2 * Real.pi * Complex.I * x) *
star (Complex.exp (2 * Real.pi * Complex.I * x)) = (1 : ℂ)
Helper for Example 5.2.6: the self-inner product of exp(2πix) equals 1.
lemma helperForExample_5_2_6_innerProduct_self_eq_one :
periodicInnerProduct expTwoPiIPeriodicFunction expTwoPiIPeriodicFunction = (1 : ℂ)
Helper for Example 5.2.6: the real part of the self-inner product is 1.
lemma helperForExample_5_2_6_realpart_selfInner_eq_one :
(periodicInnerProduct expTwoPiIPeriodicFunction expTwoPiIPeriodicFunction).re = 1
Helper for Lemma 5.2.7a: rewrite Re ⟪f,f⟫ as an interval integral of normSq.
lemma helperForLemma_5_2_7a_re_eq_intervalIntegral_normSq
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct f f).re = ∫ x in (0 : ℝ)..1, Complex.normSq (f.1 x)
Helper for Lemma 5.2.7a: a nonzero periodic function is nonzero at some point of the unit interval.
lemma helperForLemma_5_2_7a_exists_nonzero_in_unit_interval
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hne : f.1 ≠ 0) :
∃ x ∈ Set.Icc (0 : ℝ) 1, f.1 x ≠ 0
Helper for Lemma 5.2.7a: if a periodic function is nonzero, then Re ⟪f,f⟫ is strictly positive.
lemma helperForLemma_5_2_7a_re_pos_of_nonzero
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hne : f.1 ≠ 0) :
0 < (periodicInnerProduct f f).re
Helper for Lemma 5.2.7a: vanishing real part of the self-inner product is equivalent to the function being zero.
lemma helperForLemma_5_2_7a_re_eq_zero_iff_fun_zero
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct f f).re = 0 ↔ f.1 = 0
Helper for Lemma 5.2.7a: ‖f‖_2 = 0 is equivalent to vanishing real part of the self-inner product.
lemma helperForLemma_5_2_7a_l2_eq_zero_iff_re_eq_zero
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicL2Norm f = 0 ↔ (periodicInnerProduct f f).re = 0
Helper for Lemma 5.2.7b: continuity on [0,1] gives L² membership on the restricted measure.
lemma helperForLemma_5_2_7b_memLp_two_restrict_Icc
(u : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
MeasureTheory.MemLp (fun x => u.1 x) 2
(MeasureTheory.volume.restrict (Set.Icc (0 : ℝ) 1))
Helper for Lemma 5.2.7b: pointwise norm of the inner-product integrand factors.
lemma helperForLemma_5_2_7b_pointwise_norm_mul_star
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
‖f.1 x * star (g.1 x)‖ = ‖f.1 x‖ * ‖g.1 x‖
Helper for Lemma 5.2.7b: Hölder with p = q = 2 on [0,1] for norm-products.
lemma helperForLemma_5_2_7b_holder_norm_on_Icc
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
∫ x in Set.Icc (0 : ℝ) 1, ‖f.1 x‖ * ‖g.1 x‖
≤ (∫ x in Set.Icc (0 : ℝ) 1, ‖f.1 x‖ ^ (2 : ℝ)) ^ (1 / (2 : ℝ))
* (∫ x in Set.Icc (0 : ℝ) 1, ‖g.1 x‖ ^ (2 : ℝ)) ^ (1 / (2 : ℝ))
Helper for Lemma 5.2.7b: each Hölder L² factor is exactly periodicL2Norm.
lemma helperForLemma_5_2_7b_rhs_factor_eq_periodicL2Norm
(u : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(∫ x in Set.Icc (0 : ℝ) 1, ‖u.1 x‖ ^ (2 : ℝ)) ^ (1 / (2 : ℝ)) = periodicL2Norm u
Helper for Lemma 5.2.5a: swapping the two arguments in the integrand gives its complex conjugate pointwise.
lemma helperForLemma_5_2_5a_pointwise_swap_to_conj
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
g.1 x * star (f.1 x) = star (f.1 x * star (g.1 x))
Helper for Lemma 5.2.5a: conjugation commutes with integration over [0,1].
lemma helperForLemma_5_2_5a_conj_of_setIntegral_Icc
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(∫ x in Set.Icc (0 : ℝ) 1, star (f.1 x * star (g.1 x))) =
star (∫ x in Set.Icc (0 : ℝ) 1, f.1 x * star (g.1 x))
Helper for Lemma 5.2.5b: rewrite ⟪f,f⟫ as the integral of normSq.
lemma helperForLemma_5_2_5b_self_eq_setIntegral_normSq
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicInnerProduct f f =
∫ x in Set.Icc (0 : ℝ) 1, ((Complex.normSq (f.1 x) : ℝ) : ℂ)
Helper for Lemma 5.2.5b: the real part of ⟪f,f⟫ is an interval integral of normSq.
lemma helperForLemma_5_2_5b_re_eq_intervalIntegral_normSq
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct f f).re = ∫ x in (0 : ℝ)..1, Complex.normSq (f.1 x)
Helper for Lemma 5.2.5b: a nonzero periodic function is nonzero at some point in one fundamental domain.
lemma helperForLemma_5_2_5b_exists_nonzero_in_unit_interval
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hne : f.1 ≠ 0) :
∃ x ∈ Set.Icc (0 : ℝ) 1, f.1 x ≠ 0
Helper for Lemma 5.2.5b: a nonzero value on [0,1] forces strict positivity of Re ⟪f,f⟫.
lemma helperForLemma_5_2_5b_re_pos_of_exists_nonzero_in_unit_interval
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hExists : ∃ x ∈ Set.Icc (0 : ℝ) 1, f.1 x ≠ 0) :
0 < (periodicInnerProduct f f).re
Helper for Lemma 5.2.5b: the imaginary part of ⟪f,f⟫ vanishes.
lemma helperForLemma_5_2_5b_self_im_eq_zero
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct f f).im = 0
Helper for Lemma 5.2.7c: pointwise expansion of (f + g) * conj (f + g).
lemma helperForLemma_5_2_7c_pointwise_add_self_integrand_expand
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
((f.1 + g.1) x) * star ((f.1 + g.1) x) =
f.1 x * star (f.1 x) + f.1 x * star (g.1 x) + g.1 x * star (f.1 x) +
g.1 x * star (g.1 x)
Helper for Lemma 5.2.7c: expansion of ⟪f + g, f + g⟫ into diagonal and cross terms.
lemma helperForLemma_5_2_7c_inner_add_self_expand
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicInnerProduct (addPeriodicFunction f g) (addPeriodicFunction f g) =
periodicInnerProduct f f + periodicInnerProduct f g +
periodicInnerProduct g f + periodicInnerProduct g g
Helper for Lemma 5.2.7c: diagonal terms satisfy Re ⟪u, u⟫ = ‖u‖_2^2.
lemma helperForLemma_5_2_7c_re_self_eq_sq_l2Norm
(u : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct u u).re = periodicL2Norm u ^ (2 : ℕ)
Helper for Lemma 5.2.7c: the real parts of the cross terms are bounded by 2 ‖f‖_2 ‖g‖_2.
lemma helperForLemma_5_2_7c_cross_re_sum_le_two_mul
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct f g).re + (periodicInnerProduct g f).re ≤
2 * (periodicL2Norm f * periodicL2Norm g)
Helper for Lemma 5.2.7c: the real part of ⟪f+g, f+g⟫ is bounded by (‖f‖_2 + ‖g‖_2)^2.
lemma helperForLemma_5_2_7c_re_addSelf_le_sq_sum
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct (addPeriodicFunction f g) (addPeriodicFunction f g)).re ≤
(periodicL2Norm f + periodicL2Norm g) ^ (2 : ℕ)
Helper for Lemma 5.2.7d: orthogonality is preserved when swapping arguments by Hermitian symmetry.
lemma helperForLemma_5_2_7d_swap_inner_eq_zero_of_orthogonal
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hfg : periodicInnerProduct f g = 0) :
periodicInnerProduct g f = 0
Helper for Lemma 5.2.7d: the sum of the real parts of the two cross terms vanishes under orthogonality.
lemma helperForLemma_5_2_7d_cross_re_sum_eq_zero_of_orthogonal
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hfg : periodicInnerProduct f g = 0) :
(periodicInnerProduct f g).re + (periodicInnerProduct g f).re = 0
Helper for Lemma 5.2.7d: the real part of ⟪f+g, f+g⟫ equals the sum of the squared L^2 norms when ⟪f, g⟫ = 0.
lemma helperForLemma_5_2_7d_re_addSelf_eq_sq_sum_of_orthogonal
(f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hfg : periodicInnerProduct f g = 0) :
(periodicInnerProduct (addPeriodicFunction f g) (addPeriodicFunction f g)).re =
periodicL2Norm f ^ (2 : ℕ) + periodicL2Norm g ^ (2 : ℕ)
Helper for Lemma 5.2.7e: pointwise scaling of normSq under scalar multiplication.
lemma helperForLemma_5_2_7e_pointwise_normSq_smul
(c : ℂ) (f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
Complex.normSq ((smulPeriodicFunction c f).1 x) =
(‖c‖ ^ (2 : ℕ)) * Complex.normSq (f.1 x)
Helper for Lemma 5.2.7e: nonnegativity of the real part of ⟪f,f⟫.
lemma helperForLemma_5_2_7e_re_self_nonneg
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
0 ≤ (periodicInnerProduct f f).re
Helper for Lemma 5.2.7e: the real part of ⟪cf,cf⟫ scales by ‖c‖².
lemma helperForLemma_5_2_7e_re_inner_self_smul
(c : ℂ) (f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
(periodicInnerProduct (smulPeriodicFunction c f) (smulPeriodicFunction c f)).re =
(‖c‖ ^ (2 : ℕ)) * (periodicInnerProduct f f).re
Helper for Lemma 5.2.7e: pull ‖c‖² out of a square root.
lemma helperForLemma_5_2_7e_sqrt_mul_sq_norm
(c : ℂ) {t : ℝ} (ht : 0 ≤ t) :
Real.sqrt ((‖c‖ ^ (2 : ℕ)) * t) = ‖c‖ * Real.sqrt t
Helper for Lemma 5.2.5c: the product integrand defining ⟪u, v⟫ is integrable on [0,1].
lemma helperForLemma_5_2_5c_integrableOn_mul_star
(u v : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
MeasureTheory.IntegrableOn (fun x : ℝ => u.1 x * star (v.1 x)) (Set.Icc (0 : ℝ) 1)
Helper for Lemma 5.2.5c: expand the additive integrand pointwise.
lemma helperForLemma_5_2_5c_addIntegrand_expand
(f g h : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
((f.1 + g.1) x) * star (h.1 x) =
f.1 x * star (h.1 x) + g.1 x * star (h.1 x)
Helper for Lemma 5.2.5c: factor a scalar from the first-slot integrand.
lemma helperForLemma_5_2_5c_smulIntegrand_factor
(c : ℂ) (f g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (x : ℝ) :
(c • f.1) x * star (g.1 x) = c * (f.1 x * star (g.1 x))5.2 — Inner Products on Periodic Functions — Part 2
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Lemma 5.2.5c (Properties of the inner product (c), linearity in the first variable): for f, g, h ∈ C(ℝ/ℤ; ℂ) and c ∈ ℂ, ⟪f + g, h⟫ = ⟪f, h⟫ + ⟪g, h⟫ and ⟪cf, g⟫ = c ⟪f, g⟫.
lemma periodicInnerProduct_linearity_first
(f g h : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (c : ℂ) :
periodicInnerProduct (addPeriodicFunction f g) h =
periodicInnerProduct f h + periodicInnerProduct g h ∧
periodicInnerProduct (smulPeriodicFunction c f) g = c * periodicInnerProduct f g
Lemma 5.2.5d (Properties of the inner product (d), antilinearity in the second variable): for f, g, h ∈ C(ℝ/ℤ; ℂ) and c ∈ ℂ, ⟪f, g + h⟫ = ⟪f, g⟫ + ⟪f, h⟫ and ⟪f, cg⟫ = \overline{c} ⟪f, g⟫.
lemma periodicInnerProduct_antilinearity_second
(f g h : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) (c : ℂ) :
periodicInnerProduct f (addPeriodicFunction g h) =
periodicInnerProduct f g + periodicInnerProduct f h ∧
periodicInnerProduct f (smulPeriodicFunction c g) = star c * periodicInnerProduct f g
The L^∞ norm on continuous, 1-periodic complex-valued functions, taken as the supremum of ‖f(x)‖ on [0,1].
noncomputable def periodicLInftyNorm
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) : ℝ :=
sSup (Set.range (fun x : Set.Icc (0 : ℝ) 1 => ‖f.1 x‖))
Exercise 5.2.3: if f ∈ C(ℝ/ℤ; ℂ) is non-zero, then 0 < ‖f‖_2 ≤ ‖f‖_{L^∞}; conversely, for real numbers 0 < A ≤ B, there exists a non-zero f ∈ C(ℝ/ℤ; ℂ) with ‖f‖_2 = A and ‖f‖_{L^∞} = B.
theorem exercise_5_2_3 :
(∀ f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
f.1 ≠ 0 →
0 < periodicL2Norm f ∧ periodicL2Norm f ≤ periodicLInftyNorm f) ∧
(∀ A B : ℝ, 0 < A → A ≤ B →
∃ f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
f.1 ≠ 0 ∧ periodicL2Norm f = A ∧ periodicLInftyNorm f = B)
Helper for Lemma 5.2.5d: taking the star of periodicInnerProduct u v swaps the arguments.
lemma helperForLemma_5_2_5d_star_swap
(u v : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
star (periodicInnerProduct u v) = periodicInnerProduct v u
Helper for Exercise 5.2.3: 0 belongs to the fundamental interval [0,1].
lemma helperForExercise_5_2_3_zero_mem_Icc :
(0 : ℝ) ∈ Set.Icc (0 : ℝ) 1
Helper for Exercise 5.2.3: the L^∞-range set is bounded above.
lemma helperForExercise_5_2_3_lInftyRange_bddAbove
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
BddAbove (Set.range (fun x : Set.Icc (0 : ℝ) 1 => ‖f.1 x‖))
Helper for Exercise 5.2.3: the L^∞-range set is nonempty.
lemma helperForExercise_5_2_3_lInftyRange_nonempty
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
Set.Nonempty (Set.range (fun x : Set.Icc (0 : ℝ) 1 => ‖f.1 x‖))
Helper for Exercise 5.2.3: the periodic L^∞ norm is nonnegative.
lemma helperForExercise_5_2_3_lInfty_nonneg
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
0 ≤ periodicLInftyNorm f
Helper for Exercise 5.2.3: pointwise normSq is bounded by ‖f‖^2_{L^∞} on [0,1].
lemma helperForExercise_5_2_3_normSq_le_lInftySq
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(x : ℝ) (hx : x ∈ Set.Icc (0 : ℝ) 1) :
Complex.normSq (f.1 x) ≤ periodicLInftyNorm f ^ (2 : ℕ)
Helper for Exercise 5.2.3: a nonzero periodic continuous function has strictly positive L^2 norm.
lemma helperForExercise_5_2_3_nonzero_implies_l2_pos
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1})
(hne : f.1 ≠ 0) :
0 < periodicL2Norm f
Helper for Exercise 5.2.3: always ‖f‖_2 ≤ ‖f‖_{L^∞}.
lemma helperForExercise_5_2_3_l2_le_lInfty
(f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}) :
periodicL2Norm f ≤ periodicLInftyNorm f
Helper for Exercise 5.2.3: for real scalars, ‖A‖ as a complex norm is |A|.
lemma helperForExercise_5_2_3_complexNorm_ofReal
(A : ℝ) :
‖(A : ℂ)‖ = |A|
Helper for Exercise 5.2.3: scaling exp(2πix) by c gives L^∞ norm ‖c‖.
lemma helperForExercise_5_2_3_lInfty_smul_exp
(c : ℂ) :
periodicLInftyNorm (smulPeriodicFunction c expTwoPiIPeriodicFunction) = ‖c‖
Helper for Exercise 5.2.3: diagonal converse case A = B.
lemma helperForExercise_5_2_3_exists_exact_norms_on_diagonal
(A : ℝ) (hA : 0 < A) :
∃ f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
f.1 ≠ 0 ∧ periodicL2Norm f = A ∧ periodicLInftyNorm f = A
Helper for Exercise 5.2.3: choose N,t with t^2 + (1 - t)^2 / N = r for any 0 < r < 1.
lemma helperForExercise_5_2_3_choose_parameters_for_ratio
(r : ℝ) (hr0 : 0 < r) (hr1 : r < 1) :
∃ N : ℕ, 1 ≤ N ∧ ∃ t : ℝ, 0 ≤ t ∧ t ≤ 1 ∧ t ^ (2 : ℕ) + (1 - t) ^ (2 : ℕ) / N = r
Helper for Exercise 5.2.3: the Fourier mode x ↦ exp(2πinx) is continuous and 1-periodic for every integer frequency n.
lemma continuous_periodic_expMode
(n : ℤ) :
Continuous (fun x : ℝ => Complex.exp (((n : ℂ) * (2 * Real.pi * Complex.I)) * (x : ℂ))) ∧
Function.Periodic (fun x : ℝ => Complex.exp (((n : ℂ) * (2 * Real.pi * Complex.I)) * (x : ℂ))) 1
Helper for Exercise 5.2.3: integer Fourier modes as periodic continuous functions.
noncomputable def expModePeriodicFunction
(n : ℤ) :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
⟨fun x : ℝ => Complex.exp (((n : ℂ) * (2 * Real.pi * Complex.I)) * (x : ℂ)),
continuous_periodic_expMode n⟩
Helper for Exercise 5.2.3: every Fourier mode has unit pointwise magnitude.
lemma helperForExercise_5_2_3_expMode_norm_eq_one
(n : ℤ) (x : ℝ) :
‖(expModePeriodicFunction n).1 x‖ = 1
Helper for Exercise 5.2.3: Fourier modes have orthonormal inner products on [0,1].
lemma helperForExercise_5_2_3_expMode_inner
(m n : ℤ) :
periodicInnerProduct (expModePeriodicFunction m) (expModePeriodicFunction n) =
if m = n then (1 : ℂ) else 0
Helper for Exercise 5.2.3: every Fourier mode has L^2 norm squared equal to 1.
lemma helperForExercise_5_2_3_expMode_l2Sq_eq_one
(n : ℤ) :
periodicL2Norm (expModePeriodicFunction n) ^ (2 : ℕ) = (1 : ℝ)
Helper for Exercise 5.2.3: partial sums of positive Fourier modes.
noncomputable def sumExpModes : ℕ → {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1}
Helper for Exercise 5.2.3: sumExpModes N is orthogonal to any higher mode.
lemma helperForExercise_5_2_3_sumExpModes_inner_of_lt
(N k : ℕ) (hk : N < k) :
periodicInnerProduct (sumExpModes N) (expModePeriodicFunction k) = 0
Helper for Exercise 5.2.3: sumExpModes N is orthogonal to the next mode.
lemma helperForExercise_5_2_3_sumExpModes_inner_next
(N : ℕ) :
periodicInnerProduct (sumExpModes N) (expModePeriodicFunction (N + 1)) = 0
Helper for Exercise 5.2.3: constant mode is orthogonal to sumExpModes N.
lemma helperForExercise_5_2_3_constMode_inner_sumExpModes
(N : ℕ) :
periodicInnerProduct (expModePeriodicFunction 0) (sumExpModes N) = 0
Helper for Exercise 5.2.3: sumExpModes N has exact L^2 norm square N.
lemma helperForExercise_5_2_3_sumExpModes_l2Sq
(N : ℕ) :
periodicL2Norm (sumExpModes N) ^ (2 : ℕ) = (N : ℝ)
Helper for Exercise 5.2.3: value of sumExpModes N at 0 is exactly N.
lemma helperForExercise_5_2_3_sumExpModes_value_zero
(N : ℕ) :
(sumExpModes N).1 (0 : ℝ) = (N : ℂ)
Helper for Exercise 5.2.3: pointwise bound for partial Fourier sums.
lemma helperForExercise_5_2_3_sumExpModes_norm_le
(N : ℕ) (x : ℝ) :
‖(sumExpModes N).1 x‖ ≤ (N : ℝ)
Helper for Exercise 5.2.3: normalized average of the first N positive Fourier modes.
noncomputable def averageExpModesPeriodicFunction
(N : ℕ) :
{h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1} :=
smulPeriodicFunction ((1 / (N : ℝ)) : ℂ) (sumExpModes N)
Helper for Exercise 5.2.3: exact profile of the normalized mode average, including orthogonality to the constant mode, exact L^2 square, exact L^∞, and value at 0.
lemma helperForExercise_5_2_3_average_modes_profile
(N : ℕ) (hN : 1 ≤ N) :
∃ g : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
periodicInnerProduct (expModePeriodicFunction 0) g = 0 ∧
periodicL2Norm g ^ (2 : ℕ) = (1 : ℝ) / N ∧
periodicLInftyNorm g = 1 ∧ g.1 0 = 1
Helper for Exercise 5.2.3: exact blended profile with prescribed t^2 + (1 - t)^2 / N, unit L^∞, and value 1 at zero.
lemma helperForExercise_5_2_3_blend_profile
(N : ℕ) (hN : 1 ≤ N) (t : ℝ) (ht0 : 0 ≤ t) (ht1 : t ≤ 1) :
∃ u : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
periodicLInftyNorm u = 1 ∧
periodicL2Norm u ^ (2 : ℕ) = t ^ (2 : ℕ) + (1 - t) ^ (2 : ℕ) / N ∧
u.1 0 = 1
Helper for Exercise 5.2.3: strict converse case A < B.
lemma helperForExercise_5_2_3_exists_exact_norms_strict
(A B : ℝ) (hA : 0 < A) (hAB : A < B) :
∃ f : {h : ℝ → ℂ // Continuous h ∧ Function.Periodic h 1},
f.1 ≠ 0 ∧ periodicL2Norm f = A ∧ periodicLInftyNorm f = B5.3 — Trigonometric Polynomials
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
The exponential mode x ↦ exp(2π i n x) is continuous and 1-periodic on ℝ.
lemma fourierCharacter_formula_continuous_periodic (n : ℤ) :
Continuous (fun x : ℝ => Complex.exp (2 * Real.pi * Complex.I * (n : ℂ) * (x : ℂ))) ∧
Function.Periodic (fun x : ℝ => Complex.exp (2 * Real.pi * Complex.I * (n : ℂ) * (x : ℂ))) 1
Definition 5.3.1 (Characters): for every integer n, define e_n ∈ C(ℝ/ℤ; ℂ) by e_n(x) = exp(2π i n x).
noncomputable def fourierCharacter (n : ℤ) :
C(AddCircle (1 : ℝ), ℂ) :=
fourier (T := (1 : ℝ)) n
Evaluation formula for the character e_n.
lemma fourierCharacter_apply (n : ℤ) (x : ℝ) :
fourierCharacter n (x : AddCircle (1 : ℝ)) =
Complex.exp (2 * Real.pi * Complex.I * (n : ℂ) * (x : ℂ))
Lemma 5.3.5 (Characters are an orthonormal system): for integers n, m, the Fourier characters satisfy ⟪e_n, e_m⟫ = 1 when n = m and 0 otherwise, and ‖e_n‖_2 = 1, where e_k is viewed in L² as fourierLp (T := 1) (p := 2) k.
lemma fourierCharacter_orthonormal_system (n m : ℤ) :
(inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n) (fourierLp (T := (1 : ℝ)) (p := 2) m) =
(if n = m then (1 : ℂ) else 0)) ∧
‖fourierLp (T := (1 : ℝ)) (p := 2) n‖ = (1 : ℝ)
Definition 5.3.2 (Trigonometric polynomials): a function f ∈ C(ℝ/ℤ; ℂ) is a trigonometric polynomial if f = ∑_{n=-N}^{N} c_n e_n for some integer bound N ≥ 0 and coefficients c_n ∈ ℂ.
def IsTrigonometricPolynomial (f : C(AddCircle (1 : ℝ), ℂ)) : Prop :=
∃ N : ℕ, ∃ c : ℤ → ℂ,
f = Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ)) (fun n => c n • fourierCharacter n)
Corollary 5.3.6: for a trigonometric polynomial f = ∑_{n=-N}^{N} c_n e_n, the Fourier coefficient at each n in the support interval is c_n, the coefficients outside that interval vanish, and ‖f‖_2^2 = ∑_{n=-N}^{N} |c_n|^2.
theorem trigonometricPolynomial_fourierCoefficients_and_normSq (N : ℕ) (c : ℤ → ℂ) :
let f :=
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => c n • fourierLp (T := (1 : ℝ)) (p := 2) n)
(∀ n : ℤ,
n ∈ Finset.Icc (-(N : ℤ)) (N : ℤ) →
inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n) f = c n) ∧
(∀ n : ℤ,
n ∉ Finset.Icc (-(N : ℤ)) (N : ℤ) →
inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n) f = 0) ∧
‖f‖ ^ (2 : ℕ) =
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ)) (fun n => ‖c n‖ ^ (2 : ℕ))
Definition 5.3.7 (Fourier transform): for f ∈ C(ℝ/ℤ; ℂ), its Fourier transform is the function ℤ → ℂ sending n to the n-th Fourier coefficient ∫ x in (0 : ℝ)..1, f x * exp (-2π i n x).
noncomputable def fourierTransform (f : C(AddCircle (1 : ℝ), ℂ)) : ℤ → ℂ :=
fourierCoeff f
For f ∈ C(ℝ/ℤ; ℂ), the Fourier series term family is n ↦ \hat f(n) e_n.
noncomputable def fourierSeriesOnAddCircle (f : C(AddCircle (1 : ℝ), ℂ)) :
ℤ → C(AddCircle (1 : ℝ), ℂ) :=
fun n => (fourierTransform f n) • fourierCharacter n
Definition 5.3.8 (Fourier inversion, Fourier series, and Plancherel formula): for f ∈ C(ℝ/ℤ; ℂ), the Fourier series is the formal family n ↦ \hat f(n) e_n; if f is a trigonometric polynomial, then f equals a finite sum of these Fourier-series terms; and in the same finite case one has the Parseval/Plancherel identity.
def fourierSeries_inversion_plancherel_onAddCircle
(f : C(AddCircle (1 : ℝ), ℂ)) : Prop
Bridge fourierTransform to the L² inner product with Fourier modes.
lemma fourierTransform_eq_inner_fourierLp_toLp
(f : C(AddCircle (1 : ℝ), ℂ)) (n : ℤ) :
fourierTransform f n =
inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n)
(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f)
If f is a trigonometric polynomial, then f equals a finite sum of its Fourier-series terms (Fourier inversion in the finite case).
theorem trigonometricPolynomial_fourierInversion
(f : C(AddCircle (1 : ℝ), ℂ)) (hf : IsTrigonometricPolynomial f) :
∃ s : Finset ℤ,
(∀ n : ℤ, n ∉ s → fourierTransform f n = 0) ∧
f = Finset.sum s (fun n => fourierSeriesOnAddCircle f n)
If f is a trigonometric polynomial, then the finite Parseval/Plancherel identity holds: the L² norm identity on AddCircle equals a finite sum of squared Fourier coefficients.
theorem trigonometricPolynomial_plancherel_formula
(f : C(AddCircle (1 : ℝ), ℂ)) (hf : IsTrigonometricPolynomial f) :
∃ s : Finset ℤ,
(∀ n : ℤ, n ∉ s → fourierTransform f n = 0) ∧
(∫ x : AddCircle (1 : ℝ), ‖f x‖ ^ (2 : ℕ)
∂(AddCircle.haarAddCircle : MeasureTheory.Measure (AddCircle (1 : ℝ)))) =
Finset.sum s (fun n => ‖fourierTransform f n‖ ^ (2 : ℕ))
Helper for Corollary 5.3.6: coefficient extraction on a finite support set.
lemma helperForCorollary_5_3_6_inner_on_support
(s : Finset ℤ) (c : ℤ → ℂ) {n : ℤ} (hn : n ∈ s) :
inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n)
(Finset.sum s (fun k => c k • fourierLp (T := (1 : ℝ)) (p := 2) k)) = c n
Helper for Corollary 5.3.6: coefficient vanishing outside a finite support set.
lemma helperForCorollary_5_3_6_inner_off_support
(s : Finset ℤ) (c : ℤ → ℂ) {n : ℤ} (hn : n ∉ s) :
inner ℂ (fourierLp (T := (1 : ℝ)) (p := 2) n)
(Finset.sum s (fun k => c k • fourierLp (T := (1 : ℝ)) (p := 2) k)) = 0
Helper for Corollary 5.3.6: finite Parseval identity for a finite Fourier sum.
lemma helperForCorollary_5_3_6_normSq_finite_sum
(s : Finset ℤ) (c : ℤ → ℂ) :
‖Finset.sum s (fun k => c k • fourierLp (T := (1 : ℝ)) (p := 2) k)‖ ^ (2 : ℕ) =
Finset.sum s (fun k => ‖c k‖ ^ (2 : ℕ))5.4 — Periodic Convolutions — Part 1
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Theorem 5.4.1 (Weierstrass approximation theorem for trigonometric polynomials): for every continuous function f : C(ℝ/ℤ; ℂ) and every ε > 0, there exists a trigonometric polynomial P such that ‖f - P‖ ≤ ε.
theorem exists_trigonometricPolynomial_uniform_approximation
(f : C(AddCircle (1 : ℝ), ℂ)) (ε : ℝ) (hε : 0 < ε) :
∃ P : C(AddCircle (1 : ℝ), ℂ),
IsTrigonometricPolynomial P ∧ ‖f - P‖ ≤ ε
Continuity of the function underlying periodic convolution.
lemma periodicConvolution_continuous
(f g : C(AddCircle (1 : ℝ), ℂ)) :
Continuous (fun x : AddCircle (1 : ℝ) =>
∫ y in Set.Icc (0 : ℝ) 1, f (y : AddCircle (1 : ℝ)) * g (x - (y : AddCircle (1 : ℝ))))
Definition 5.4.2: for continuous f, g : C(ℝ/ℤ; ℂ), the periodic convolution is the continuous function on ℝ/ℤ given by (f * g)(x) = ∫ y ∈ [0,1], f(y) g(x - y) dy with y viewed in ℝ/ℤ.
noncomputable def periodicConvolution (f g : C(AddCircle (1 : ℝ), ℂ)) : C(AddCircle (1 : ℝ), ℂ) :=
⟨(fun x : AddCircle (1 : ℝ) =>
∫ y in Set.Icc (0 : ℝ) 1, f (y : AddCircle (1 : ℝ)) * g (x - (y : AddCircle (1 : ℝ)))),
periodicConvolution_continuous f g⟩
Lemma 5.4.4a (Basic properties of periodic convolution (a), Closure): if f, g ∈ C(ℝ/ℤ; ℂ), then f * g ∈ C(ℝ/ℤ; ℂ). In this file, f * g is represented by periodicConvolution f g.
lemma periodicConvolution_continuous_closure
(f g : C(AddCircle (1 : ℝ), ℂ)) :
Continuous (periodicConvolution f g)
Lemma 5.4.4b (Basic properties of periodic convolution (b), Commutativity): for f, g : C(ℝ/ℤ; ℂ), one has f * g = g * f, represented here as periodicConvolution f g = periodicConvolution g f.
lemma periodicConvolution_comm
(f g : C(AddCircle (1 : ℝ), ℂ)) :
periodicConvolution f g = periodicConvolution g f
Lemma 5.4.4c (Basic properties of periodic convolution (c), Bilinearity): for f, g, h : C(ℝ/ℤ; ℂ) and c : ℂ, one has f * (g + h) = f * g + f * h, (f + g) * h = f * h + g * h, and c (f * g) = (c f) * g = f * (c g), represented here using periodicConvolution.
lemma periodicConvolution_bilinearity
(f g h : C(AddCircle (1 : ℝ), ℂ)) (c : ℂ) :
periodicConvolution f (g + h) = periodicConvolution f g + periodicConvolution f h ∧
periodicConvolution (f + g) h = periodicConvolution f h + periodicConvolution g h ∧
c • periodicConvolution f g = periodicConvolution (c • f) g ∧
periodicConvolution (c • f) g = periodicConvolution f (c • g)
Lemma 5.4.5: for f ∈ C(ℝ/ℤ; ℂ) and n ∈ ℤ, one has f * e_n = \hat f(n)e_n. More generally, for a trigonometric polynomial P = ∑_{k=-N}^{N} c_k e_k, periodic convolution acts termwise: f * P = ∑_{k=-N}^{N} \hat f(k)c_k e_k, hence f * P is trigonometric.
lemma periodicConvolution_fourierCharacter_and_trigSum
(f : C(AddCircle (1 : ℝ), ℂ)) (n : ℤ) (N : ℕ) (c : ℤ → ℂ) :
periodicConvolution f (fourierCharacter n) = (fourierTransform f n) • fourierCharacter n ∧
periodicConvolution f
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ)) (fun k => c k • fourierCharacter k))
=
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k => (fourierTransform f k * c k) • fourierCharacter k) ∧
IsTrigonometricPolynomial
(periodicConvolution f
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ)) (fun k => c k • fourierCharacter k)))
Definition 5.4.5 (Periodic approximation to the identity): for ε > 0 and 0 < δ < 1/2, a continuous periodic function f : C(ℝ/ℤ; ℂ) is a periodic (ε, δ) approximation to the identity iff (a) f(x) is real and nonnegative for all x, and ∫_[0,1] f = 1, and (b) |f(x)| < ε whenever δ ≤ |x| ≤ 1 - δ.
def IsPeriodicApproximationToIdentity
(ε δ : ℝ) (f : C(AddCircle (1 : ℝ), ℂ)) : Prop :=
0 < ε ∧
0 < δ ∧
δ < (1 / 2 : ℝ) ∧
(∀ x : AddCircle (1 : ℝ), (f x).im = 0 ∧ 0 ≤ (f x).re) ∧
(∫ y in Set.Icc (0 : ℝ) 1, f (y : AddCircle (1 : ℝ))) = (1 : ℂ) ∧
∀ x : ℝ, δ ≤ |x| → |x| ≤ 1 - δ → ‖f (x : AddCircle (1 : ℝ))‖ < ε
Helper for Theorem 5.4.1: extract a Fourier-span element that approximates a target continuous function within a prescribed sup-norm error.
lemma helperForTheorem_5_4_1_dense_span_witness
(f : C(AddCircle (1 : ℝ), ℂ)) (ε : ℝ) (hε : 0 < ε) :
∃ P : C(AddCircle (1 : ℝ), ℂ),
P ∈ Submodule.span ℂ (Set.range (fourier (T := (1 : ℝ)))) ∧ ‖f - P‖ < ε
Helper for Theorem 5.4.1: every element of the Fourier span has a finite Finsupp expansion in the characters fourierCharacter n.
lemma helperForTheorem_5_4_1_span_element_has_finsupp_expansion
{P : C(AddCircle (1 : ℝ), ℂ)}
(hP : P ∈ Submodule.span ℂ (Set.range (fourier (T := (1 : ℝ))))) :
∃ c : ℤ →₀ ℂ, P = c.sum (fun n a => a • fourierCharacter n)
Helper for Theorem 5.4.1: the support of a finitely supported coefficient family is contained in a symmetric interval [-N, N], with N = c.support.sup Int.natAbs.
lemma helperForTheorem_5_4_1_support_in_symmetric_Icc
(c : ℤ →₀ ℂ) :
let N : ℕ := c.support.sup Int.natAbs
∀ n ∈ c.support, n ∈ Finset.Icc (-(N : ℤ)) (N : ℤ)
Helper for Theorem 5.4.1: every finite Fourier-character expansion is a trigonometric polynomial in the sense of Definition 5.3.2.
lemma helperForTheorem_5_4_1_isTrig_of_finsupp_expansion
(c : ℤ →₀ ℂ) :
IsTrigonometricPolynomial (c.sum (fun n a => a • fourierCharacter n))
Helper for Lemma 5.4.4b: rewrite the periodic-convolution integral on Set.Icc as an evaluation of convolution on AddCircle (1 : ℝ).
lemma helperForLemma_5_4_4b_eval_eq_addcircle_convolution
(f g : C(AddCircle (1 : ℝ), ℂ)) (x : AddCircle (1 : ℝ)) :
(∫ y in Set.Icc (0 : ℝ) 1, f (y : AddCircle (1 : ℝ)) * g (x - (y : AddCircle (1 : ℝ)))) =
MeasureTheory.convolution (fun t : AddCircle (1 : ℝ) => f t)
(fun t : AddCircle (1 : ℝ) => g t)
(ContinuousLinearMap.mul ℂ ℂ)
(MeasureTheory.volume : MeasureTheory.Measure (AddCircle (1 : ℝ))) x
Helper for Lemma 5.4.4b: for complex-valued convolution, replacing mul by mul.flip does not change the result.
lemma helperForLemma_5_4_4b_mul_convolution_eq_flip_convolution
(F G : AddCircle (1 : ℝ) → ℂ) :
MeasureTheory.convolution G F (ContinuousLinearMap.mul ℂ ℂ)
(MeasureTheory.volume : MeasureTheory.Measure (AddCircle (1 : ℝ))) =
MeasureTheory.convolution G F (ContinuousLinearMap.mul ℂ ℂ).flip
(MeasureTheory.volume : MeasureTheory.Measure (AddCircle (1 : ℝ)))
Helper for Lemma 5.4.4b: convolution on AddCircle (1 : ℝ) with complex multiplication is commutative.
lemma helperForLemma_5_4_4b_addcircle_convolution_comm
(F G : AddCircle (1 : ℝ) → ℂ) :
MeasureTheory.convolution F G (ContinuousLinearMap.mul ℂ ℂ)
(MeasureTheory.volume : MeasureTheory.Measure (AddCircle (1 : ℝ))) =
MeasureTheory.convolution G F (ContinuousLinearMap.mul ℂ ℂ)
(MeasureTheory.volume : MeasureTheory.Measure (AddCircle (1 : ℝ)))
Helper for Lemma 5.4.5: factor a character at a difference into a product of an x-factor and a y-factor.
lemma helperForLemma_5_4_5_character_sub_factorization
(n : ℤ) (x y : AddCircle (1 : ℝ)) :
fourierCharacter n (x - y) = fourierCharacter n x * fourierCharacter (-n) y
Helper for Lemma 5.4.5: identify the interval integral appearing in the character-convolution computation with the Fourier coefficient.
lemma helperForLemma_5_4_5_intervalIntegral_eq_fourierTransform
(f : C(AddCircle (1 : ℝ), ℂ)) (n : ℤ) :
(∫ y in Set.Icc (0 : ℝ) 1,
f (y : AddCircle (1 : ℝ)) * fourierCharacter (-n) (y : AddCircle (1 : ℝ)))
= fourierTransform f n
Helper for Lemma 5.4.5: periodic convolution is scalar-linear in the second argument.
lemma helperForLemma_5_4_5_convolution_right_smul
(f g : C(AddCircle (1 : ℝ), ℂ)) (c : ℂ) :
periodicConvolution f (c • g) = c • periodicConvolution f g
Helper for Lemma 5.4.5: each Fourier character is an eigenfunction of periodic convolution with eigenvalue equal to the corresponding Fourier coefficient of the left factor.
lemma helperForLemma_5_4_5_convolution_with_character
(f : C(AddCircle (1 : ℝ), ℂ)) (n : ℤ) :
periodicConvolution f (fourierCharacter n) = (fourierTransform f n) • fourierCharacter n
Helper for Lemma 5.4.5: convolution with a trigonometric polynomial acts termwise on its Fourier expansion.
lemma helperForLemma_5_4_5_convolution_finite_fourier_sum
(f : C(AddCircle (1 : ℝ), ℂ)) (N : ℕ) (c : ℤ → ℂ) :
periodicConvolution f
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ)) (fun k => c k • fourierCharacter k))
=
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k => (fourierTransform f k * c k) • fourierCharacter k)5.4 — Periodic Convolutions — Part 2
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Lemma 5.4.6: for every ε > 0 and 0 < δ < 1/2, there exists a trigonometric polynomial P that is a periodic (ε, δ) approximation to the identity.
lemma exists_trigonometricPolynomial_isPeriodicApproximationToIdentity
(ε δ : ℝ) (hε : 0 < ε) (hδ₀ : 0 < δ) (hδ₁ : δ < (1 / 2 : ℝ)) :
∃ P : C(AddCircle (1 : ℝ), ℂ),
IsTrigonometricPolynomial P ∧ IsPeriodicApproximationToIdentity ε δ P
Helper for Lemma 5.4.6: Fejér polynomial on AddCircle (1 : ℝ) with frequency cutoff N.
noncomputable def helperForLemma_5_4_6_fejerPolynomial (N : ℕ) :
C(AddCircle (1 : ℝ), ℂ) :=
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k =>
((((N + 1 - Int.natAbs k : ℕ) : ℂ) / ((N + 1 : ℕ) : ℂ))) • fourierCharacter k)
Helper for Lemma 5.4.6: the Fejér polynomial is a trigonometric polynomial.
lemma helperForLemma_5_4_6_fejer_isTrig (N : ℕ) :
IsTrigonometricPolynomial (helperForLemma_5_4_6_fejerPolynomial N)
Helper for Lemma 5.4.6: character integral on [0,1] is 1 at frequency 0 and vanishes at nonzero frequencies.
lemma helperForLemma_5_4_6_intervalIntegral_fourierCharacter_eq_ite (k : ℤ) :
(∫ y in Set.Icc (0 : ℝ) 1, fourierCharacter k (y : AddCircle (1 : ℝ)))
= (if k = 0 then (1 : ℂ) else 0)
Helper for Lemma 5.4.6: rewrite the Fejér mass integral as a weighted finite character-integral sum.
lemma helperForLemma_5_4_6_fejerIntegral_as_weightedCharacterSum (N : ℕ) :
(∫ y in Set.Icc (0 : ℝ) 1, helperForLemma_5_4_6_fejerPolynomial N (y : AddCircle (1 : ℝ)))
= Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k =>
((((N + 1 - Int.natAbs k : ℕ) : ℂ) / ((N + 1 : ℕ) : ℂ)) *
(∫ y in Set.Icc (0 : ℝ) 1, fourierCharacter k (y : AddCircle (1 : ℝ)))))
Helper for Lemma 5.4.6: the Fejér polynomial has total mass 1 on [0,1].
lemma helperForLemma_5_4_6_fejer_mass_one (N : ℕ) :
(∫ y in Set.Icc (0 : ℝ) 1, helperForLemma_5_4_6_fejerPolynomial N (y : AddCircle (1 : ℝ)))
= (1 : ℂ)
Helper for Lemma 5.4.6: Archimedean index choice turning an O((N+1)⁻¹) bound into a strict < ε estimate.
lemma helperForLemma_5_4_6_chooseIndexFromTailBound
{ε C : ℝ} (hε : 0 < ε) (hC : 0 < C) :
∃ N : ℕ, C / (N + 1 : ℝ) < ε
Helper for Lemma 5.4.6: convert an integer-difference equation between natural numbers into a nonnegative shift equation.
lemma helperForLemma_5_4_6_intDiff_eq_natShift (i j m : ℕ) :
((i : ℤ) - (j : ℤ) = (m : ℤ)) ↔ i = j + m
Helper for Lemma 5.4.6: convert an integer-difference equation between natural numbers into a negative shift equation.
lemma helperForLemma_5_4_6_intDiff_eq_negNatShift (i j m : ℕ) :
((i : ℤ) - (j : ℤ) = - (m : ℤ)) ↔ j = i + m
Helper for Lemma 5.4.6: count pairs (i,j) in range (N+1) × range (N+1) with i = j + m.
lemma helperForLemma_5_4_6_pairCount_nonnegShift (N m : ℕ) :
((Finset.product (Finset.range (N + 1)) (Finset.range (N + 1))).filter
(fun p : ℕ × ℕ => p.1 = p.2 + m)).card = N + 1 - m
Helper for Lemma 5.4.6: count pairs (i,j) in range (N+1) × range (N+1) with j = i + m.
lemma helperForLemma_5_4_6_pairCount_nonnegShift_swapped (N m : ℕ) :
((Finset.product (Finset.range (N + 1)) (Finset.range (N + 1))).filter
(fun p : ℕ × ℕ => p.2 = p.1 + m)).card = N + 1 - m
Helper for Lemma 5.4.6: the number of pairs in range (N+1) × range (N+1) with integer difference k is N+1-|k|.
lemma helperForLemma_5_4_6_pairCount_eq_weight (N : ℕ) (k : ℤ) :
((Finset.product (Finset.range (N + 1)) (Finset.range (N + 1))).filter
(fun p : ℕ × ℕ => ((p.1 : ℤ) - (p.2 : ℤ) = k))).card = N + 1 - Int.natAbs k
Helper for Lemma 5.4.6: integer differences of indices from range (N+1) lie in the symmetric interval [-N, N].
lemma helperForLemma_5_4_6_diff_mem_Icc
(N i j : ℕ)
(hi : i ∈ Finset.range (N + 1))
(hj : j ∈ Finset.range (N + 1)) :
((i : ℤ) - (j : ℤ)) ∈ Finset.Icc (-(N : ℤ)) (N : ℤ)
Helper for Lemma 5.4.6: group the double range sum by fixed difference k = i - j.
lemma helperForLemma_5_4_6_doubleRangeSum_groupByDifference
(N : ℕ) (z : ℂ) :
Finset.sum (Finset.range (N + 1))
(fun i => Finset.sum (Finset.range (N + 1)) (fun j => z ^ i * star (z ^ j)))
=
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k =>
Finset.sum
((Finset.product (Finset.range (N + 1)) (Finset.range (N + 1))).filter
(fun p : ℕ × ℕ => (p.1 : ℤ) - (p.2 : ℤ) = k))
(fun p => z ^ p.1 * star (z ^ p.2)))
Helper for Lemma 5.4.6: on the fiber where i - j = k, the term z^i * conj(z^j) with z = exp(i 2π x) collapses to the k-mode exponential.
lemma helperForLemma_5_4_6_expTerm_onDifferenceFiber
(x : ℝ) (i j : ℕ) (k : ℤ)
(hDiff : ((i : ℤ) - (j : ℤ) = k)) :
let z : ℂ := Complex.exp (Complex.I * (2 * Real.pi * x))
z ^ i * star (z ^ j)
= Complex.exp (2 * Real.pi * Complex.I * (k : ℂ) * (x : ℂ))
Helper for Lemma 5.4.6: convert the weighted k-sum on [-N,N] into the double-range geometric sum.
lemma helperForLemma_5_4_6_weightedIccSum_eq_doubleRangeSum
(N : ℕ) (x : ℝ) :
let z : ℂ := Complex.exp (Complex.I * (2 * Real.pi * x))
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun k =>
(((N + 1 - Int.natAbs k : ℕ) : ℂ) *
Complex.exp (2 * Real.pi * Complex.I * (k : ℂ) * (x : ℂ))))
= Finset.sum (Finset.range (N + 1))
(fun i => Finset.sum (Finset.range (N + 1))
(fun j => z ^ i * star (z ^ j)))
Helper for Lemma 5.4.6: rewrite the weighted Fourier-coefficient sum for the Fejér kernel value into a normalized double geometric sum.
lemma helperForLemma_5_4_6_fejerCoeff_weightedSum_eq_doubleGeomSum
(N : ℕ) (x : ℝ) :
let z : ℂ := Complex.exp (Complex.I * (2 * Real.pi * x))
let D : ℂ := ((1 : ℂ) / ((N + 1 : ℕ) : ℂ)) *
(Finset.sum (Finset.range (N + 1))
(fun i => Finset.sum (Finset.range (N + 1))
(fun j => z ^ i * star (z ^ j))) )
helperForLemma_5_4_6_fejerPolynomial N (x : AddCircle (1 : ℝ)) = D
Helper for Lemma 5.4.6: the normalized double geometric sum equals the normalized squared geometric-sum norm, with vanishing imaginary part.
lemma helperForLemma_5_4_6_doubleGeomSum_eq_normSq_geomSum
(N : ℕ) (z : ℂ) :
let S : ℂ := Finset.sum (Finset.range (N + 1)) (fun j => z ^ j)
let D : ℂ := ((1 : ℂ) / ((N + 1 : ℕ) : ℂ)) *
(Finset.sum (Finset.range (N + 1))
(fun i => Finset.sum (Finset.range (N + 1))
(fun j => z ^ i * star (z ^ j))) )
D.re = ((1 : ℝ) / (N + 1 : ℝ)) * ‖S‖ ^ 2 ∧ D.im = 0
Helper for Lemma 5.4.6: Fejér-kernel evaluation at a real point as a normalized squared geometric-sum norm, with vanishing imaginary part.
lemma helperForLemma_5_4_6_fejer_eval_eq_normSqGeometricSum
(N : ℕ) (x : ℝ) :
(helperForLemma_5_4_6_fejerPolynomial N (x : AddCircle (1 : ℝ))).re
= ((1 : ℝ) / (N + 1 : ℝ)) *
‖Finset.sum (Finset.range (N + 1))
(fun j => (Complex.exp (Complex.I * (2 * Real.pi * x))) ^ j)‖ ^ 2 ∧
(helperForLemma_5_4_6_fejerPolynomial N (x : AddCircle (1 : ℝ))).im = 0
Helper for Lemma 5.4.6: geometric-sum norm bound away from the resonance exp(2πix) = 1.
lemma helperForLemma_5_4_6_geomSum_norm_le_two_div_normExpSubOne
(N : ℕ) (x : ℝ)
(hx : Complex.exp (Complex.I * (2 * Real.pi * x)) ≠ 1) :
‖Finset.sum (Finset.range (N + 1))
(fun j => (Complex.exp (Complex.I * (2 * Real.pi * x))) ^ j)‖
≤ 2 / ‖Complex.exp (Complex.I * (2 * Real.pi * x)) - 1‖
Helper for Lemma 5.4.6: on the annulus δ ≤ |x| ≤ 1 - δ with 0 < δ < 1/2, the denominator ‖exp(2πix) - 1‖ is uniformly bounded below by a positive constant.
lemma helperForLemma_5_4_6_annulus_expSubOne_lowerBound
{δ : ℝ} (hδ₀ : 0 < δ) (hδ₁ : δ < (1 / 2 : ℝ)) :
∃ mδ > 0,
∀ x : ℝ, δ ≤ |x| → |x| ≤ 1 - δ →
mδ ≤ ‖Complex.exp (Complex.I * (2 * Real.pi * x)) - 1‖
Helper for Lemma 5.4.6: combining the norm-square Fejér formula, geometric-sum denominator estimate, and annulus denominator lower bound gives the quantitative tail estimate.
lemma helperForLemma_5_4_6_fejer_tailBound_from_formula
{δ mδ : ℝ} (hmδ : 0 < mδ)
(hLower : ∀ x : ℝ, δ ≤ |x| → |x| ≤ 1 - δ →
mδ ≤ ‖Complex.exp (Complex.I * (2 * Real.pi * x)) - 1‖)
(N : ℕ) (x : ℝ) (hxδ : δ ≤ |x|) (hx1 : |x| ≤ 1 - δ) :
‖helperForLemma_5_4_6_fejerPolynomial N (x : AddCircle (1 : ℝ))‖
≤ (4 / (mδ ^ 2)) / (N + 1 : ℝ)
Helper for Lemma 5.4.6: Fejér kernels are pointwise real/nonnegative and satisfy uniform annulus decay ‖F_N(x)‖ ≤ Cδ/(N+1) when δ ≤ |x| ≤ 1-δ.
lemma helperForLemma_5_4_6_fejer_real_nonneg_and_tailBound
{δ : ℝ} (hδ₀ : 0 < δ) (hδ₁ : δ < (1 / 2 : ℝ)) :
(∀ N : ℕ,
∀ x : AddCircle (1 : ℝ),
(helperForLemma_5_4_6_fejerPolynomial N x).im = 0 ∧
0 ≤ (helperForLemma_5_4_6_fejerPolynomial N x).re) ∧
∃ Cδ > 0,
∀ N : ℕ,
∀ x : ℝ,
δ ≤ |x| → |x| ≤ 1 - δ →
‖helperForLemma_5_4_6_fejerPolynomial N (x : AddCircle (1 : ℝ))‖
≤ Cδ / (N + 1 : ℝ)
Helper for Lemma 5.4.6: quantitative Fejér-kernel concentration yields a periodic (ε, δ) approximation to the identity for some index.
lemma helperForLemma_5_4_6_exists_fejerPeriodicApproximation
(ε δ : ℝ) (hε : 0 < ε) (hδ₀ : 0 < δ) (hδ₁ : δ < (1 / 2 : ℝ)) :
∃ N : ℕ, IsPeriodicApproximationToIdentity ε δ (helperForLemma_5_4_6_fejerPolynomial N)5.5 — The Fourier and Plancherel Theorems
Lean authors: Zichen Wang, Wanli Ma, Shu Miao, Zaiwen Wen.
Theorem 5.5.1 (Fourier theorem): for every f ∈ C(ℝ/ℤ; ℂ), the symmetric Fourier partial sums ∑_{n=-N}^{N} \hat f(n) e_n converge to f in the L² metric, i.e. lim_{N→∞} ‖f - ∑_{n=-N}^{N} \hat f(n)e_n‖₂ = 0.
theorem fourier_series_tendsto_in_L2
(f : C(AddCircle (1 : ℝ), ℂ)) :
Filter.Tendsto
(fun N : ℕ =>
‖(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f) -
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourierLp (T := (1 : ℝ)) (p := 2) n)))‖)
Filter.atTop (nhds (0 : ℝ))
Theorem 5.5.3: let f ∈ C(ℝ/ℤ; ℂ). If ∑_{n∈ℤ} ‖\hat f(n)‖ converges, then the symmetric Fourier partial sums ∑_{n=-N}^{N} \hat f(n)e_n converge uniformly to f, i.e. lim_{N→∞} ‖f - ∑_{n=-N}^{N} \hat f(n)e_n‖∞ = 0 (with ‖·‖ the sup norm on C(ℝ/ℤ; ℂ)).
theorem fourier_series_tendsto_uniform_of_summable_fourierCoefficients
(f : C(AddCircle (1 : ℝ), ℂ))
(habs : Summable (fun n : ℤ => ‖fourierCoeff f n‖)) :
Filter.Tendsto
(fun N : ℕ =>
‖f -
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourier (T := (1 : ℝ)) n)))‖)
Filter.atTop (nhds (0 : ℝ))
Theorem 5.5.4 (Plancherel / Parseval): for any f ∈ C(ℝ/ℤ; ℂ), the series ∑_{n∈ℤ} |\hat f(n)|^2 converges and the L² norm identity ‖f‖₂^2 = ∑_{n=-∞}^{∞} |\hat f(n)|^2 holds.
theorem continuousMap_norm_sq_eq_tsum_sq_fourierCoeff
(f : C(AddCircle (1 : ℝ), ℂ)) :
Summable (fun n : ℤ => ‖fourierCoeff f n‖ ^ 2) ∧
‖ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f‖ ^ 2 =
∑' n : ℤ, ‖fourierCoeff f n‖ ^ 2
Helper for Theorem 5.5.1: the Fourier series of f viewed in L² has the expected coefficient formula and sums to toLp f.
lemma helperForTheorem_5_5_1_hasSum_fourier_series_toLp
(f : C(AddCircle (1 : ℝ), ℂ)) :
HasSum
(fun n : ℤ => (fourierCoeff f n) • (fourierLp (T := (1 : ℝ)) (p := 2) n))
(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f)
Helper for Theorem 5.5.1: symmetric integer windows [-N, N] tend to atTop in Finset ℤ.
lemma helperForTheorem_5_5_1_tendsto_symmetricIntegerWindows :
Filter.Tendsto (fun N : ℕ => Finset.Icc (-(N : ℤ)) (N : ℤ))
Filter.atTop Filter.atTop
Helper for Theorem 5.5.1: symmetric Fourier partial sums converge to toLp f in L².
lemma helperForTheorem_5_5_1_tendsto_partialSums_in_Lp
(f : C(AddCircle (1 : ℝ), ℂ)) :
Filter.Tendsto
(fun N : ℕ =>
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourierLp (T := (1 : ℝ)) (p := 2) n)))
Filter.atTop
(nhds (ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f))
Helper for Theorem 5.5.1: the L² norm of the Fourier approximation error tends to 0.
lemma helperForTheorem_5_5_1_normError_tendsto_zero
(f : C(AddCircle (1 : ℝ), ℂ)) :
Filter.Tendsto
(fun N : ℕ =>
‖(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f) -
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourierLp (T := (1 : ℝ)) (p := 2) n)))‖)
Filter.atTop (nhds (0 : ℝ))
Helper for Theorem 5.5.3: absolute summability of Fourier coefficients implies summability of the coefficients themselves.
lemma helperForTheorem_5_5_3_summable_fourierCoeff
(f : C(AddCircle (1 : ℝ), ℂ))
(habs : Summable (fun n : ℤ => ‖fourierCoeff f n‖)) :
Summable (fun n : ℤ => fourierCoeff f n)
Helper for Theorem 5.5.3: summable Fourier coefficients force symmetric Fourier partial sums to converge to f in C(ℝ/ℤ; ℂ).
lemma helperForTheorem_5_5_3_tendsto_symmetricPartialSums
(f : C(AddCircle (1 : ℝ), ℂ))
(hcoeff : Summable (fun n : ℤ => fourierCoeff f n)) :
Filter.Tendsto
(fun N : ℕ =>
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourier (T := (1 : ℝ)) n)))
Filter.atTop
(nhds f)
Helper for Theorem 5.5.3: convergence of symmetric partial sums to f implies convergence of the sup-norm error to 0.
lemma helperForTheorem_5_5_3_normError_tendsto_zero
(f : C(AddCircle (1 : ℝ), ℂ))
(hpartial :
Filter.Tendsto
(fun N : ℕ =>
Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourier (T := (1 : ℝ)) n)))
Filter.atTop
(nhds f)) :
Filter.Tendsto
(fun N : ℕ =>
‖f -
(Finset.sum (Finset.Icc (-(N : ℤ)) (N : ℤ))
(fun n => (fourierCoeff f n) • (fourier (T := (1 : ℝ)) n)))‖)
Filter.atTop
(nhds (0 : ℝ))
Helper for Theorem 5.5.4: Parseval's HasSum identity for toLp f, with Fourier coefficients rewritten in terms of the original continuous map f.
lemma helperForTheorem_5_5_4_hasSum_sq_fourierCoeff_toLp
(f : C(AddCircle (1 : ℝ), ℂ)) :
HasSum
(fun n : ℤ => ‖fourierCoeff f n‖ ^ 2)
(∫ t : AddCircle (1 : ℝ),
‖(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f) t‖ ^ 2
∂AddCircle.haarAddCircle)
Helper for Theorem 5.5.4: for toLp f, the square of the L² norm equals the integral of the pointwise squared norm.
lemma helperForTheorem_5_5_4_normSq_toLp_eq_integral_normSq
(f : C(AddCircle (1 : ℝ), ℂ)) :
‖ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f‖ ^ 2 =
∫ t : AddCircle (1 : ℝ),
‖(ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f) t‖ ^ 2
∂AddCircle.haarAddCircle
Helper for Theorem 5.5.4: the Parseval tsum identity for continuous maps on AddCircle 1.
lemma helperForTheorem_5_5_4_tsum_sq_fourierCoeff_eq_normSq
(f : C(AddCircle (1 : ℝ), ℂ)) :
(∑' n : ℤ, ‖fourierCoeff f n‖ ^ 2) =
‖ContinuousMap.toLp (E := ℂ) 2 AddCircle.haarAddCircle ℂ f‖ ^ 2
Helper for Theorem 5.5.4: summability of squared Fourier coefficients for continuous maps on AddCircle 1.
lemma helperForTheorem_5_5_4_summable_sq_fourierCoeff
(f : C(AddCircle (1 : ℝ), ℂ)) :
Summable (fun n : ℤ => ‖fourierCoeff f n‖ ^ 2)