Documentation

TauCeti.Analysis.SpecialFunctions.Trigonometric.Orthogonality

Orthogonality of the sine system on [0, π] #

The functions θ ↦ sin (k θ), for k a positive natural number, are pairwise orthogonal on the interval [0, π], each of square norm π / 2:

∫₀^π sin (m θ) · sin (n θ) dθ = (π / 2) · δ_{mn}.

This file proves that identity and its normalized form, in which the system √(2/π) · sin (k ·) is orthonormal.

The proof is the product-to-sum identity 2 sin x sin y = cos (x - y) - cos (x + y) followed by the single computation ∫₀^π cos (k θ) dθ = 0 for a nonzero integer k, which is TauCeti.integral_cos_int_mul below. That computation is elementary — the antiderivative sin (k θ) / k vanishes at both endpoints because sin vanishes at every integer multiple of π — and is obtained here from intervalIntegral.mul_integral_comp_mul_left and integral_cos.

The cosine counterpart of the orthogonality relation is already available in this repository, in the Chebyshev vocabulary: TauCeti.integral_chebyshevCosine_mul_chebyshevCosine_eq_ite (TauCeti/Analysis/SpecialFunctions/Trigonometric/Chebyshev/Cosine/Transfer.lean) records ∫₀^π cos (m θ) cos (n θ) dθ through the Chebyshev T measure. Nothing there covers the sine system, whose square norm is π / 2 in every degree — the cosine system is the one whose degree-0 mode has square norm π instead — so the two families are stated separately, and this file is deliberately independent of the Chebyshev measure-theoretic development. In the other direction the single-mode integral does serve that development: the two cosine-side single-mode integrals TauCeti.integral_chebyshevCosine_zero and TauCeti.integral_chebyshevCosine_of_ne_zero are read off TauCeti.integral_cos_int_mul below, so the computation is performed once.

Main results #

References #

The sine system on [0, π] is the classical Fourier sine basis; see for instance E. M. Stein, R. Shakarchi, Fourier Analysis: An Introduction, Princeton (2003), Chapter 2.

The normalized form is the identity that the engine case of TauCetiRoadmap/RepresentationTheory/CompactGroups/README.md reduces the character orthonormality of SU(2) to; see TauCeti/RepresentationTheory/SU2/Weyl/Orthogonality.lean.

theorem TauCeti.integral_cos_int_mul (k : ℤ) :
∫ (θ : ℝ) in 0..Real.pi, Real.cos (↑k * θ) = if k = 0 then Real.pi else 0

∫₀^π cos (k θ) dθ for an integer k: it is π in degree 0, and vanishes otherwise, because the antiderivative sin (k θ) / k vanishes at both endpoints.

theorem TauCeti.integral_sin_nat_mul_mul_sin_nat_mul (m n : ℕ) :
∫ (θ : ℝ) in 0..Real.pi, Real.sin (↑m * θ) * Real.sin (↑n * θ) = if m = n ∧ m ≠ 0 then Real.pi / 2 else 0

Orthogonality of the sine system on [0, π]. For natural numbers m and n, ∫₀^π sin (m θ) sin (n θ) dθ is π / 2 when m = n is nonzero, and 0 otherwise.

The degenerate mode m = n = 0 is absorbed into the right-hand side rather than hypothesised away: there the integrand vanishes identically, so the integral is 0, not π / 2. The condition m = n ∧ m ≠ 0 is symmetric in m and n up to logical equivalence, since under m = n the requirements m ≠ 0 and n ≠ 0 coincide.

theorem TauCeti.integral_sin_succ_mul_sin_succ (m n : ℕ) :
∫ (θ : ℝ) in 0..Real.pi, Real.sin ((↑m + 1) * θ) * Real.sin ((↑n + 1) * θ) = if m = n then Real.pi / 2 else 0

The sine system indexed by its positive frequencies m + 1, so that the degenerate mode of TauCeti.integral_sin_nat_mul_mul_sin_nat_mul is out of range and the right-hand side is a plain Kronecker delta: ∫₀^π sin ((m+1) θ) sin ((n+1) θ) dθ = (π / 2) · δ_{mn}.

theorem TauCeti.two_div_pi_mul_integral_sin_succ_mul_sin_succ (m n : ℕ) :
2 / Real.pi * ∫ (θ : ℝ) in 0..Real.pi, Real.sin ((↑m + 1) * θ) * Real.sin ((↑n + 1) * θ) = if m = n then 1 else 0

The sine system, normalized. With the 2/π normalization the family √(2/π) · sin ((m+1) ·) is orthonormal on [0, π]: (2/π) ∫₀^π sin ((m+1) θ) sin ((n+1) θ) dθ = δ_{mn}.

This is the form the character orthonormality of SU(2) reduces to, and TauCeti.SU2.character_symPower_orthonormal_torusExp (TauCeti/RepresentationTheory/SU2/Weyl/Orthogonality.lean) is proved by rewriting with it.