Documentation

LeanPool.OperatorTheory.Operator.Crouzeix.VonNeumann

Von Neumann's inequality (L4.1) #

For a contraction T on a complex Hilbert space E and a polynomial p, ‖p(T)‖ ≤ sup_{‖z‖ ≤ 1} ‖p(z)‖.

Route (task-spec L4.1, via dilation): the Schäffer dilation (L3.1) gives a Hilbert space H, an inner-product-preserving V : E →L[ℂ] H and a unitary U on H with V† U^n V = T^n for all n. Then

  1. V† p(U) V = p(T) by linearity (adjoint_aeval_apply_of_forall_pow);
  2. ‖p(T)‖ ≤ ‖V†‖ ‖p(U)‖ ‖V‖ ≤ ‖p(U)‖ since ‖V†‖ = ‖V‖ ≤ 1 (norm_aeval_le_norm_aeval_of_dilation);
  3. for unitary U the element p(U) is normal, so ‖p(U)‖ equals its spectral radius; the spectral mapping theorem and σ(U) ⊆ 𝕊¹ ⊆ closedBall 0 1 bound this by the sup-norm of p on the closed unit disk (norm_aeval_le_polynomialSupNorm_closedBall_of_mem_unitary).

Main declarations #

Requires [CompleteSpace E] (adjoints, the C*-algebra structure on E →L[ℂ] E).

The polynomial sup-norm #

The polynomial sup-norm is nonnegative.

theorem norm_eval_le_polynomialSupNorm (p : Polynomial ℂ) {X : Set ℂ} (hX : BddAbove ((fun (z : ℂ) => ‖Polynomial.eval z p‖) '' X)) {z : ℂ} (hz : z ∈ X) :

If the values ‖p.eval z‖, z ∈ X, are bounded above, then each of them is at most polynomialSupNorm p X.

On a compact set the values ‖p.eval z‖ are bounded above.

Polynomials in a unitary #

Von Neumann's inequality for unitaries, in any C*-algebra: for U unitary, ‖p(U)‖ ≤ sup_{‖z‖ ≤ 1} ‖p(z)‖. Since p(U) is normal, ‖p(U)‖ is its spectral radius, and σ(p(U)) = p(σ(U)) with σ(U) contained in the unit circle.

Compression through a power dilation #

theorem adjoint_aeval_apply_of_forall_pow {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℂ E] [CompleteSpace E] {H : Type u_2} [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H] (T : E →L[ℂ] E) (V : E →L[ℂ] H) (U : H →L[ℂ] H) (hpow : ∀ (n : ℕ) (x : E), (ContinuousLinearMap.adjoint V) ((U ^ n) (V x)) = (T ^ n) x) (p : Polynomial ℂ) (x : E) :

If V† U^n V = T^n for all n, then V† p(U) V = p(T) for every polynomial p.

theorem norm_aeval_le_norm_aeval_of_dilation {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℂ E] [CompleteSpace E] {H : Type u_2} [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H] (T : E →L[ℂ] E) (V : E →L[ℂ] H) (U : H →L[ℂ] H) (hV : ∀ (x y : E), inner ℂ (V x) (V y) = inner ℂ x y) (hpow : ∀ (n : ℕ) (x : E), (ContinuousLinearMap.adjoint V) ((U ^ n) (V x)) = (T ^ n) x) (p : Polynomial ℂ) :

If V preserves inner products and V† U^n V = T^n for all n, then ‖p(T)‖ ≤ ‖p(U)‖ for every polynomial p.

Von Neumann's inequality #

theorem vonNeumann_inequality_of_exists_unitary_power_dilation {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℂ E] [CompleteSpace E] (T : E →L[ℂ] E) (hdil : ∃ (H : Type u) (x : NormedAddCommGroup H) (x_1 : InnerProductSpace ℂ H) (x_2 : CompleteSpace H) (V : E →L[ℂ] H) (U : H →L[ℂ] H), (∀ (x_3 y : E), inner ℂ (V x_3) (V y) = inner ℂ x_3 y) ∧ U ∈ unitary (H →L[ℂ] H) ∧ ∀ (n : ℕ) (x_3 : E), (ContinuousLinearMap.adjoint V) ((U ^ n) (V x_3)) = (T ^ n) x_3) (p : Polynomial ℂ) :

Von Neumann's inequality for an operator admitting a unitary power dilation: if there are a Hilbert space H, an inner-product-preserving V : E →L[ℂ] H and a unitary U on H with V† U^n V = T^n for all n (the conclusion of L3.1), then ‖p(T)‖ ≤ sup_{‖z‖ ≤ 1} ‖p(z)‖ for every polynomial p.

Von Neumann's inequality: for a contraction T and a polynomial p, ‖p(T)‖ ≤ sup_{‖z‖ ≤ 1} ‖p(z)‖.