Keyboard shortcuts

Press ← or → to navigate between chapters

Press ? to show this help

Press Esc to hide this help

Half-Line Mellin Kernel

Abstract

The exact Mellin separator kernel and dual norm on the positive half-line.

Work in H=Lp(C,2,volume restricted to (0,infinity)). Let beta=Re(rho)>1/2 and rho differ from one. Positive real bases use the principal complex power. The kernel is conjugated because the first argument of the complex inner product is conjugate-linear.

Definition 1.1 (Real source vector).

Formalization. D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.realSourceVector (✓ std3).

Citation. Luis Báez-Duarte (2002). A strengthening of the Nyman-Beurling criterion for the Riemann Hypothesis. URL: https://arxiv.org/abs/math/0202141v2.

Commentary.

F_a is the Lp class of x |-> ofReal(fract(1/(a*x))). This realizes the Introduction’s real-dilation family in the complex Lp space. The construction proves square integrability before taking the quotient.

Theorem 1.2 (Actual representative).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.realSourceVector_coe_ae (✓ std3). ∎

Source. Repository-derived.

Commentary.

The equality is almost everywhere for positiveMeasure. Real values are embedded into C, and the measure and exponent remain those of H.

Theorem 1.3 (Natural sources agree).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.realSourceVector_nat (✓ std3). ∎

Source. Repository-derived.

Commentary.

Equality in the Lp quotient identifies the real-parameter construction with the original natural source family, including n=1.

Theorem 1.4 (Reciprocal tail).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.realSourceVector_tail (✓ std3). ∎

Source. Repository-derived.

Commentary.

Since 0 < 1/(a*x) < 1 on this tail, taking fractional part changes nothing. This is a pointwise identity for every real a at least one.

Definition 1.5 (Kernel vector).

Formalization. D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineKernel (✓ std3).

Source. Repository-derived.

Commentary.

The near-zero piece has square norm x^(2*beta-2); the tail has square norm norm(rho-1)^(-2)*x^(-2). Their integrability constructs the Lp vector.

Theorem 1.6 (Conjugated representative).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineKernel_coe_ae (✓ std3). ∎

Source. Repository-derived.

Commentary.

This representative identity is almost everywhere for positiveMeasure. The two open intervals are disjoint; values at endpoints have measure zero.

Theorem 1.7 (Exact kernel energy).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineKernel_norm_sq (✓ std3). ∎

Source. Repository-derived.

Commentary.

Disjoint support makes the square norm a sum. The power integral over (0,1) is 1/(2*beta-1); the integral of x^(-2) over (1,infinity) is one.

Definition 1.8 (Continuous complex-linear functional).

Formalization. D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineFunctional (✓ std3).

Source. Repository-derived.

Commentary.

The Riesz inner-product map defines J as a continuous C-linear map from H to C.

Theorem 1.9 (Representative-independent integral formula).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineFunctional_integral (✓ std3). ∎

Source. Repository-derived.

Commentary.

For every g equal to the Lp representative almost everywhere for positiveMeasure, g(x)*x^(rho-1) is integrable on (0,1), and g(x)/x is integrable on (1,infinity). The displayed identity holds for every such g, so it is independent of the chosen representative. Cauchy-Schwarz gives integrability, and rho-1 nonzero allows cancellation of the tail coefficient.

Theorem 1.10 (Exact operator norm).

Proof. Machine-checked in Lean as D5/S3/Observer/Hilbert/NymanHalflineMellinKernel.halflineFunctional_norm_sq (✓ std3). ∎

Source. Repository-derived.

Commentary.

The isometry of the Riesz map identifies the operator norm with the kernel norm. The result is an equality, with both support contributions retained.

References