How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Qm Math Coulomb

1 · Prerequisites

2 · Summary

The Coulomb layer of the non-relativistic quantum mechanics mathematical prerequisites, entirely mathematical: the three-dimensional Hardy inequality and the resulting infinitesimal kinetic bound, the relative compactness and essential spectrum of the Coulomb operator, the polar-coordinate unitary reduction to radial operators with their transported endpoint domain, the Kummer and Laguerre solution theory of the negative spectrum, the two-body reduced-mass separation, the dipole matrix-element domain and selection rules, and the Jost-Volterra construction of the continuum with its spectral density and onto radial transform.

Two local prerequisites open the page: L^2 kernel operators and Hilbert-Schmidt operators on R^3 and L^2 kernel operators are bounded and compact supply the $L^{2}$ -kernel calculus used by the relative-compactness argument, and Closed semibounded symmetric forms and their associated operators fixes the closed semibounded form terminology used by the radial realisation. Every item carries domain: mathematics and depends only on mathematical suppliers from qm-math-composites, qm-math-oscillator, qm-math-intervals-and-contacts, qm-math-orbital-and-spin and published mathematics; nothing here asserts a physical interpretation, measurement model or empirical claim, and the companion page qm-math-coulomb-examples stays mathematics as well.

Three-dimensional Hardy bound proves $∥ u / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u ∥_{2}$ on $W^{1, 2} (R^{3})$ by integration by parts with a cutoff and a density argument, and Coulomb infinitesimal kinetic bound converts it into the bound $∥ u / ∣ \cdot ∣ ∥_{2} \leq η ∥ Δ u ∥_{2} + η^{- 1} ∥ u ∥_{2}$ , so that $C = - Δ - γ / ∣ \cdot ∣$ on $H^{2} (R^{3})$ is self-adjoint and bounded below by Kato-Rellich. Coulomb relative compactness factors the free resolvent through an $L^{2}$ convolution kernel and splits the Coulomb multiplier into a Hilbert-Schmidt annulus plus small short-range and tail pieces, giving $H_{0}$ -compactness; Coulomb essential spectrum applies Weyl's theorem to get $σ_{e s s} (C) = [0, \infty)$ and the discreteness of the negative spectrum.

Angular radial unitary reduction constructs the unitary polar decomposition $ψ \mapsto (u_{l m})$ , $u_{l m} = r \int ψ \overline{Y_{l m}}$ into $⨁ L^{2} ((0, \infty), d r)$ , proves the polar form of the Laplacian and the single-block reduction $- Δ \leftrightarrow - u^{''} + l (l + 1) u / r^{2}$ ; Transported radial endpoint realization defines the radial operator $A_{l}$ through the closed form $\int (∣ u^{'} - u / r ∣^{2} + l (l + 1) ∣ u ∣^{2} / r^{2} - γ ∣ u ∣^{2} / r)$ , proves that it is self-adjoint with $U C U^{*} = ⨁ A_{l}$ , that its domain is the transported $H^{2}$ domain, and that its elements satisfy the differential equation distributionally almost everywhere, with classical $C^{2}$ solutions for the eigenfunctions. Kummer recurrence and termination solves the radial equation in terms of the Kummer series and the Laguerre polynomials and identifies the terminating solutions; Laguerre bound energy exhaustion shows that non-terminating solutions grow exponentially and therefore excludes them, so that the negative spectrum of $A_{l}$ is exactly ${- γ^{2} / (4 n^{2}) : n \geq l + 1}$ ; Negative-subspace completeness and degeneracy proves that these bound states form a complete orthonormal family of the negative spectral subspace with degeneracy $n^{2}$ and do not span $L^{2}$ .

Ground radial probability and units computes the normalised ground vector $(γ^{3} / 8 π)^{1 / 2} e^{- r / a_{0}}$ , the radial law $4 r^{2} a_{0}^{- 3} e^{- 2 r / a_{0}}$ , its mean $3 a_{0} / 2$ and mode $a_{0}$ , and records the dimension convention $a_{0} = ℏ^{2} / (m κ)$ ; Two-body reduced mass separates the two-body problem by the unit-determinant coordinate map $(x_{1}, x_{2}) \mapsto (R, r)$ , proves $U_{Φ} H_{2} U_{Φ}^{- 1} = T_{R} \otimes I + I \otimes C_{μ}$ with reduced mass $μ = m_{1} m_{2} / (m_{1} + m_{2})$ , and shows $σ (H_{2}) = [- μ κ^{2} / (2 ℏ^{2}), \infty)$ with no eigenvalues; Atomic dipole matrix-element domain places the bound states in the domain of the coordinate multipliers, factors the dipole pairings into radial and angular parts, applies the angular selection rules $Δ l = \pm 1$ , $Δ m = q$ , and records the oscillator comparison $X ψ_{n} \propto \sqrt{n} ψ_{n - 1} + \sqrt{n + 1} ψ_{n + 1}$ .

The continuum layer is Coulomb Jost Volterra solution, which constructs the regular solution by an entire power series and the Jost-type solution by a modified Volterra equation with controlled asymptotics $e^{i (k r + a \log r)} (1 + O (1 / r))$ , proves the Wronskian identities $W (\overline{f}, f) = 2 i k$ , $W_{l} \neq 0$ and excludes positive eigenvalues; Coulomb continuum spectral density derives the resolvent Green kernel $- φ (\min) f (\max) / W_{l}$ , its boundary value $Im G (k^{2} + i 0) = k φ φ / ∣ W_{l} ∣^{2}$ , the spectral density $ρ_{l} (E) = \sqrt{E} / (π ∣ W_{l} (\sqrt{E}) ∣^{2})$ and the absence of point and singular continuous spectrum on $(0, \infty)$ , including the scaling argument excluding the zero eigenvalue; and Onto radial continuum transform defines the compact-source transform $(T_{l} g) (E) = \sqrt{ρ_{l} (E)} \int φ_{l} (\sqrt{E}, r) g (r) d r$ , extends it to a contraction $T$ on the full radial $L^{2}$ space, and proves that its restriction ${\hat{T}}_{l}$ to the positive spectral subspace is unitary onto $L^{2} ((0, \infty), d E)$ , with $T = {\hat{T}}_{l} E_{A_{l}} ((0, \infty))$ , diagonalises $A_{l}$ as multiplication by $E$ , and assembles the complete spectral resolution of $C$ as bound data plus continuum data in every angular channel.

3 · Logical flowchart

4 · Definitions, postulates, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

L^2 kernel operators and Hilbert-Schmidt operators on R^3

Definition

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Write $L^{2} = L^{2} (R^{3}; C)$ for the complex Hilbert space of Lebesgue (equivalence classes of) square-integrable functions with inner product $⟨ f, g ⟩ = \int_{R^{3}} f \overline{g}$ linear in the first argument (The space $L^{p} (μ)$ as the quotient by null functions, Complex Lp classes and Euclidean test-function conventions, Hilbert space), and write $L^{2} (R^{3} \times R^{3}; C)$ for the same construction over the product with Lebesgue measure; identify Lebesgue measure on $R^{6}$ with the completion of the product of Lebesgue measures on the two $R^{3}$ factors.

Kernel data and sections. Let $k \in L^{2} (R^{3} \times R^{3}; C)$ , an almost-everywhere equivalence class, and choose a finite-valued Lebesgue-measurable representative $k_{0} : R^{3} \times R^{3} \to C$ . When $f \in L^{2}$ is used, choose likewise a finite-valued Lebesgue-measurable representative $f_{0} : R^{3} \to C$ . For almost every $x \in R^{3}$ , the section $k_{x} (y) : = k_{0} (x, y)$ is measurable, by applying the completed-product Fubini theorem to the positive and negative parts of the real and imaginary components of $k_{0}$ (Tonelli and Fubini for the completed product, with only almost-everywhere section measurability).

Domain and action. For $f \in L^{2}$ say that the integral defining $T_{k} f$ exists at $x \in R^{3}$ if the section product $y \mapsto k_{0} (x, y) f_{0} (y)$ is measurable and lies in $L^{1} (R^{3}; C)$ , and at such points put $(T_{k} f) (x) : = \int_{R^{3}} k_{0} (x, y) f_{0} (y) d y .$ Define $D (T_{k}) : = {f \in L^{2} : x \mapsto (T_{k} f) (x) exists for almost every x and agrees a.e. with a class in L^{2}},$ and for $f \in D (T_{k})$ let $T_{k} f$ denote that class. The assignment $f \mapsto T_{k} f$ is called the $L^{2}$ -kernel operator with kernel $k$ .

This definition is independent of both representative choices. If $k_{1}$ is another representative of $k$ , its disagreement set with $k_{0}$ is null for the completed product Lebesgue measure. Fubini applied to the indicator of that set gives $k_{0} (x, \cdot) = k_{1} (x, \cdot)$ almost everywhere for almost every $x$ . For each such $x$ and any representative of $f$ , the two section products therefore have the same $L^{1}$ status and, when integrable, the same integral. Replacing $f_{0}$ by another representative changes the section product only on a null set of $y$ , so it too preserves integrability and the integral. Consequently the almost-everywhere sections, the subset $D (T_{k}) \subseteq L^{2}$ , and the resulting output class are unchanged by either representative choice.

Hilbert-Schmidt operators. A linear operator $T$ on $L^{2}$ is called a Hilbert-Schmidt operator if there is $k \in L^{2} (R^{3} \times R^{3}; C)$ with $D (T) = D (T_{k})$ and $T = T_{k}$ ; such a $k$ is called an $L^{2}$ kernel of $T$ .

This definition asserts nothing about whether $D (T_{k})$ is all of $L^{2}$ , whether $T_{k}$ is bounded, or whether $T_{k}$ is compact; those properties are proved in L^2 kernel operators are bounded and compact. A finite product kernel $\sum_{i = 1}^{n} f_{i} (x) \overline{g_{i} (y)}$ with $f_{i}, g_{i} \in L^{2}$ is also a class in $L^{2} (R^{3} \times R^{3}; C)$ , as the same lemma records, so $T_{k}$ is defined for it whenever that membership is used.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

L^2 kernel operators are bounded and compact

Statement

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Let $L^{2} = L^{2} (R^{3}; C)$ and let $k \in L^{2} (R^{3} \times R^{3}; C)$ , with $T_{k}$ the $L^{2}$ -kernel operator of L^2 kernel operators and Hilbert-Schmidt operators on R^3. Then:

(a) $D (T_{k}) = L^{2}$ ; $T_{k}$ is linear and bounded with $∥ T_{k} f ∥_{2} \leq ∥ k ∥_{L^{2} (R^{3} \times R^{3})} ∥ f ∥_{2}$ for every $f \in L^{2}$ , so $∥ T_{k} ∥ \leq ∥ k ∥_{L^{2} (R^{3} \times R^{3})}$ ;

(b) if $n \geq 1$ , $f_{1}, \dots, f_{n}, g_{1}, \dots, g_{n} \in L^{2}$ and $k (x, y) = \sum_{i = 1}^{n} f_{i} (x) \overline{g_{i} (y)}$ a.e., then $k$ is a class in $L^{2} (R^{3} \times R^{3}; C)$ , the range of $T_{k}$ is contained in $span {f_{1}, \dots, f_{n}}$ , and $T_{k}$ is compact;

(d) for every sequence $(k_{m})$ in $L^{2} (R^{3} \times R^{3}; C)$ with $∥ k_{m} - k ∥_{L^{2} (R^{3} \times R^{3})} \to 0$ one has $T_{k_{m}} - T_{k} = T_{k_{m} - k}$ and $∥ T_{k_{m}} - T_{k} ∥ \leq ∥ k_{m} - k ∥_{L^{2} (R^{3} \times R^{3})} \to 0$ ; in particular $T_{k}$ is the operator-norm limit of finite-rank kernel operators.

No statement is made about the spectrum of $T_{k}$ , about its singular values, or about $L^{2}$ kernels of a given compact operator.

Facts & Assumptions

Given: the Hilbert space $L^{2}$ and its product $L^{2} (R^{3} \times R^{3}; C)$ ; a kernel $k$ of the stated class; the operators $T_{k}$ of L^2 kernel operators and Hilbert-Schmidt operators on R^3.

[F1]

Cauchy-Schwarz for $L^{2}$ : $\int ∣ h g ∣ \leq ∥ h ∥_{2} ∥ g ∥_{2}$ whenever $h, g \in L^{2}$ . Completed-product Tonelli gives measurable section integrals and $\int_{R^{3}} (\int_{R^{3}} ∣ k (x, y) ∣^{2} d y) d x = ∥ k ∥_{L^{2} (R^{3} \times R^{3})}^{2}$ ; completed-product Fubini gives measurable integrals of integrable real and imaginary components (Cauchy-Schwarz inequality for $L^{2}$ , Tonelli and Fubini for the completed product, with only almost-everywhere section measurability).

[F2]

Density of box step functions: for $1 \leq p < \infty$ the finite linear combinations of indicators of half-open boxes in $R^{n}$ with rational endpoints and rational coefficients are dense in $L^{p} (R^{n})$ (Rational box-step functions form a countable dense subset of $L^{p} (R^{n})$ for $1 \leq p < \infty$ ).

[F3]

A bounded linear operator between normed spaces whose range has a basis of finite length is compact (Bounded finite rank operators are compact). Under Countable Choice, an operator-norm limit of compact operators from a normed space into a Banach space is compact, provided the limit is a bounded linear operator (Norm limit of compact operators is compact). Here the target $L^{2}$ is a Banach space because it is a Hilbert space (Hilbert space), and step 1.1 supplies boundedness for each kernel operator to which these results are applied.

[F4]

$L^{2}$ elements are almost-everywhere classes with the linear-first pairing $⟨ f, g ⟩ = \int f \overline{g}$ . The kernel definition proves that changes of kernel or input representative preserve the section integrals almost everywhere (The space $L^{p} (μ)$ as the quotient by null functions, Complex Lp classes and Euclidean test-function conventions, L^2 kernel operators and Hilbert-Schmidt operators on R^3).

Proof

1.1F1F4algebra

[Boundedness and full domain] Fix $f \in L^{2}$ . For a.e. $x$ the function $y \mapsto k (x, y) f (y)$ is measurable, and by [F1] $\int ∣ k (x, y) f (y) ∣ d y \leq ∥ f ∥_{2} (\int ∣ k (x, y) ∣^{2} d y)^{1 / 2} < \infty$ for a.e. $x$ , the last integral being finite for a.e. $x$ with $\int (\int ∣ k (x, y) ∣^{2} d y) d x = ∥ k ∥_{2}^{2} < \infty$ ; hence $(T_{k} f) (x)$ exists for a.e. $x$ and $∣ (T_{k} f) (x) ∣ \leq ∥ f ∥_{2} (\int ∣ k (x, y) ∣^{2} d y)^{1 / 2}$ . For each bounded box $B \subset R^{3}$ , integrating the preceding bound gives $\int_{B} \int ∣ k (x, y) f (y) ∣ d y d x \leq ∣ B ∣^{1 / 2} ∥ k ∥_{2} ∥ f ∥_{2} < \infty$ . Applying [F1] to the real and imaginary components of $1_{B} (x) k (x, y) f (y)$ shows that $T_{k} f$ is measurable on $B$ up to a null set; a countable box exhaustion proves measurability on $R^{3}$ . Integrating the square and using completed-product Tonelli gives $∥ T_{k} f ∥_{2}^{2} \leq ∥ f ∥_{2}^{2} \int \int ∣ k (x, y) ∣^{2} d x d y = ∥ k ∥_{2}^{2} ∥ f ∥_{2}^{2}$ , so $D (T_{k}) = L^{2}$ , $T_{k} f \in L^{2}$ with the stated bound, and $T_{k}$ is linear because the integral is linear in $f$ for each $x$ . This is (a).

2.1step 1.1F1F3F4algebra

[Finite product kernels] Let $k (x, y) = \sum_{i = 1}^{n} f_{i} (x) \overline{g_{i} (y)}$ a.e. with $f_{i}, g_{i} \in L^{2}$ . By Tonelli and [F1], $\int \int ∣ k ∣^{2} \leq (\sum_{i} ∥ f_{i} ∥_{2} ∥ g_{i} ∥_{2})^{2} < \infty$ , so $k$ is a class in $L^{2} (R^{3} \times R^{3}; C)$ ; moreover for $f \in L^{2}$ step 1.1 applies and, for a.e. $x$ , $(T_{k} f) (x) = \sum_{i} f_{i} (x) \int \overline{g_{i} (y)} f (y) d y = \sum_{i} ⟨ f, g_{i} ⟩ f_{i} (x)$ by linearity of the integral and [F4]; hence the range of $T_{k}$ lies in the finite-dimensional space $span {f_{1}, \dots, f_{n}}$ . By [F3] $T_{k}$ is compact. This is (b).

3.1step 2.1F2F4algebra

[Box step functions are finite product kernels] A half-open box in $R^{6}$ is a product $B \times B^{'}$ of half-open boxes $B, B^{'} \subseteq R^{3}$ , and its indicator satisfies $1_{B \times B^{'}} (x, y) = 1_{B} (x) 1_{B^{'}} (y)$ for every $(x, y)$ ; the indicators of bounded boxes lie in $L^{2} (R^{3}; C)$ . Hence every finite linear combination of indicators of half-open boxes in $R^{6}$ with coefficients in $Q + i Q$ is a finite product kernel as in step 2.1, and its kernel operator is compact by step 2.1.

4.1step 1.1step 3.1F2F3algebra

[Norm approximation and compactness] Let $k \in L^{2} (R^{3} \times R^{3}; C)$ . For each integer $m \geq 0$ , apply [F2] with $n = 6$ , $p = 2$ and tolerance $1 / (m + 1)$ separately to the real $L^{2}$ functions $Re k$ and $Im k$ . By the assumed Countable Choice, select real rational box steps $a_{m}, b_{m}$ for all $m \geq 0$ with $∥ a_{m} - Re k ∥_{2} < 1 / (m + 1)$ and $∥ b_{m} - Im k ∥_{2} < 1 / (m + 1)$ . Then $s_{m} : = a_{m} + i b_{m}$ is a finite box step with coefficients in $Q + i Q$ , and $∥ s_{m} - k ∥_{2}^{2} = ∥ a_{m} - Re k ∥_{2}^{2} + ∥ b_{m} - Im k ∥_{2}^{2} < 2 / (m + 1)^{2}$ . For each $m$ , step 3.1 makes $T_{s_{m}}$ compact, and by linearity of the kernel operator $T_{k} - T_{s_{m}} = T_{k - s_{m}}$ ; step 1.1 applied to the kernel $k - s_{m}$ gives $∥ T_{k} - T_{s_{m}} ∥ \leq ∥ k - s_{m} ∥_{L^{2} (R^{3} \times R^{3})} \to 0$ . By [F3] the operator-norm limit $T_{k}$ is compact, which proves (c). For any sequence $(k_{m})$ as in (d), linearity gives $T_{k_{m}} - T_{k} = T_{k_{m} - k}$ , and step 1.1 gives $∥ T_{k_{m}} - T_{k} ∥ \leq ∥ k_{m} - k ∥_{2} \to 0$ , proving (d).

5.1step 1.1step 2.1step 4.1∎

[Conclusion] Part (a) is step 1.1, part (b) is step 2.1, and parts (c) and (d) are step 4.1.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Closed semibounded symmetric forms and their associated operators

Definition

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Let $(V, ⟨ \cdot, \cdot ⟩_{V})$ and $(H, ⟨ \cdot, \cdot ⟩_{H})$ be complex Hilbert spaces (Hilbert space, Real and complex inner product spaces, with the inner product linear in the first argument) and let $i : V \to H$ be a linear map with dense image.

Symmetric forms. A sesquilinear form on $V$ is a map $a : V \times V \to C$ that is linear in its first argument and conjugate-linear in its second; it is symmetric if $a (u, v) = \overline{a (v, u)}$ for all $u, v \in V$ , in which case $a (u, u) \in R$ and $Re a (u, v) = Re \overline{a (v, u)}$ .

Bounded and semibounded. Such a form is bounded if there is $C > 0$ with $∣ a (u, v) ∣ \leq C ∥ u ∥_{V} ∥ v ∥_{V}$ for all $u, v \in V$ , and semibounded below with bound $M \geq 0$ if $a (u, u) \geq - M ∥ i (u) ∥_{H}^{2}$ for every $u \in V$ .

Form seminorm and closedness. Let $M \geq 0$ be a bound as above and define the form seminorm $∥ u ∥_{a, M} : = (Re a (u, u) + (M + 1) ∥ i (u) ∥_{H}^{2})^{1 / 2} (u \in V) .$ It is a seminorm because $a (u, v) + (M + 1) ⟨ i (u), i (v) ⟩_{H}$ is a positive semidefinite Hermitian form; its Cauchy–Schwarz inequality follows by expanding its value on $u + t v$ and minimizing over the scalar $t$ , and gives the triangle inequality. It need not separate points for a general dense-image map $i$ . The form $a$ is closed if there are such an $M$ and constants $c, C > 0$ with $c ∥ u ∥_{V} \leq ∥ u ∥_{a, M} \leq C ∥ u ∥_{V}$ for every $u \in V$ ; then the form seminorm is a norm, and $(V, ∥ \cdot ∥_{a, M})$ is complete because $V$ is. If $i$ is injective, it is already a norm before closedness, since its square is at least $∥ i (u) ∥_{H}^{2}$ .

Associated operator. With $a$ and $i$ as above, define $D (A) : = {u \in V : there is f \in H with a (u, v) = ⟨ f, i (v) ⟩_{H} for every v \in V}, A u : = f .$ The element $f$ is unique when it exists, because two candidates differ by an element of $H$ orthogonal to the dense subspace $i (V)$ , hence coincide; $A$ is called the operator associated with $a$ . This definition asserts neither that $A$ is densely defined, nor that $A$ is symmetric, nor that $A$ is self-adjoint or bounded below; the specific closed semibounded form of Transported radial endpoint realization establishes those properties for its own associated operator by a direct computation. When $V$ is a subspace of $H$ and $i$ is the inclusion, $∥ i (u) ∥_{H}$ is written $∥ u ∥_{H}$ and the pairing $⟨ f, i (v) ⟩_{H}$ is written $⟨ f, v ⟩$ .

Remarks

For $V = C^{2}$ , $H = C$ , $i (u, v) = u$ and $a = 0$ , the map has dense image but the form seminorm vanishes on $(0, 1)$ . Thus the general dense-image setting requires the seminorm qualification; the radial consumer uses the actual inclusion and proves closedness directly.

LemmaStatement: AI-adaptedProof: AI-adaptedOpen item page →

Three-dimensional Hardy bound

Statement

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Let $W^{1, 2} (R^{3}; C)$ be the Sobolev space of Integer-order Sobolev spaces and their norms with the weak gradient $\nabla u$ of Weak derivative of a locally integrable function, write $r (x) : = ∣ x ∣$ , and write $∥ u ∥_{2}^{2} = \int_{R^{3}} ∣ u (x) ∣^{2} d x$ for the $L^{2}$ norm of The space $L^{p} (μ)$ as the quotient by null functions and Complex Lp classes and Euclidean test-function conventions. Then:

(a) for every $u \in C_{c}^{\infty} (R^{3} ∖ {0}; C)$ ,

$\int_{R^{3}} \frac{∣ u (x) ∣^{2}}{∣ x ∣^{2}} d x \leq 4 \int_{R^{3}} ∣ \nabla u (x) ∣^{2} d x;$

(b) the same inequality holds for every $u \in C_{c}^{\infty} (R^{3}; C)$ ;

(c) for every $u \in W^{1, 2} (R^{3}; C)$ the a.e.-defined function $x \mapsto u (x) / ∣ x ∣$ lies in $L^{2} (R^{3}; C)$ and satisfies $∥ u / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u ∥_{2}$ .

No constant smaller than $4$ is claimed, and no statement is made about $∥ u / ∣ \cdot ∣ ∥_{2}$ for $u$ outside $W^{1, 2}$ .

Facts & Assumptions

Given: the Sobolev space $W^{1, 2} (R^{3}; C)$ with weak gradient $\nabla u$ ; the function $r (x) = ∣ x ∣$ ; a function $u$ of one of the stated regularity classes.

[F1]

Divergence theorem and its vector forms: for a bounded domain with a finite piecewise $C^{1}$ presentation and a $C^{1}$ vector field $F$ on a neighbourhood of its closure, the integral of $div F$ equals the outward flux; in particular, if $F$ is $C^{1}$ on $R^{3} ∖ {0}$ and compactly supported there, then $\int_{R^{3}} div F = 0$ (Divergence for finite piecewise C1 presentations, Vector forms: the boundary integrals of $f n$ and of $n \times F$ ).

[F2]

Cauchy-Schwarz for $L^{2}$ : $\int ∣ f g ∣ \leq ∥ f ∥_{2} ∥ g ∥_{2}$ (Cauchy-Schwarz inequality for $L^{2}$ ).

[F3]

$C_{c}^{\infty} (R^{3}; C)$ is dense in $W^{1, 2} (R^{3}; C)$ in the Sobolev norm (Compactly supported smooth functions are dense in W^{k,p}(R^n), Integer-order Sobolev spaces and their norms).

[F4]

If $f_{n} \to f$ in $L^{2}$ , then some subsequence converges to $f$ pointwise a.e. (Complex Lp completeness and almost-everywhere subsequences).

[F5]

If $f_{n} \to f$ a.e. and $∣ f_{n} ∣ \leq g$ with $g \in L^{1}$ , then $\int f_{n} \to \int f$ (Dominated convergence).

[F6]

For $u \in C_{c}^{\infty} (R^{3}; C)$ the function $x \mapsto ∣ u (x) ∣^{2} / ∣ x ∣^{2}$ is integrable: choosing $R > 0$ with $supp u \subseteq B (0, R)$ and bounding $∣ u ∣ \leq ∥ u ∥_{\infty}$ , its integral is at most $∥ u ∥_{\infty}^{2} \int_{∣ x ∣ \leq R} ∣ x ∣^{- 2} d x = 4 π R ∥ u ∥_{\infty}^{2} < \infty$ (The space $L^{p} (μ)$ as the quotient by null functions).

[F7]

The explicit radial cutoff $χ_{0} \in C_{c}^{\infty} (R^{3})$ of Explicit compactly supported smooth cutoffs equals one on $∣ x ∣ \leq 1$ , zero on $∣ x ∣ \geq 2$ , and satisfies $0 \leq χ_{0} \leq 1$ . Thus $χ : = 1 - χ_{0}$ is smooth and bounded, vanishes near zero, equals one outside radius two, and has compactly supported bounded gradient.

Proof

1.1algebra

[The divergence of the radial weight] On $R^{3} ∖ {0}$ one has $div (x / ∣ x ∣^{2}) = \sum_{i = 1}^{3} \partial_{i} (x_{i} ∣ x ∣^{- 2}) = 3 ∣ x ∣^{- 2} - 2 ∣ x ∣^{- 2} = ∣ x ∣^{- 2}$ , because the derivative of $∣ x ∣^{- 2}$ in the direction $x$ is $- 2 ∣ x ∣^{- 4} \sum_{i} x_{i}^{2} = - 2 ∣ x ∣^{- 2}$ .

2.1step 1.1F1algebra

[Integration by parts on the punctured space] Let $u \in C_{c}^{\infty} (R^{3} ∖ {0}; C)$ and put $W (x) : = ∣ u (x) ∣^{2} x / ∣ x ∣^{2}$ . Then $W$ is $C^{1}$ on $R^{3} ∖ {0}$ and compactly supported there, and step 1.1 gives $div W = \nabla (∣ u ∣^{2}) \cdot x / ∣ x ∣^{2} + ∣ u ∣^{2} / ∣ x ∣^{2}$ . Since $\int_{R^{3}} div W = 0$ by [F1], and $\nabla (∣ u ∣^{2}) = 2 Re (\overline{u} \nabla u)$ , we obtain $\int_{R^{3}} ∣ u ∣^{2} / ∣ x ∣^{2} d x = - 2 Re \int_{R^{3}} \overline{u} (x) x \cdot \nabla u (x) ∣ x ∣^{- 2} d x$ .

3.1step 2.1F2F6algebra

[Hardy bound away from the origin] With $u$ as in step 2.1, the absolute value of the right-hand integral is at most $\int (∣ u ∣ / ∣ x ∣) ∣ \nabla u ∣ \leq ∥ u / ∣ \cdot ∣ ∥_{2} ∥ \nabla u ∥_{2}$ by [F2], while the left-hand side is $∥ u / ∣ \cdot ∣ ∥_{2}^{2}$ ; both quantities are finite by [F6]. Hence $∥ u / ∣ \cdot ∣ ∥_{2}^{2} \leq 2 ∥ u / ∣ \cdot ∣ ∥_{2} ∥ \nabla u ∥_{2}$ , and if the factor is zero the conclusion is immediate, and otherwise division by its positive value yields $∥ u / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u ∥_{2}$ , which is (a).

4.1step 3.1F5F6F7algebra

[Removing the puncture] Let $u \in C_{c}^{\infty} (R^{3}; C)$ and use the radial $χ = 1 - χ_{0} \in C^{\infty} (R^{3}; [0, 1])$ of [F7], with $χ = 0$ for $∣ x ∣ \leq 1$ , $χ = 1$ for $∣ x ∣ \geq 2$ and $∣ \nabla χ ∣ \leq C$ ; put $χ_{ε} (x) : = χ (x / ε)$ . Then $u_{ε} : = χ_{ε} u \in C_{c}^{\infty} (R^{3} ∖ {0}; C)$ , because its support is contained in the closed set ${∣ x ∣ \geq ε} \cap supp u$ and hence stays a positive distance from the origin, so step 3.1 applies: $∥ u_{ε} / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u_{ε} ∥_{2}$ . As $ε ↓ 0$ the functions $u_{ε} / ∣ x ∣$ converge to $u / ∣ x ∣$ pointwise on ${∣ x ∣ > 0}$ and are dominated by $∣ u ∣ / ∣ x ∣ \in L^{2}$ by [F6], so $∥ u_{ε} / ∣ \cdot ∣ ∥_{2} \to ∥ u / ∣ \cdot ∣ ∥_{2}$ by [F5] applied to the squared moduli. Moreover $\nabla u_{ε} = χ_{ε} \nabla u + u \nabla χ_{ε}$ with $∥ χ_{ε} \nabla u - \nabla u ∥_{2}^{2} \leq \int_{∣ x ∣ < 2 ε} ∣ \nabla u ∣^{2} \to 0$ and, with $M : = ∥ u ∥_{\infty}$ , $∥ u \nabla χ_{ε} ∥_{2}^{2} \leq C^{2} ε^{- 2} \int_{∣ x ∣ \leq 2 ε} M^{2} d x \leq C^{2} M^{2} \frac{4 π (2 ε)^{3}}{3 ε^{2}} = O (ε) \to 0$ ; hence $∥ \nabla u_{ε} ∥_{2} \to ∥ \nabla u ∥_{2}$ , and passing to the limit in the display gives (b).

5.1step 4.1F3F4algebra∎

[Density extension to $W^{1, 2}$ ] Let $u \in W^{1, 2} (R^{3}; C)$ and choose $u_{n} \in C_{c}^{\infty} (R^{3}; C)$ with $∥ u_{n} - u ∥_{W^{1, 2}} \to 0$ by [F3]. Step 4.1 applied to $u_{n} - u_{m}$ gives $∥ u_{n} / ∣ \cdot ∣ - u_{m} / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla (u_{n} - u_{m}) ∥_{2}$ , so $(u_{n} / ∣ \cdot ∣)$ is Cauchy in $L^{2}$ ; let $w \in L^{2}$ be its limit, so $∥ w ∥_{2} \leq 2 ∥ \nabla u ∥_{2}$ because $∥ \nabla u_{n} ∥_{2} \to ∥ \nabla u ∥_{2}$ . By [F4] some subsequence satisfies $u_{n_{k}} \to u$ and $u_{n_{k}} / ∣ \cdot ∣ \to w$ pointwise a.e.; on the conull set where $u_{n_{k}} \to u$ and $∣ x ∣ > 0$ we have $u_{n_{k}} (x) / ∣ x ∣ \to u (x) / ∣ x ∣$ , so $w = u / ∣ \cdot ∣$ a.e. and (c) follows.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Coulomb infinitesimal kinetic bound

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , let $H = L^{2} (R^{3}; C)$ (Hilbert space, The space $L^{p} (μ)$ as the quotient by null functions), and let $H_{0} : = - Δ$ be the free operator of Real multiplication operators and free plus bounded-potential Schrödinger operators with domain $D (H_{0}) = W^{2, 2} (R^{3}; C)$ (Integer-order Sobolev spaces and their norms): $H_{0} = F_{2}^{*} M_{4 π^{2} ∣ ξ ∣^{2}} F_{2}$ and $H_{0} u = - Δ u$ for $u \in D (H_{0})$ , where $Δ u$ is the distributional Laplacian. Write $M$ for the multiplication operator $M u : = \frac{u}{∣ \cdot ∣}, D (M) : = {u \in H : \frac{u}{∣ \cdot ∣} \in H},$ and write $∥ u ∥_{2} = ∥ u ∥_{H}$ . Then:

(a) for every $u \in D (H_{0}) = W^{2, 2} (R^{3}; C)$ one has $u / ∣ \cdot ∣ \in L^{2}$ and

$∥ \frac{u}{∣ \cdot ∣} ∥_{2} \leq 2 \sqrt{∥ u ∥_{2} ∥ Δ u ∥_{2}}, ∥ \frac{u}{∣ \cdot ∣} ∥_{2} \leq η ∥ Δ u ∥_{2} + \frac{1}{η} ∥ u ∥_{2} (η > 0);$

(b) $M$ is a self-adjoint operator on $H$ with $D (H_{0}) \subseteq D (M)$ , symmetric on $D (H_{0})$ , with $H_{0}$ -bound zero in the sense that for every $η > 0$ the pair $(η, η^{- 1})$ is admissible for $M$ relative to $H_{0}$ ;

(c) for every $γ > 0$ the operator $C : = H_{0} - γ M, D (C) : = D (H_{0}) = W^{2, 2} (R^{3}; C),$ is self-adjoint on $H$ , bounded below, and acts by $C u = - Δ u - γ u / ∣ \cdot ∣$ for every $u \in D (C)$ ; for every $η \in (0, 1 / γ)$ one has $C \geq - γ η^{- 1} / (1 - γ η)$ in the quadratic-form sense.

No sharp lower bound of $C$ is claimed here, and no physical interpretation of $C$ , $H_{0}$ , $γ$ or $Δ$ is asserted.

Facts & Assumptions

Given: the free operator $H_{0} = - Δ$ with $D (H_{0}) = W^{2, 2} (R^{3}; C)$ ; the multiplier $M$ ; the constant $γ > 0$ ; a function $u$ of the stated regularity.

[F1]

$H_{0} = F_{2}^{*} M_{4 π^{2} ∣ ξ ∣^{2}} F_{2}$ is self-adjoint on $D (H_{0}) = {f : ∣ ξ ∣^{2} F_{2} f \in L^{2}} = W^{2, 2} (R^{3}; C)$ as a set with equivalent norms, $H_{0} \geq 0$ in the quadratic-form sense, and $H_{0} f = - Δ f$ for Schwartz $f$ (Real multiplication operators and free plus bounded-potential Schrödinger operators, Integer-order W^{k,2} and H^k agree with equivalent norms, Integer-order Sobolev spaces and their norms).

[F2]

For every $u \in W^{1, 2} (R^{3}; C)$ one has $u / ∣ \cdot ∣ \in L^{2}$ and $∥ u / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u ∥_{2}$ (Three-dimensional Hardy bound).

[F3]

Plancherel: $F_{2}$ is unitary and for the weak derivatives of $u \in W^{2, 2}$ the multipliers satisfy $F_{2} (\partial_{i} u) (ξ) = 2 π i ξ_{i} F_{2} u (ξ)$ and $F_{2} (Δ u) (ξ) = - 4 π^{2} ∣ ξ ∣^{2} F_{2} u (ξ)$ a.e.; equivalently $∥ \nabla u ∥_{2}^{2} = \int 4 π^{2} ∣ ξ ∣^{2} ∣ F_{2} u ∣^{2}$ and $∥ Δ u ∥_{2}^{2} = \int 16 π^{4} ∣ ξ ∣^{4} ∣ F_{2} u ∣^{2}$ (Plancherel theorem, Distributional derivatives are polynomial Fourier multipliers, Integer-order W^{k,2} and H^k agree with equivalent norms).

[F4]

Cauchy-Schwarz: $\int ∣ f g ∣ \leq ∥ f ∥_{2} ∥ g ∥_{2}$ (Cauchy-Schwarz inequality for $L^{2}$ ).

[F5]

Kato-Rellich: if $A$ is self-adjoint, $B$ symmetric, $D (A) \subseteq D (B)$ and the $A$ -bound of $B$ is less than one with admissible pair $(a, b)$ , then $A + B$ on $D (A)$ is self-adjoint and, when $A \geq γ_{0}$ , satisfies $A + B \geq γ_{0} - \max {a ∣ γ_{0} ∣ + b, b / (1 - a)}$ (Kato-Rellich theorem, Relative boundedness with respect to an operator).

[F6]

If $a$ is a real finite-valued measurable function on $R^{3}$ , then $M_{a} f : = a f$ on ${f \in H : a f \in H}$ is self-adjoint, and it is symmetric on every subspace on which both sides of the pairing are integrable, in particular on the set of $f$ with $a f \in H$ (Real multiplication operators and free plus bounded-potential Schrödinger operators (a), Adjoint of a densely defined operator, Symmetric, self-adjoint and essentially self-adjoint operators).

Proof

1.1F3F4algebra

[Kinetic inequality] Let $u \in W^{2, 2} (R^{3}; C)$ and put $g : = F_{2} u$ . By [F3], $∥ \nabla u ∥_{2}^{2} = \int 4 π^{2} ∣ ξ ∣^{2} ∣ g (ξ) ∣^{2} d ξ$ and $∥ Δ u ∥_{2}^{2} = \int (4 π^{2} ∣ ξ ∣^{2})^{2} ∣ g (ξ) ∣^{2} d ξ$ , while $∥ u ∥_{2}^{2} = \int ∣ g ∣^{2}$ . Writing $4 π^{2} ∣ ξ ∣^{2} ∣ g ∣^{2} = (∣ g ∣) (4 π^{2} ∣ ξ ∣^{2} ∣ g ∣)$ and applying [F4] to the two factors gives $∥ \nabla u ∥_{2}^{2} \leq ∥ u ∥_{2} ∥ Δ u ∥_{2}$ .

2.1step 1.1F2algebra

[Infinitesimal Coulomb bound] Combining step 1.1 with [F2] yields $∥ u / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla u ∥_{2} \leq 2 \sqrt{∥ u ∥_{2} ∥ Δ u ∥_{2}}$ . For $η > 0$ the arithmetic-geometric mean inequality $2 \sqrt{a b} \leq η a + η^{- 1} b$ applied to $a = ∥ Δ u ∥_{2}$ , $b = ∥ u ∥_{2}$ gives $2 \sqrt{∥ u ∥_{2} ∥ Δ u ∥_{2}} \leq η ∥ Δ u ∥_{2} + η^{- 1} ∥ u ∥_{2}$ . This proves (a).

3.1step 2.1F1F3F6algebra

[Relative bound of the multiplier] By [F6] applied to $a (x) = 1 / ∣ x ∣$ (finite-valued on $R^{3} ∖ {0}$ and arbitrary, say $0$ , at the origin, a null set) the operator $M$ is self-adjoint on $D (M) = {f : f / ∣ \cdot ∣ \in H}$ , and $M$ is symmetric on $D (M)$ . Step 2.1 shows $D (H_{0}) = W^{2, 2} \subseteq D (M)$ and $∥ M u ∥_{2} = ∥ u / ∣ \cdot ∣ ∥_{2} \leq η ∥ Δ u ∥_{2} + η^{- 1} ∥ u ∥_{2} = η ∥ H_{0} u ∥_{2} + η^{- 1} ∥ u ∥_{2}$ for every $u \in D (H_{0})$ , since $Δ u = - H_{0} u$ on $D (H_{0})$ by [F1] and [F3]. Hence $(η, η^{- 1})$ is an admissible pair for $M$ relative to $H_{0}$ , and the $H_{0}$ -bound of $M$ is zero; this is (b).

4.1step 3.1F1F3F5algebra∎

[Self-adjointness of the Coulomb operator] Fix $γ > 0$ and let $η \in (0, 1 / γ)$ . Then $B : = - γ M$ is symmetric with $D (H_{0}) \subseteq D (B)$ and admissible pair $(a, b) = (γ η, γ η^{- 1})$ with $a < 1$ ; indeed $∥ B u ∥_{2} = γ ∥ M u ∥_{2} \leq γ η ∥ H_{0} u ∥_{2} + γ η^{- 1} ∥ u ∥_{2}$ by step 3.1. By [F5] with $A = H_{0} \geq 0$ (so $γ_{0} = 0$ ) the operator $C = H_{0} + B$ on $D (H_{0})$ is self-adjoint with $C \geq - \max {0, b / (1 - a)} = - γ η^{- 1} / (1 - γ η)$ . Moreover $C u = H_{0} u - γ u / ∣ \cdot ∣ = - Δ u - γ u / ∣ \cdot ∣$ for $u \in D (H_{0})$ , by [F1] and [F3]; this is (c).

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Coulomb relative compactness

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , let $H_{0} = - Δ$ with $D (H_{0}) = W^{2, 2} (R^{3}; C)$ be the free operator of Real multiplication operators and free plus bounded-potential Schrödinger operators, let $M u = u / ∣ \cdot ∣$ be the multiplier of Coulomb infinitesimal kinetic bound with $D (M) \supseteq D (H_{0})$ , and let $V : = - γ M$ , so that $V f = - γ f / ∣ \cdot ∣$ . For $z \notin R$ write $R_{0} (z) : = (H_{0} - z)^{- 1}$ (Resolvent and spectrum of an unbounded operator, Resolvent of a self-adjoint operator: nonreal resolvents and the estimate). Then:

(a) for every $z \notin R$ there holds $R_{0} (z) = F_{2}^{*} M_{m_{z}} F_{2}$ with $m_{z} (ξ) : = (4 π^{2} ∣ ξ ∣^{2} - z)^{- 1} \in L^{2} (R^{3})$ , and $∥ R_{0} (z) f ∥_{\infty} \leq ∥ m_{z} ∥_{2} ∥ f ∥_{2}$ for every $f \in L^{2} (R^{3})$ ;

(b) for every $z \notin R$ there is $g_{z} \in L^{2} (R^{3})$ with $R_{0} (z) f = g_{z} * f$ for every $f \in L^{2}$ ;

(c) for every bounded measurable $χ$ with compact support the operator $M_{χ} R_{0} (z)$ is Hilbert–Schmidt, hence compact;

(d) $V$ is $H_{0}$ -compact (Relative compactness with respect to an operator): $V$ is graph-norm bounded with respect to $H_{0}$ and $V R_{0} (z)$ is a compact operator for one, equivalently every, $z \notin R$ ; explicitly $V R_{0} (z) = - γ M R_{0} (z)$ is compact.

No statement is made about compactness of $M$ itself, about point spectrum or about any other spectral property of $H_{0} + V$ ; no physical interpretation is asserted.

Facts & Assumptions

Given: the self-adjoint free operator $H_{0} = - Δ$ with $D (H_{0}) = W^{2, 2} (R^{3}; C)$ ; the multiplier $M$ ; the number $γ > 0$ ; a nonreal $z$ ; a bounded compactly supported $χ$ .

[F1]

$H_{0} = F_{2}^{*} M_{4 π^{2} ∣ ξ ∣^{2}} F_{2}$ is self-adjoint with $D (H_{0}) = {f : ∣ ξ ∣^{2} F_{2} f \in L^{2}} = W^{2, 2} (R^{3}; C)$ (as a set, with equivalent norms), and $H_{0} \geq 0$ (Real multiplication operators and free plus bounded-potential Schrödinger operators, Relative boundedness with respect to an operator, Symmetric, self-adjoint and essentially self-adjoint operators).

[F2]

$M$ is self-adjoint with $D (H_{0}) \subseteq D (M)$ and $H_{0}$ -bound zero: for every $η > 0$ , $∥ M f ∥_{2} \leq η ∥ H_{0} f ∥_{2} + η^{- 1} ∥ f ∥_{2}$ (Coulomb infinitesimal kinetic bound (b), Relative boundedness with respect to an operator).

[F3]

$F_{2}$ is unitary on $L^{2}$ and extends the $L^{1}$ transform $\hat{f} (ξ) = \int f (x) e^{- 2 π i x \cdot ξ} d x$ (Plancherel theorem, Fourier transform on complex L1 classes).

[F4]

For Schwartz $f, g$ the convolution $f * g$ is Schwartz and $F_{2} (f * g) = \hat{f} \hat{g}$ (Schwartz convolution and product laws, Schwartz space and its seminorms).

[F5]

If $k \in L^{2} (R^{3} \times R^{3}; C)$ then the kernel operator $T_{k}$ of L^2 kernel operators and Hilbert-Schmidt operators on R^3 is defined on all of $L^{2}$ , is bounded with $∥ T_{k} f ∥_{2} \leq ∥ k ∥_{2} ∥ f ∥_{2}$ , and is compact; in particular it is a Hilbert–Schmidt operator in the sense of that definition (L^2 kernel operators are bounded and compact).

[F6]

For nonnegative product-measurable functions the iterated and double integrals agree (Tonelli's theorem for nonnegative measurable functions on a sigma-finite product).

[F7]

Cauchy-Schwarz and dominated convergence in their $L^{2}$ and integral forms (Cauchy-Schwarz inequality for $L^{2}$ , Dominated convergence).

[F8]

If $T$ is compact and $B$ is bounded then $B T$ is compact (Compositions with a compact operator are compact, Compact linear operator).

[F9]

A uniform limit of compact operators is compact (Norm limit of compact operators is compact).

[F10]

The Schwartz classes are dense in $L^{2} (R^{3}; C)$ (Schwartz space is dense in L2).

Proof

1.1F1F3givenalgebra

[The resolvent as a Fourier multiplier] Fix $z \notin R$ and define $m_{z} (ξ) : = (4 π^{2} ∣ ξ ∣^{2} - z)^{- 1}$ . Since $∣ 4 π^{2} ∣ ξ ∣^{2} - z ∣ \geq ∣ Im z ∣$ one has $∣ m_{z} (ξ) ∣ \leq ∣ Im z ∣^{- 1}$ everywhere; and whenever $4 π^{2} ∣ ξ ∣^{2} \geq 2 ∣ Re z ∣$ one has $∣ 4 π^{2} ∣ ξ ∣^{2} - z ∣ \geq 4 π^{2} ∣ ξ ∣^{2} - ∣ Re z ∣ \geq \frac{1}{2} 4 π^{2} ∣ ξ ∣^{2}$ , so that $∣ m_{z} (ξ) ∣ \leq \frac{1}{2 π^{2} ∣ ξ ∣^{2}}$ . With $ρ : = \max {1, \sqrt{2 ∣ Re z ∣} / (2 π)} > 0$ , so that $4 π^{2} ρ^{2} \geq 2 ∣ Re z ∣$ this yields $\int_{R^{3}} ∣ m_{z} ∣^{2} \leq ∣ Im z ∣^{- 2} ∣ B (0, ρ) ∣ + \frac{1}{4 π^{4}} \int_{∣ ξ ∣ > ρ} ∣ ξ ∣^{- 4} d ξ < \infty$ , because $\int_{∣ ξ ∣ > ρ} ∣ ξ ∣^{- 4} d ξ = 4 π \int_{ρ}^{\infty} r^{- 2} d r$ ; so $m_{z} \in L^{2} (R^{3})$ ; moreover $m_{z} (ξ) (4 π^{2} ∣ ξ ∣^{2} - z) = 1$ for a.e. $ξ$ . Define $R_{0} (z) : = F_{2}^{*} M_{m_{z}} F_{2}$ ; then for $f \in L^{2}$ the vector $R_{0} (z) f$ lies in $D (H_{0})$ , because $∣ ξ ∣^{2} m_{z} (ξ) F_{2} f (ξ) = (F_{2} f (ξ) + z m_{z} (ξ) F_{2} f (ξ)) / (4 π^{2})$ is a sum of two $L^{2}$ functions, and $(H_{0} - z) R_{0} (z) f = f$ since $F_{2} ((H_{0} - z) R_{0} (z) f) = (4 π^{2} ∣ ξ ∣^{2} - z) m_{z} (ξ) F_{2} f = F_{2} f$ ; conversely for $g \in D (H_{0})$ the function $(4 π^{2} ∣ ξ ∣^{2} - z) F_{2} g$ is in $L^{2}$ , so $R_{0} (z) (H_{0} - z) g$ is defined and its transform is $m_{z} (ξ) (4 π^{2} ∣ ξ ∣^{2} - z) F_{2} g = F_{2} g$ . Hence $R_{0} (z) = (H_{0} - z)^{- 1}$ and $∥ R_{0} (z) ∥ \leq ∣ Im z ∣^{- 1}$ .

2.1step 1.1F3F4F7algebra

[ $L^{\infty}$ bound of the free resolvent] For Schwartz $f$ the function $m_{z} F_{2} f$ lies in $L^{1} \cap L^{2}$ , so $R_{0} (z) f$ has the everywhere-defined representative $x \mapsto \int m_{z} (ξ) F_{2} f (ξ) e^{2 π i x \cdot ξ} d ξ$ , whose modulus is at most $∥ m_{z} ∥_{2} ∥ F_{2} f ∥_{2} = ∥ m_{z} ∥_{2} ∥ f ∥_{2}$ by [F7]. For general $f \in L^{2}$ choose Schwartz $f_{n} \to f$ in $L^{2}$ ; then $R_{0} (z) f_{n} \to R_{0} (z) f$ in $L^{2}$ and $∥ R_{0} (z) f_{n} - R_{0} (z) f_{m} ∥_{\infty} \leq ∥ m_{z} ∥_{2} ∥ f_{n} - f_{m} ∥_{2}$ , so the representatives converge uniformly to a bounded function $h$ with $∥ h ∥_{\infty} \leq ∥ m_{z} ∥_{2} ∥ f ∥_{2}$ ; uniform convergence gives $L_{l o c}^{2}$ convergence and hence $h = R_{0} (z) f$ a.e., so $∥ R_{0} (z) f ∥_{\infty} \leq ∥ m_{z} ∥_{2} ∥ f ∥_{2}$ .

2.2step 1.1F3F4F6F7F10algebra

[Convolution kernel] Put $g_{z} : = F_{2}^{*} m_{z} \in L^{2}$ by [F3]. We first prove $R_{0} (z) f = g_{z} * f$ for Schwartz $f$ . Choose Schwartz functions $φ_{n}$ with $φ_{n} \to g_{z}$ in $L^{2}$ ([F10]). For every $n$ both $φ_{n}$ and $f$ are Schwartz, so [F4] gives $F_{2} (φ_{n} * f) = \hat{φ_{n}} \hat{f}$ ; here $\hat{φ_{n}} \to F_{2} g_{z} = m_{z}$ in $L^{2}$ by the isometry [F3], while $∣ \hat{f} (ξ) ∣ \leq ∥ f ∥_{1} < \infty$ for every $ξ$ by the defining integral of the $L^{1}$ transform [F3] and the integrability of the Schwartz function $f$ ; hence $F_{2} (φ_{n} * f) = \hat{φ_{n}} \hat{f} \to m_{z} \hat{f} = F_{2} (R_{0} (z) f)$ in $L^{2}$ . Since $F_{2}$ is an isometry, $φ_{n} * f \to R_{0} (z) f$ in $L^{2}$ . On the other hand $φ_{n} * f \to g_{z} * f$ in $L^{2}$ : for $h : = φ_{n} - g_{z}$ and a.e. $x$ , $\int ∣ h (x - y) ∣ ∣ f (y) ∣ d y = \int ∣ h (x - y) ∣ \sqrt{∣ f (y) ∣} \sqrt{∣ f (y) ∣} d y \leq (\int ∣ h (x - y) ∣^{2} ∣ f (y) ∣ d y)^{1 / 2} ∥ f ∥_{1}^{1 / 2}$ by [F7], so Tonelli [F6] gives $∥ h * f ∥_{2}^{2} \leq ∥ f ∥_{1} \iint ∣ h (x - y) ∣^{2} ∣ f (y) ∣ d y d x = ∥ f ∥_{1}^{2} ∥ h ∥_{2}^{2} \to 0$ . Therefore $R_{0} (z) f = g_{z} * f$ for Schwartz $f$ . This identity also gives $∥ g_{z} * f ∥_{2} = ∥ R_{0} (z) f ∥_{2} = ∥ m_{z} F_{2} f ∥_{2} \leq ∥ m_{z} ∥_{\infty} ∥ f ∥_{2}$ for Schwartz $f$ , so $f \mapsto g_{z} * f$ is bounded on the dense subspace of Schwartz classes and extends uniquely to a bounded operator on $L^{2}$ by [F10], and this extension agrees with the bounded operator $R_{0} (z)$ on the dense subspace; hence the extension equals $R_{0} (z)$ on every $f \in L^{2}$ . It also equals the ordinary convolution representative: for fixed $x$ , Cauchy–Schwarz gives $\int ∣ g_{z} (x - y) f (y) ∣ d y \leq ∥ g_{z} ∥_{2} ∥ f ∥_{2}$ , and for Schwartz $f_{n} \to f$ it gives $\sup_{x} ∣ g_{z} * (f_{n} - f) (x) ∣ \leq ∥ g_{z} ∥_{2} ∥ f_{n} - f ∥_{2} \to 0$ . This is uniform convergence, while $R_{0} (z) f_{n} \to R_{0} (z) f$ in $L^{2}$ . On every bounded set both limits agree by $L^{2}$ convergence and uniform convergence, proving $R_{0} (z) f = g_{z} * f$ almost everywhere for the ordinary convolution as well.

3.1step 2.2F5F6algebra

[Localized resolvent is Hilbert–Schmidt] Let $χ$ be bounded with compact support. By step 2.2, $M_{χ} R_{0} (z) f = χ \cdot (g_{z} * f)$ , so the operator $M_{χ} R_{0} (z)$ has the kernel $k (x, y) : = χ (x) g_{z} (x - y)$ . The function $k$ is product-measurable and $\int \int ∣ k (x, y) ∣^{2} d x d y = \int ∣ χ (x) ∣^{2} (\int ∣ g_{z} (x - y) ∣^{2} d y) d x = ∥ χ ∥_{2}^{2} ∥ g_{z} ∥_{2}^{2} < \infty$ by [F6]; hence $M_{χ} R_{0} (z) = T_{k}$ is Hilbert–Schmidt in the sense of L^2 kernel operators and Hilbert-Schmidt operators on R^3 and compact by [F5].

3.2step 2.1F2F3givenalgebra

[Short-range and tail estimates] For $ε > 0$ let $B_{ε} : = {x : ∣ x ∣ < ε}$ . By step 2.1, $∥ M_{γ ∣ \cdot ∣^{- 1} 1_{B_{ε}}} R_{0} (z) f ∥_{2} \leq γ (\int_{B_{ε}} ∣ x ∣^{- 2} d x)^{1 / 2} ∥ R_{0} (z) f ∥_{\infty} \leq γ \sqrt{4 π ε} ∥ m_{z} ∥_{2} ∥ f ∥_{2}$ , so the operator norm is at most $γ \sqrt{4 π ε} ∥ m_{z} ∥_{2}$ . For $R > 0$ , $∥ M_{γ ∣ \cdot ∣^{- 1} 1_{{∣ x ∣ > R}}} R_{0} (z) ∥ \leq γ R^{- 1} ∥ R_{0} (z) ∥ \leq γ R^{- 1} ∣ Im z ∣^{- 1}$ .

4.1step 3.1step 3.2F9algebra

[Compactness of $M R_{0} (z)$ ] Fix $R > ε > 0$ and choose a bounded measurable $χ$ with $0 \leq χ \leq 1$ , $χ = 1$ on ${ε \leq ∣ x ∣ \leq R}$ and $χ = 0$ off ${ε / 2 \leq ∣ x ∣ \leq 2 R}$ ; this is possible with compact support. Since $χ = 1$ on the closed annulus ${ε \leq ∣ x ∣ \leq R}$ , the function $1 - χ$ vanishes there and satisfies the pointwise domination $0 \leq γ \frac{1 - χ (x)}{∣ x ∣} \leq γ \frac{1_{B_{ε}} (x)}{∣ x ∣} + γ \frac{1_{{∣ x ∣ > R}} (x)}{∣ x ∣}$ . Hence $γ M R_{0} (z) = M_{γ ∣ \cdot ∣^{- 1} χ} R_{0} (z) + M_{γ ∣ \cdot ∣^{- 1} (1 - χ)} R_{0} (z)$ , where the first summand is compact by step 3.1 and the second has operator norm at most $γ \sqrt{4 π ε} ∥ m_{z} ∥_{2} + γ R^{- 1} ∣ Im z ∣^{- 1}$ by step 3.2, because its multiplier is pointwise dominated by the sum of the two multipliers of step 3.2. Given $δ > 0$ choose $ε$ and $R$ making this bound smaller than $δ$ . Thus $γ M R_{0} (z)$ lies within $δ$ of a compact operator for every $δ > 0$ , hence is compact by [F9].

5.1step 4.1F2F8∎

[Conclusion] By step 4.1 and [F8], $V R_{0} (z) = - γ M R_{0} (z)$ is compact; the graph-norm boundedness of $V$ with respect to $H_{0}$ is [F2]. Hence $V$ is $H_{0}$ -compact in the sense of Relative compactness with respect to an operator, and the equivalent formulations in (a)–(c) are steps 1.1, 2.1, 2.2 and 3.1.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Coulomb essential spectrum

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , let $H_{0} = - Δ$ with $D (H_{0}) = W^{2, 2} (R^{3}; C)$ be the free operator of Real multiplication operators and free plus bounded-potential Schrödinger operators, let $M$ be the multiplier $M u = u / ∣ \cdot ∣$ of Coulomb infinitesimal kinetic bound, and let $C : = H_{0} - γ M, D (C) = W^{2, 2} (R^{3}; C),$ be the self-adjoint operator of Coulomb infinitesimal kinetic bound (c). Write $σ_{d}$ , $σ_{e s s}$ , $E$ for the discrete spectrum, essential spectrum and spectral projection valued measure of Discrete and essential spectrum of a self-adjoint operator and Spectral theorem for unbounded self-adjoint operators (PVM form). Then:

(a) $C$ is bounded below and $σ_{e s s} (C) = [0, + \infty)$ ;

(b) every point of $σ (C) \cap (- \infty, 0)$ lies in $σ_{d} (C)$ : it is an eigenvalue of $C$ of finite multiplicity isolated in $σ (C)$ , and the set of such eigenvalues is at most countable;

(c) the negative spectral subspace $E ((- \infty, 0)) L^{2} (R^{3}; C)$ is the closed linear span of the eigenspaces of $C$ for its negative eigenvalues.

No claim is made about the multiplicity or the number of the negative eigenvalues, about the nonnegative part of the spectrum beyond $σ_{e s s} (C) = [0, \infty)$ , or about any physical interpretation.

Facts & Assumptions

Given: the operators $H_{0}$ , $M$ , $C$ ; a nonreal number is used only inside the proof; $γ > 0$ .

[F1]

$H_{0} = F_{2}^{*} M_{4 π^{2} ∣ ξ ∣^{2}} F_{2}$ is self-adjoint with $D (H_{0}) = W^{2, 2} (R^{3}; C)$ and $H_{0} \geq 0$ ; for $u \in D (H_{0})$ one has $F_{2} (H_{0} u) (ξ) = 4 π^{2} ∣ ξ ∣^{2} F_{2} u (ξ)$ a.e. (Real multiplication operators and free plus bounded-potential Schrödinger operators, Distributional derivatives are polynomial Fourier multipliers).

[F2]

$C = H_{0} - γ M$ is self-adjoint on $D (C) = D (H_{0})$ and bounded below; $M$ is $H_{0}$ -bounded with bound zero and $M R_{0} (z)$ is compact for every nonreal $z$ , where $R_{0} (z) = (z - H_{0})^{- 1}$ (Coulomb infinitesimal kinetic bound, Coulomb relative compactness, Relative compactness with respect to an operator, Resolvent of a self-adjoint operator: nonreal resolvents and the estimate).

[F3]

Weyl's theorem: if $A, C$ are self-adjoint and $R_{A} (z) - R_{C} (z)$ is compact for one nonreal $z$ , then $σ_{e s s} (A) = σ_{e s s} (C)$ (Weyl's theorem: invariance of the essential spectrum).

[F4]

For a self-adjoint operator $A$ with PVM $E$ : $λ \in σ_{e s s} (A)$ if and only if $rank E ((λ - ε, λ + ε)) = \infty$ for every $ε > 0$ ; for every Borel set $B$ and vector $f$ the scalar set function $E_{f} (B) : = ⟨ E (B) f, f ⟩$ is a finite positive measure of mass $∥ f ∥^{2}$ ; and $E ({λ})$ is the orthogonal projection onto $\ker (A - λ)$ (Discrete and essential spectrum of a self-adjoint operator, Projection valued measure, Unbounded Borel functional calculus: domains, products, spectral mapping).

[F5]

If $T$ is compact and $B$ is bounded, then $B T$ and $T B$ are compact (Compositions with a compact operator are compact, Compact linear operator).

[F6]

Normalized indicator functions of pairwise disjoint measurable sets of finite positive measure form an orthonormal family; a nonempty open ball in $R^{3}$ has positive measure and contains infinitely many pairwise disjoint nonempty open balls (Orthonormal families, complete orthonormal systems and Hilbert bases, The space $L^{p} (μ)$ as the quotient by null functions).

Proof

1.1F1F4F6algebra

[Essential spectrum of the free operator] Fix $λ \geq 0$ and $ε > 0$ and put $S_{ε} : = {ξ \in R^{3} : 4 π^{2} ∣ ξ ∣^{2} \in (λ - ε, λ + ε)}$ . If $λ > 0$ choose $ξ_{0}$ with $4 π^{2} ∣ ξ_{0} ∣^{2} = λ$ ; continuity of $ξ \mapsto 4 π^{2} ∣ ξ ∣^{2}$ gives a nonempty open ball around $ξ_{0}$ contained in $S_{ε}$ . If $λ = 0$ take the ball ${∣ ξ ∣ < \sqrt{ε} / (2 π)}$ . In either case $S_{ε}$ contains a nonempty open ball, hence by [F6] contains infinitely many pairwise disjoint nonempty open balls and has infinite-dimensional $L^{2}$ space; by [F1] the projection $E_{H_{0}} ((λ - ε, λ + ε)) = F_{2}^{*} M_{1_{S_{ε}}} F_{2}$ has range $F_{2}^{*} M_{1_{S_{ε}}} L^{2}$ , which is infinite dimensional. By [F4] applied to $H_{0}$ we get $λ \in σ_{e s s} (H_{0})$ . Conversely, if $λ < 0$ and $0 < ε < ∣ λ ∣$ , then $(λ - ε, λ + ε) \cap [0, \infty) = \emptyset$ , so $S_{ε} = \emptyset$ and $E_{H_{0}} ((λ - ε, λ + ε)) = 0$ ; since the projection vanishes on an open interval around $λ$ , [F4] shows $λ \notin σ (H_{0})$ . Hence $σ_{e s s} (H_{0}) = [0, \infty)$ .

2.1step 1.1F2F5algebra

[The resolvent difference is compact] Fix a nonreal $z$ and use the convention $R_{A} (z) = (z - A)^{- 1}$ of Resolvent and spectrum of an unbounded operator. Both resolvents are bounded by [F2] and have range $D (C) = D (H_{0})$ . For $f \in L^{2}$ put $g = R_{H_{0}} (z) f$ ; then $(z - H_{0}) g = f$ and $(z - C) g = f + γ M g$ . Applying $R_{C} (z)$ gives $g = R_{C} (z) f + γ R_{C} (z) M g$ , hence $R_{C} (z) - R_{H_{0}} (z) = - γ R_{C} (z) M R_{H_{0}} (z)$ . The factor $M R_{H_{0}} (z)$ is compact by [F2] (changing the sign of the free resolvent preserves compactness), and $R_{C} (z)$ is bounded; therefore this difference is compact by [F5].

3.1step 2.1step 1.1F2F3

[Weyl's theorem and the essential spectrum] By step 2.1 the resolvent difference of $C$ and $H_{0}$ is compact at every nonreal $z$ , so [F3] gives $σ_{e s s} (C) = σ_{e s s} (H_{0}) = [0, \infty)$ by step 1.1. The boundedness below of $C$ is [F2]. This is (a).

4.1step 3.1F4algebra

[Negative spectrum is discrete] The spectrum of the self-adjoint operator $C$ is the disjoint union of $σ_{d} (C)$ and $σ_{e s s} (C) = [0, \infty)$ by [F4]; hence every $λ \in σ (C) \cap (- \infty, 0)$ lies in $σ_{d} (C)$ , which by definition means that $λ$ is an eigenvalue of $C$ , is isolated in $σ (C)$ and has finite-dimensional eigenspace. Each such $λ$ admits a rational open interval $(p, q)$ containing $λ$ whose intersection with the spectrum is exactly ${λ}$ . Fix an enumeration of ${(p, q) \in Q^{2} : p < q}$ and assign to $λ$ its least eligible pair. Two different spectral points cannot receive the same pair, because that interval has exactly one spectral point. This is an injection into a countable set, so the negative eigenvalue set is at most countable; no arbitrary choices of overlapping right endpoints are used. This is (b).

5.1step 4.1F4algebra∎

[Completeness of the negative eigenspaces] Let $S$ be the closed linear span of the eigenspaces of $C$ for negative eigenvalues and let $P_{-} : = E ((- \infty, 0))$ . Every eigenvector of $C$ with negative eigenvalue $λ$ satisfies $E ({λ}) u = u$ by [F4], hence lies in $P_{-} L^{2}$ because ${λ} \subseteq (- \infty, 0)$ ; therefore $S \subseteq P_{-} L^{2}$ . For the reverse inclusion let $f \in P_{-} L^{2}$ be orthogonal to $S$ and put $μ : = E_{f}$ . Since $f = P_{-} f$ and $E_{f} (B) = ∥ E (B) f ∥^{2}$ , the measure $μ$ is concentrated on $(- \infty, 0)$ , i.e. $μ (R) = μ ((- \infty, 0))$ , and by step 4.1 the set ${λ_{1}, λ_{2}, \dots} = σ (C) \cap (- \infty, 0)$ is at most countable and consists of eigenvalues. By [F4] the support of $μ$ is contained in $σ (C)$ : if $B \cap σ (C) = \emptyset$ then $E (B) = 0$ . Hence $μ (R) = μ ({λ_{n}}) = \sum_{n} μ ({λ_{n}}) = \sum_{n} ∥ E ({λ_{n}}) f ∥^{2}$ , and each term vanishes because $E ({λ_{n}}) f$ is an eigenvector with negative eigenvalue and $f ⊥ S$ . Therefore $μ = 0$ , so $∥ f ∥^{2} = 0$ and $f = 0$ ; thus $P_{-} L^{2} \subseteq S$ and (c) holds.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Angular radial unitary reduction

Statement

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Let $S^{2} = {x \in R^{3} : ∣ x ∣ = 1}$ , let $σ$ be the polar surface measure on $S^{2}$ (The polar surface set function on the unit sphere), and let $(Y_{l m})_{l \geq 0, - l \leq m \leq l}$ be the complete orthonormal family of $L^{2} (S^{2}, σ)$ of Sphere ONB. For pointwise formulas use the representatives ${\tilde{Y}}_{l m}$ of Associated Legendre function normalization, and write $Y_{l m} (Ω)$ for their values. Explicitly, with $P_{l}$ and $b_{l m}$ as in that supplier, for $0 \leq m \leq l$ , ${\tilde{Y}}_{l m} (Ω) = \frac{(- 1)^{m} b_{l m}}{\sqrt{2 π}} (Ω_{1} + i Ω_{2})^{m} P_{l}^{(m)} (Ω_{3}), {\tilde{Y}}_{l, - m} = (- 1)^{m} \overline{{\tilde{Y}}_{l m}} (1 \leq m \leq l) .$ These are restrictions of polynomials in the Cartesian coordinates, hence smooth on the entire sphere, and represent the stated $L^{2}$ classes. Write $L^{2} = L^{2} (R^{3}; C)$ with Lebesgue measure and inner product linear in the first argument (The space $L^{p} (μ)$ as the quotient by null functions, Hilbert space), and let $K : = ⨁_{(l, m)} L^{2} ((0, \infty), d r; C)$ be the Hilbert direct sum over $l \geq 0$ , $- l \leq m \leq l$ . For $ψ \in L^{2}$ and each $(l, m)$ define the coefficient function and the reduced function $c_{l m} (r) : = \int_{S^{2}} ψ (r Ω) \overline{Y_{l m} (Ω)} d σ (Ω) (r > 0, a.e.), u_{l m} (r) : = r c_{l m} (r) .$ Then:

(a) for every $ψ \in L^{2}$ and every $(l, m)$ the function $u_{l m}$ is a well-defined element of $L^{2} ((0, \infty), d r)$ depending only on the class of $ψ$ , and $U ψ : = (u_{l m})_{l, m}$ is a linear isometry of $L^{2}$ into (indeed onto) $K$ with $∥ U ψ ∥_{K}^{2} = \sum_{l \geq 0} \sum_{m = - l}^{l} \int_{0}^{\infty} ∣ u_{l m} (r) ∣^{2} d r = ∥ ψ ∥_{L^{2}}^{2};$

(b) $U$ is unitary, and for every $(u_{l m}) \in K$ the finite-subset sums $\sum_{l \leq N, ∣ m ∣ \leq l} u_{l m} (r) Y_{l m} (Ω) / r$ converge in $L^{2}$ to a function $ψ$ with $U ψ = (u_{l m})$ ;

(c) on $R^{3} ∖ {0}$ every $f \in C^{2}$ satisfies the following identity. Here radial derivatives differentiate $F (r, Ω) = f (r Ω)$ with $Ω$ fixed, and in angle coordinates $Δ_{S^{2}} F : = \partial_{θ}^{2} F + \cot θ \partial_{θ} F + \sin^{- 2} θ \partial_{φ}^{2} F$ , extended continuously to the poles by the Cartesian identity proved below. For the smooth representatives of $Y_{l m}$ this agrees with the negative of the diagonal sphere operator in Sphere Laplacian diagonal domain:

$Δ f = \partial_{r}^{2} f + \frac{2}{r} \partial_{r} f + \frac{1}{r^{2}} Δ_{S^{2}} f;$

(d) for every $l \geq 0$ , $m$ , and $u \in C_{c}^{\infty} ((0, \infty); C)$ , the function $ψ : = u (r) Y_{l m} (Ω) / r$ lies in $C_{c}^{\infty} (R^{3} ∖ {0}) \subseteq H^{2} (R^{3}; C)$ , its image under $U$ is the finite-support vector with $(l, m)$ -entry $u$ , and $- Δ ψ = \frac{1}{r} (- u^{''} (r) + \frac{l (l + 1)}{r^{2}} u (r)) Y_{l m} (Ω), \int_{R^{3}} ∣ \nabla ψ ∣^{2} d x = \int_{0}^{\infty} (∣ u^{'} (r) - \frac{u (r)}{r} ∣^{2} + \frac{l (l + 1)}{r^{2}} ∣ u (r) ∣^{2}) d r .$

All statements are purely mathematical assertions about $L^{2}$ spaces, the surface measure and the Laplacian; no physical interpretation and no statement about operators with unbounded potentials is included.

Facts & Assumptions

Given: the sphere $S^{2}$ with polar measure $σ$ and the complete orthonormal family $(Y_{l m})$ ; a function $ψ \in L^{2} (R^{3})$ ; a function $f \in C^{2} (R^{3} ∖ {0})$ ; indices $l, m$ and $u \in C_{c}^{\infty} ((0, \infty))$ .

[F1]

Angle coordinates: $\int_{S^{2}} G d σ = \int_{(0, π) \times (0, 2 π)} G (Φ (θ, φ)) \sin θ d (θ, φ)$ for Borel $G \geq 0$ , and $σ (S^{2}) = 4 π$ (Sphere integrals in angle coordinates, The polar surface set function on the unit sphere).

[F2]

$(Y_{l m})$ is a complete orthonormal family of $L^{2} (S^{2}, σ)$ : for every $F \in L^{2} (S^{2}, σ)$ , $F = \sum_{l, m} ⟨ F, Y_{l m} ⟩ Y_{l m}$ in the finite-subset net and $∥ F ∥^{2} = \sum_{l, m} ∣ ⟨ F, Y_{l m} ⟩ ∣^{2}$ (Sphere ONB, Parseval equivalences for an orthonormal family, Orthonormal families, complete orthonormal systems and Hilbert bases).

[F3]

Polar coordinates: $\int_{R^{3}} g d x = \int_{0}^{\infty} r^{2} \int_{S^{2}} g (r Ω) d σ (Ω) d r$ for every Borel $g \geq 0$ (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma); iterated integrals of nonnegative completed-product-measurable functions may be interchanged (Tonelli's theorem for nonnegative measurable functions on a sigma-finite product, Tonelli and Fubini for the completed product, with only almost-everywhere section measurability).

[F4]

The sphere Laplacian: for $F$ in the finite span of the $Y_{l m}$ and each chart point with $\sin θ \neq 0$ one has $S F (θ, φ) = (A F) (θ, φ)$ , where $S = - \frac{1}{\sin θ} \partial_{θ} (\sin θ \partial_{θ}) - \frac{1}{\sin^{2} θ} \partial_{φ}^{2}$ and $A$ is the diagonal operator with $A Y_{l m} = l (l + 1) Y_{l m}$ ; in particular the chart expression of the spherical Laplacian acts on $Y_{l m}$ by multiplication by $- l (l + 1)$ (Sphere Laplacian diagonal domain, Associated Legendre function normalization, The Laplacian of a $C^{2}$ function and of a $C^{2}$ vector field).

[F5]

$C_{c}^{\infty} (R^{3}; C)$ is dense in $L^{2} (R^{3}; C)$ (Compactly supported smooth functions are dense in W^{k,p}(R^n)).

[F6]

Cauchy-Schwarz and dominated convergence (Cauchy-Schwarz inequality for $L^{2}$ , Dominated convergence).

[F7]

Chain rule and partial derivatives: the total derivative computes directional derivatives, and $C^{k}$ maps are closed under sums, products and compositions, so the formal polar-coordinate differentiation below is legitimate for $C^{2}$ functions (The chain rule for total derivatives: $D (g \circ f) (a) = D g (f (a)) \circ D f (a)$ , A total derivative computes every directional derivative, and its matrix is the Jacobian, $C^{k}$ Euclidean maps are closed under componentwise algebra and composition, $C^{k}$ Euclidean maps and diffeomorphisms, Directional derivatives and partial derivatives of a map $U \subseteq R^{m} \to R^{n}$ ).

[F8]

The Lebesgue sigma-algebra on $R^{n}$ is the completion of Borel Lebesgue measure, and any real-valued Lebesgue-measurable function agrees almost everywhere with a Borel-measurable function; for a complex-valued function apply this to its real and imaginary parts ( $L (R^{n})$ is exactly the completion of the restriction of $λ_{n}$ to the Borel sets, A function measurable for a completion is almost everywhere equal to one measurable for the original sigma-algebra).

[F9]

Under CC, the divergence theorem for a bounded piecewise $C^{1}$ domain evaluates the integral of a $C^{1}$ vector field's divergence by its boundary flux (Divergence for finite piecewise C1 presentations). The fundamental theorem of calculus integrates a differentiable real function's integrable derivative to its endpoint difference; apply it to real and imaginary components for complex functions (The second fundamental theorem: if $G$ is differentiable on $[a, b]$ with $G^{'} = f$ and $f$ is integrable, then $\int_{a}^{b} f = G (b) - G (a)$ ).

Proof

1.1F1F2F3F8givenalgebra

[Polar measurability and the angular expansion] For continuous compactly supported $ψ$ , $(r, Ω) \mapsto ψ (r Ω)$ is continuous on $(0, \infty) \times S^{2}$ , and the polar formula [F3] gives $\int_{R^{3}} ∣ ψ ∣^{2} d x = \int_{0}^{\infty} r^{2} \int_{S^{2}} ∣ ψ (r Ω) ∣^{2} d σ d r$ . For a general $L^{2}$ class, take a finite Lebesgue-measurable representative and use [F8] on its real and imaginary parts to obtain Borel representatives, setting them to zero on the Borel null sets where they are not finite; their combination is a finite Borel representative $ψ_{0}$ . If $ψ_{0}$ and ${\tilde{ψ}}_{0}$ are two such representatives, their Borel disagreement set is Lebesgue-null; applying [F3] to its indicator shows that their polar pullbacks agree for $r^{2} d r d σ$ -almost every $(r, Ω)$ , so the angular coefficient functions agree for almost every $r$ . Thus the coefficients depend only on the $L^{2}$ class. Since $(r, Ω) \mapsto ψ_{0} (r Ω)$ is Borel, completed-product Fubini [F3] gives measurable sections and Parseval [F2] gives $\int_{S^{2}} ∣ ψ_{0} (r Ω) ∣^{2} d σ = \sum_{l, m} ∣ c_{l m} (r) ∣^{2}$ for almost every $r$ . The polar formula [F3] then yields $\int_{R^{3}} ∣ ψ ∣^{2} d x = \int_{0}^{\infty} r^{2} \sum_{l, m} ∣ c_{l m} (r) ∣^{2} d r$ .

1.2F4F7algebra

[Polar form of the Laplacian] Let $f \in C^{2} (R^{3} ∖ {0})$ , put $F (r, Ω) : = f (r Ω)$ , and use the chart $Ω (θ, φ) = (\sin θ \cos φ, \sin θ \sin φ, \cos θ)$ for $0 < θ < π$ . Write $D^{2} f$ for the Hessian matrix. The chain rule gives $\partial_{r} F = \nabla f (r Ω) \cdot Ω$ and $\partial_{r}^{2} F = D^{2} f (r Ω) [Ω, Ω]$ . For $a \in {θ, φ}$ one has $\partial_{a} F = r \nabla f (r Ω) \cdot \partial_{a} Ω$ and $\partial_{a}^{2} F = r^{2} D^{2} f (r Ω) [\partial_{a} Ω, \partial_{a} Ω] + r \nabla f (r Ω) \cdot \partial_{a}^{2} Ω$ . Using $Δ_{S^{2}} = \partial_{θ}^{2} + \cot θ \partial_{θ} + \sin^{- 2} θ \partial_{φ}^{2}$ , the identities $\partial_{θ} Ω \otimes \partial_{θ} Ω + \sin^{- 2} θ \partial_{φ} Ω \otimes \partial_{φ} Ω = I - Ω \otimes Ω$ and $\partial_{θ}^{2} Ω + \cot θ \partial_{θ} Ω + \sin^{- 2} θ \partial_{φ}^{2} Ω = - 2 Ω$ give $Δ_{S^{2}} F = r^{2} (Δ f - D^{2} f [Ω, Ω]) - 2 r \partial_{r} F$ . Substitution of $\partial_{r}^{2} F = D^{2} f [Ω, Ω]$ yields $\partial_{r}^{2} F + 2 r^{- 1} \partial_{r} F + r^{- 2} Δ_{S^{2}} F = Δ f (r Ω)$ , proving (c) on this chart. Both sides are continuous on $R^{3} ∖ {0}$ , so the identity also holds on the polar axis by continuity.

2.1step 1.1F3F5F6algebra

[The isometry] By step 1.1 and Tonelli [F3] applied to the nonnegative function $r^{2} \sum_{l, m} ∣ c_{l m} (r) ∣^{2}$ one has $\int_{R^{3}} ∣ ψ ∣^{2} d x = \int_{0}^{\infty} r^{2} \sum_{l, m} ∣ c_{l m} (r) ∣^{2} d r = \sum_{l, m} \int_{0}^{\infty} ∣ u_{l m} (r) ∣^{2} d r$ for continuous compactly supported $ψ$ , and both sides are finite; the linear map $U$ is thus isometric on the dense subspace $C_{c} (R^{3})$ (density [F5] applies to $C_{c}^{\infty} \subseteq C_{c}$ ), and for $ψ \in L^{2}$ the a.e. definition of $u_{l m}$ gives the same isometry by continuity and [F6].

2.2step 1.2F4algebra

[Reduction of the free Laplacian on one block] Let $u \in C_{c}^{\infty} ((0, \infty))$ and $ψ (x) : = f (r) Y_{l m} (Ω)$ with $f (r) : = u (r) / r$ and $x = r Ω$ . The polynomial restriction formula in the statement follows from $(\sin θ)^{m} e^{i m φ} = (Ω_{1} + i Ω_{2})^{m}$ and [F4], including the cut and poles by continuity. Then $ψ \in C_{c}^{\infty} (R^{3} ∖ {0})$ , because $supp u$ is a compact subset of $(0, \infty)$ and $Y_{l m}$ is smooth on $S^{2}$ . Since $Y_{l m}$ lies in the finite span, [F4] gives $Δ_{S^{2}} ψ (r \cdot) = - l (l + 1) ψ (r \cdot)$ on the chart; by step 1.2, $- Δ ψ = - \partial_{r}^{2} ψ - 2 r^{- 1} \partial_{r} ψ - r^{- 2} Δ_{S^{2}} ψ = (- f^{''} (r) - 2 f^{'} (r) / r + l (l + 1) f (r) / r^{2}) Y_{l m}$ , and substituting $f = u / r$ gives $- f^{''} - 2 f^{'} / r + l (l + 1) f / r^{2} = (- u^{''} + l (l + 1) u / r^{2}) / r$ , which is the first identity of (d).

3.1step 2.1F2F3F8algebra

[Surjectivity and inverse] For a finite-support vector $(u_{l m}) \in K$ , extend each radial entry by zero to $R$ , apply [F8] with $n = 1$ to its real and imaginary parts, set any Borel null set of nonfinite values to zero, and restrict the resulting finite-valued Borel representatives to $(0, \infty)$ . Define $V (u_{l m}) : = \sum_{l, m} u_{l m} (r) Y_{l m} (Ω) / r$ for $r > 0$ (assign any value at the origin). This is Borel; using [F3] and [F2], $∥ V (u_{l m}) ∥_{L^{2}}^{2} = \int_{0}^{\infty} r^{2} \int_{S^{2}} ∣ \sum u_{l m} (r) Y_{l m} (Ω) / r ∣^{2} d σ d r = \sum_{l, m} \int ∣ u_{l m} ∣^{2} d r$ , the cross terms vanishing because $\int Y_{l m} \overline{Y_{l^{'} m^{'}}} d σ = δ_{l l^{'}} δ_{m m^{'}}$ by [F2]; hence $V$ is isometric on the dense finite-support subspace of $K$ and extends to a linear isometry $K \to L^{2}$ . On finite sums the definitions give $U (V (u_{l m})) = (u_{l m})$ , and on $C_{c} (R^{3})$ the angular expansion shows $V (U ψ) = ψ$ ; by density and continuity these identities extend, so $V = U^{- 1}$ and $U$ is unitary.

3.2step 2.2F9algebra

[The free form on one block] With $u, ψ$ as in step 2.2, $\int_{R^{3}} ∣ \nabla ψ ∣^{2} d x = ⟨ - Δ ψ, ψ ⟩$ since [F9] applied componentwise on a ball containing the support to the compactly supported $C^{1}$ field $\overline{ψ} \nabla ψ$ gives $\int ∣ \nabla ψ ∣^{2} = \int (- Δ ψ) \overline{ψ}$ with zero boundary flux; substituting step 2.2 and using $∥ Y_{l m} ∥_{L^{2} (S^{2})} = 1$ yields $\int_{R^{3}} ∣ \nabla ψ ∣^{2} d x = \int_{0}^{\infty} (- u^{''} (r) \overline{u (r)} + l (l + 1) ∣ u (r) ∣^{2} / r^{2}) d r$ . Integration by parts from [F9] in the compact support of $u$ gives $\int_{0}^{\infty} - u^{''} \overline{u} d r = \int_{0}^{\infty} ∣ u^{'} ∣^{2} d r$ , and the elementary identity $- 2 Re (u^{'} \overline{u}) / r + ∣ u ∣^{2} / r^{2} = - (∣ u ∣^{2} / r)^{'}$ integrates to zero over the compact support, so $\int ∣ u^{'} ∣^{2} = \int ∣ u^{'} - u / r ∣^{2}$ ; combining, the right-hand side equals $\int_{0}^{\infty} (∣ u^{'} - u / r ∣^{2} + l (l + 1) ∣ u ∣^{2} / r^{2}) d r$ , which is the second identity of (d).

4.1step 2.1step 3.1step 1.2step 2.2step 3.2∎

[Conclusion] On the dense subspace of finite sums of one-block functions the map $U$ is given by the displayed coefficient formula, so clause (a) follows from step 2.1 for all $ψ \in L^{2}$ ; clause (b) is step 3.1; clauses (c) and (d) are steps 1.2, 2.2 and 3.2.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Transported radial endpoint realization

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ \geq 0$ , let $C : = - Δ - γ / ∣ \cdot ∣$ on $D (C) = W^{2, 2} (R^{3}; C)$ be the self-adjoint operator of Coulomb infinitesimal kinetic bound (c) when $γ > 0$ , and the free Fourier multiplier when $γ = 0$ , and let $U$ be the unitary of Angular radial unitary reduction from $L^{2} (R^{3})$ onto $K = ⨁_{(l, m)} L^{2} ((0, \infty), d r)$ , with coefficients $u_{l m} = r c_{l m}$ . For $l \geq 0$ let $V_{l}$ be the closure of $C_{c}^{\infty} ((0, \infty); C)$ in $L^{2} (0, \infty; C)$ for the norm $∥ u ∥_{V_{l}}^{2} : = \int_{0}^{\infty} (∣ u^{'} (r) - \frac{u (r)}{r} ∣^{2} + \frac{l (l + 1)}{r^{2}} ∣ u (r) ∣^{2} + ∣ u (r) ∣^{2}) d r,$ realized as the space of $u \in L^{2}$ with $u^{'} - \frac{u}{r} \in L^{2}$ and $\int ∣ u ∣^{2} / r^{2} < \infty$ in the distributional sense (Densely defined, closed and closable operators, and cores, The space $L^{p} (μ)$ as the quotient by null functions), and define $a_{l} (u, v) : = \int_{0}^{\infty} [\overline{(v^{'} (r) - \frac{v (r)}{r})} (u^{'} (r) - \frac{u (r)}{r}) + \frac{l (l + 1)}{r^{2}} u (r) \overline{v (r)} - \frac{γ}{r} u (r) \overline{v (r)}] d r .$ Then:

(a) $V_{l}$ is a dense linear subspace of $L^{2} (0, \infty)$ that is a Hilbert space for $∥ \cdot ∥_{V_{l}}$ , the inclusion $V_{l} ↪ L^{2} (0, \infty)$ is continuous, and $a_{l}$ is a bounded symmetric sesquilinear form on $V_{l}$ that is closed and semibounded below;

(b) the operator $A_{l}$ defined by $D (A_{l}) : = {u \in V_{l} : \exists f \in L^{2} (0, \infty) with a_{l} (u, v) = ⟨ f, v ⟩ \forall v \in V_{l}}, A_{l} u : = f,$ is well defined, self-adjoint and bounded below; $D (A_{l})$ is dense in $V_{l}$ in the norm $∥ \cdot ∥_{V_{l}}$ and $a_{l} (u, u) = ⟨ A_{l} u, u ⟩$ for $u \in D (A_{l})$ ;

(c) for every $u \in D (A_{l})$ the distributional identity $- u^{''} + (l (l + 1) / r^{2} - γ / r) u = A_{l} u$ holds on $(0, \infty)$ ; consequently $u \in H_{l o c}^{2} (0, \infty)$ , $u$ is of class $C^{1}$ with $u^{'}$ locally absolutely continuous, and the identity holds pointwise almost everywhere on $(0, \infty)$ ; if moreover $A_{l} u = E u$ for some constant $E \in C$ (in particular if $u$ is an eigenfunction of $A_{l}$ ), then $u$ is of class $C^{2}$ on $(0, \infty)$ and the differential equation holds pointwise there;

(d) block identity. For $ψ \in L^{2} (R^{3})$ with $U ψ = (u_{l m})$ one has $ψ \in D (C)$ if and only if $u_{l m} \in D (A_{l})$ for every $(l, m)$ and $\sum_{l, m} (∥ u_{l m} ∥^{2} + ∥ A_{l} u_{l m} ∥^{2}) < \infty$ ; then $U (C ψ) = (A_{l} u_{l m})_{l, m}$ . Equivalently $U C U^{*} = ⨁_{(l, m)} A_{l}$ on $K$ ;

(e) eigenfunction transport. If $u \in D (A_{l})$ and $A_{l} u = E u$ for some $E \in R$ , then the function $ψ (x) : = u (r) Y_{l 0} (Ω) / r$ (with $x = r Ω$ ) lies in $H^{2} (R^{3}; C)$ , satisfies $C ψ = E ψ$ , and $r \mapsto u (r) / r$ is bounded on $(0, 1]$ ; moreover $ψ$ is bounded and continuous on $R^{3}$ .

No endpoint boundary condition is imposed by hand: the domain $D (A_{l})$ is the form-domain realization transported from $D (C) = H^{2} (R^{3})$ , and no physical interpretation of $A_{l}$ , $C$ , $U$ or $γ$ is asserted.

Facts & Assumptions

Given: the constants and operators $γ, C, U$ ; the form domain $V_{l}$ and form $a_{l}$ ; a parameter $μ$ chosen below; function spaces as in the statement.

[F1]

$U$ is unitary and its inverse is the finite-subset-net sum $ψ = \sum_{l, m} u_{l m} (r) Y_{l m} (Ω) / r$ ; for $ψ \in C_{c}^{\infty} (R^{3} ∖ {0})$ the coefficients are $u_{l m} = r \int ψ (r Ω) \overline{Y_{l m}} d σ$ (Angular radial unitary reduction).

[F2]

$C$ is self-adjoint on $D (C) = H^{2} (R^{3}; C)$ , $C \geq - M_{0}$ for a finite $M_{0}$ , and $⟨ C ψ, ψ ⟩ = \int_{R^{3}} ∣ \nabla ψ ∣^{2} d x - γ \int_{R^{3}} ∣ ψ ∣^{2} ∣ x ∣^{- 1} d x$ for $ψ \in D (C)$ (Coulomb infinitesimal kinetic bound, Plancherel theorem, Distributional derivatives are polynomial Fourier multipliers).

[F3]

$∣ u (r) ∣ \leq r^{1 / 2} ∥ u / r ∥_{L^{2} (r, \infty)} + r^{1 / 2} (\int_{r}^{\infty} ∣ u^{'} - \frac{u}{r} ∣^{2})^{1 / 2}$ -type endpoint bounds are not assumed; instead the following elementary inequalities are used: $2 ∣ a b ∣ \leq ε a^{2} + ε^{- 1} b^{2}$ , $∣ \int f \overline{g} ∣ \leq ∥ f ∥_{2} ∥ g ∥_{2}$ , and $C_{c}^{\infty} ((0, \infty))$ is dense in $L^{2} (0, \infty)$ (Cauchy-Schwarz inequality for $L^{2}$ , Compactly supported smooth functions are dense in W^{k,p}(R^n)).

[F4]

Orthogonal decomposition by a closed subspace is available under CC (Orthogonal decomposition by a closed subspace). Riesz representation in a Hilbert space: every bounded conjugate-linear functional on $(V_{l}, ∥ \cdot ∥_{V_{l}})$ is represented uniquely by an element of $V_{l}$ (Riesz representation for Hilbert spaces).

[F5]

Range criterion: a densely defined symmetric operator $T$ is self-adjoint if and only if it is closed and $ran (T - z) = H$ for every nonreal $z$ ; if additionally $T$ is symmetric and $ran (T + μ) = H$ for one real $μ$ and $\ker (T^{*} + μ) = {0}$ , then $T$ is self-adjoint (Range criterion for self-adjointness, Symmetric, self-adjoint and essentially self-adjoint operators).

[F7]

On $C_{c}^{\infty} (R^{3} ∖ {0})$ one has $\int ∣ \nabla ψ ∣^{2} = \int (- Δ ψ) \overline{ψ}$ by the divergence theorem applied to the compactly supported field $\overline{ψ} \nabla ψ$ , and $C_{c}^{\infty} (R^{3} ∖ {0})$ is dense in $H^{1} (R^{3}; C)$ : for $u \in H^{1}$ and the radial cutoffs $η_{ε}$ with $η_{ε} = 1$ off $B (0, \sqrt{ε})$ , $0$ on $B (0, ε)$ and $∣ \nabla η_{ε} ∣ \leq c / (r \log (1 / ε))$ , the Hardy bound of Three-dimensional Hardy bound gives $∥ u \nabla η_{ε} ∥_{2}^{2} \leq 4 c^{2} ∥ \nabla u ∥_{2}^{2} / \log (1 / ε)^{2} \to 0$ and dominated convergence gives $∥ η_{ε} u - u ∥_{H^{1}} \to 0$ (Divergence for finite piecewise C1 presentations, Vector forms: the boundary integrals of $f n$ and of $n \times F$ , Dominated convergence, Integer-order Sobolev spaces and their norms).

[F8]

For every Borel function $φ$ with $∣ φ ∣ \leq 1$ and every $ψ \in H^{2}$ , the identity $F_{2} (- Δ ψ) (ξ) = 4 π^{2} ∣ ξ ∣^{2} F_{2} ψ (ξ)$ and Cauchy-Schwarz give $∣ ψ (x) ∣ \leq ∥ (1 + ∣ ξ ∣^{2}) F_{2} ψ ∥_{2} ∥ (1 + ∣ ξ ∣^{2})^{- 1} ∥_{2}$ , so $H^{2} (R^{3}) \subseteq C_{b} (R^{3})$ with a continuous representative (Plancherel theorem, Distributional derivatives are polynomial Fourier multipliers, Cauchy-Schwarz inequality for $L^{2}$ ).

[F9]

One-dimensional Sobolev regularity: for $- \infty < a < b < \infty$ every class in $W^{1, 2} (a, b; C)$ has exactly one continuous representative that is absolutely continuous on $[a, b]$ , with $u (x) - u (y) = \int_{y}^{x} D u (t) d t$ ; applied to $u$ and then to $u^{'}$ , a class with weak derivatives $u^{'}, u^{''} \in L^{2} (a, b)$ therefore has a $C^{1}$ representative with $u^{'}$ absolutely continuous and $u^{'} (x) - u^{'} (y) = \int_{y}^{x} u^{''} (t) d t$ (One-dimensional $W^{1, p}$ functions have unique absolutely continuous representatives, Integer-order Sobolev spaces and their norms).

[F10]

Polar coordinates: $\int_{R^{3}} g d x = \int_{0}^{\infty} r^{2} \int_{S^{2}} g (r Ω) d σ (Ω) d r$ for every Borel $g \geq 0$ ; in particular for $ψ (x) = u (r) Y_{l m} (Ω) / r$ with $u \in C_{c}^{\infty} ((0, \infty))$ one has $\int_{R^{3}} ∣ ψ ∣^{2} ∣ x ∣^{- 1} d x = \int_{0}^{\infty} ∣ u (r) ∣^{2} r^{- 1} d r$ (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma, Sphere ONB).

Proof

1.1givenF3algebra

Put $D_{0} u = u^{'} - u / r$ on $C_{c}^{\infty} (0, \infty)$ . This operator is closable: if $u_{n} \to 0$ and $D_{0} u_{n} \to g$ in $L^{2}$ , then on every compact subinterval $u_{n} / r \to 0$ in $L^{2}$ and $u_{n}^{'} \to 0$ distributionally, so $g = 0$ distributionally and almost everywhere. Integration of $(∣ u ∣^{2} / r)^{'}$ on the core gives $∥ D_{0} u ∥_{2}^{2} = ∥ u^{'} ∥_{2}^{2}$ and $∥ u / r ∥_{2}^{2} = - 2 Re \int (u / r) \overline{D_{0} u}$ ; consequently $∥ u / r ∥_{2} \leq 2 ∥ D_{0} u ∥_{2}$ . Applied to differences, these identities make the defining $V_{l}$ norm equivalent to the $H^{1}$ norm for each fixed $l$ , and identify its completion in $L^{2}$ with $H_{0}^{1} (0, \infty)$ . In particular the graph completion has injective first projection, its derivative is distributional, and it is a Hilbert space densely and continuously embedded in $L^{2}$ . No fixed- $l$ equivalence constant will be used in an infinite channel sum.

1.2F1F2F7F10algebra

Let $W$ be the Hilbert direct sum of $V_{l}$ over $(l, m)$ , with squared norm $\sum ∥ u_{l m} ∥_{V_{l}}^{2}$ . On finite compact smooth radial sums, the corrected angular-radial block identity and Euclidean integration by parts give $∥ ψ ∥_{H^{1}}^{2} = \sum ∥ u_{l m} ∥_{V_{l}}^{2}$ and $q (ψ, φ) = \sum a_{l} (u_{l m}, v_{l m})$ , where $q = \int \nabla ψ \cdot \overline{\nabla φ} - γ \int ψ \overline{φ} / ∣ x ∣$ . Polar integration gives the actual $r^{- 1}$ potential weight. To establish the necessary full $H^{1}$ density, first use smooth density and the origin-removing cutoffs of [F7], followed by compact support and mollification, to reduce to $ψ \in C_{c}^{\infty} (R^{3} ∖ {0})$ .

2.1step 1.1F3F9algebra

The trace at zero of an $H^{1} (0, 1)$ function exists by $∣ u (t) - u (s) ∣ \leq ∣ t - s ∣^{1 / 2} ∥ u^{'} ∥_{2}$ , and is continuous in its $H^{1}$ norm: choose a point where $∣ u ∣ \leq ∥ u ∥_{L^{2} (0, 1)}$ and use this estimate to bound $∣ u (0) ∣$ by $∥ u ∥_{2} + ∥ u^{'} ∥_{2}$ . Thus the completion in step 1.1 has zero trace. Conversely $u, u^{'} - u / r, u / r \in L^{2}$ gives $u^{'} \in L^{2}$ and a trace; its trace must be zero since otherwise $\int ∣ u ∣^{2} / r^{2}$ diverges. Take a smooth cutoff $χ_{ε}$ vanishing below $ε$ , one beyond $2 ε$ , with $∣ χ_{ε}^{'} ∣ \leq c / ε$ ; then $∥ χ_{ε}^{'} u ∥_{2}^{2} \leq 4 c^{2} \int_{0}^{2 ε} ∣ u ∣^{2} / r^{2} \to 0$ , and all omitted function, derivative and weighted tails tend to zero. A cutoff between $R$ and $2 R$ adds a derivative bounded by $(c / R) ∥ u ∥_{L^{2} (R, 2 R)}$ . Mollify the resulting compactly supported functions on intervals away from zero; $H^{1}$ convergence and boundedness of $1 / r$ there give core convergence in $V_{l}$ . This proves exactly the stated distributional realization, including $l = 0$ , and the first part of (a).

2.2step 1.2F1F7F10algebra

For such an annular smooth $ψ$ , let $ψ_{N} = P_{N} ψ$ be the angular truncation $l \leq N$ . All its radial coefficients are smooth with the same annular support. The polar formula gives $Δ ψ_{N} = P_{N} Δ ψ$ : radial differentiation commutes with the finite projection, and angular integration by parts gives $⟨ - Δ_{S^{2}} ψ (r \cdot), Y_{l m} ⟩ = λ_{l} ⟨ ψ (r \cdot), Y_{l m} ⟩$ . This latter identity uses only the supplied finite-span eigenidentity, with a local Green argument for the other smooth function: excise caps of angular radius $δ$ , integrate the chart expression, cancel periodic azimuth boundaries, and let $δ ↓ 0$ ; smooth intrinsic gradients are bounded and cap boundaries have length $O (δ)$ , so their terms vanish. Angular completeness and polar integration give $ψ_{N} \to ψ$ and $Δ ψ_{N} \to Δ ψ$ in $L^{2}$ . For the smooth annular residual $R_{N}$ , integration by parts gives $∥ \nabla R_{N} ∥^{2} = ⟨ - Δ R_{N}, R_{N} ⟩ \leq ∥ Δ R_{N} ∥ ∥ R_{N} ∥ \to 0$ . Thus finite compact smooth radial sums are $H^{1}$ -dense with the gradient tail actually controlled.

3.1step 1.1step 2.1F3F4algebra

Write $D u = u^{'} - u / r$ and $λ_{l} = l (l + 1)$ . Hardy and Cauchy–Schwarz give $∣ \int u \overline{v} / r ∣ \leq 2 ∥ D u ∥_{2} ∥ v ∥_{2}$ ; the derivative and angular terms are bounded by their corresponding $V_{l}$ norms. Hence $a_{l}$ is bounded and symmetric. The inequality $2 γ x y \leq x^{2} / 2 + 2 γ^{2} y^{2}$ gives the uniform bound $a_{l} [u] \geq \frac{1}{2} ∥ D u ∥_{2}^{2} + λ_{l} ∥ u / r ∥_{2}^{2} - 2 γ^{2} ∥ u ∥_{2}^{2}$ . With $μ = 2 γ^{2} + 1$ , the shifted form $b_{l} = a_{l} + μ ⟨ \cdot, \cdot ⟩$ is a complete inner product with $\frac{1}{2} ∥ u ∥_{V_{l}}^{2} \leq b_{l} [u] \leq c_{γ} ∥ u ∥_{V_{l}}^{2}$ , uniformly in $l$ . Riesz representation on $(V_{l}, b_{l})$ solves $b_{l} (u, v) = ⟨ f, v ⟩$ for every $f \in L^{2}$ , so the actual domain definition gives $u \in D (A_{l})$ and $(A_{l} + μ) u = f$ . Testing $v = u$ gives uniqueness. This proves closedness and semiboundedness of the form and surjectivity of its associated shifted operator.

4.1step 3.1step 2.1F5algebra

For $u \in C_{c}^{\infty} (0, \infty)$ , integration by parts gives $a_{l} (u, v) = ⟨ L_{l} u, v ⟩$ first for core $v$ , where $L_{l} u = - u^{''} + (λ_{l} / r^{2} - γ / r) u \in L^{2}$ . Both sides are continuous in $V_{l}$ , so the identity extends to every $v \in V_{l}$ . Thus the core is contained in $D (A_{l})$ and this operator domain is $L^{2}$ -dense before any adjoint is taken. The representing $f$ is unique by $V_{l}$ density. Symmetry of the form gives symmetry of $A_{l}$ . For $y \in D (A_{l}^{*})$ , solve $(A_{l} + μ) x = (A_{l}^{*} + μ) y$ by step 3.1; then $y - x \in \ker (A_{l}^{*} + μ) = ran (A_{l} + μ)^{⊥} = 0$ . Hence $A_{l} = A_{l}^{*}$ on exactly its defined domain, and $⟨ A_{l} u, u ⟩ = a_{l} [u] \geq - 2 γ^{2} ∥ u ∥^{2}$ .

4.2step 1.1step 3.1step 1.2step 2.2F3algebra

Finite-support families with core coordinates are dense in $W$ , since each $V_{l}$ is the core completion. Step 2.2 therefore extends the finite-core isometry onto $W$ , with its $L^{2}$ extension agreeing with $U$ : $U (H^{1}) = W$ . The full polarized form identity extends by continuity. In particular the infinite Coulomb sum is uniformly bounded by $\sum ∣ \int u_{l m} \overline{v_{l m}} / r ∣ \leq 2 (\sum ∥ D u_{l m} ∥^{2})^{1 / 2} (\sum ∥ v_{l m} ∥^{2})^{1 / 2}$ ; kinetic and angular sums have their corresponding Cauchy–Schwarz bounds. This uses no equivalence constant growing with $l$ .

5.1step 3.1step 4.1F4algebra

If $w \in V_{l}$ is $b_{l}$ -orthogonal to $D (A_{l})$ , the solution $u_{f} = (A_{l} + μ)^{- 1} f$ from step 3.1 gives $0 = b_{l} (u_{f}, w) = ⟨ f, w ⟩$ for every $f \in L^{2}$ ; therefore $w = 0$ . Orthogonal decomposition in the Hilbert space $(V_{l}, b_{l})$ implies that the closure of $D (A_{l})$ is all $V_{l}$ . Norm equivalence makes this the required density in $∥ \cdot ∥_{V_{l}}$ and completes (a),(b). This proves form density directly, without a spectral-form assertion.

5.2step 4.1step 2.1F9algebra

For $u \in D (A_{l})$ , test its form identity against core $v$ . Distributional integration by parts gives $- u^{''} + (λ_{l} / r^{2} - γ / r) u = A_{l} u$ ; the cross terms in $(u^{'} - u / r) \overline{(v^{'} - v / r)}$ cancel by integrating $(u \overline{v} / r)^{'}$ . On a compact interval away from zero the potential coefficient is bounded, so $u^{''} \in L^{2}$ there and step 2.1 gives $u, u^{'} \in L^{2}$ . Thus $u \in H_{l o c}^{2}$ , and [F9] gives its $C^{1}$ representative and locally absolutely continuous derivative. If $A_{l} u = E u$ , the right side for $u^{''}$ is continuous, so integration makes $u^{'}$ of class $C^{1}$ and the equation holds everywhere with $u \in C^{2}$ . This proves (c) without regularity at the origin.

5.3step 4.1F5algebra

Define $\hat{A} = ⨁ A_{l}$ on exactly the square-summable graph domain in (d). Finite-support domain vectors are dense and coordinate symmetry gives symmetry. Self-adjointness of each $A_{l}$ gives $∥ (A_{l} \pm i)^{- 1} ∥ \leq 1$ , by $∥ (A_{l} \pm i) u ∥^{2} = ∥ A_{l} u ∥^{2} + ∥ u ∥^{2}$ . For $f \in K$ , take $u_{l m} = (A_{l} \pm i)^{- 1} f_{l m}$ ; the coordinate norm sum is bounded by $∥ f ∥^{2}$ , and $A_{l} u_{l m} = f_{l m} \mp i u_{l m}$ has square-summable norms. Thus both ranges of $\hat{A} \pm i$ are all $K$ , proving self-adjointness by the symmetric range criterion. Coordinate closedness also directly proves the graph is closed.

6.1step 4.2step 5.3F1F2F7algebra

For $ψ \in H^{2} = D (C)$ and $φ \in H^{1}$ , weak Fourier integration by parts and three-dimensional Hardy give $q (ψ, φ) = ⟨ C ψ, φ ⟩$ : prove it first for core $φ$ and extend by $H^{1}$ continuity, using $ψ / ∣ x ∣ \in L^{2}$ . By step 4.2, $u = U ψ \in W$ . Fix one coordinate $j = (l, m)$ and $v \in V_{l}$ ; then $φ = U^{- 1} (v δ_{j}) \in H^{1}$ and the full form identity gives $a_{l} (u_{j}, v) = ⟨ (U C ψ)_{j}, v ⟩$ . Hence $u_{j} \in D (A_{l})$ and $A_{l} u_{j} = (U C ψ)_{j}$ . The latter coordinates are square summable, so $C \subseteq U^{*} \hat{A} U$ . Both operators are self-adjoint; taking adjoints reverses the inclusion, giving equality. This proves the full iff graph-domain statement and action formula (d), also for $γ = 0$ using its explicit free operator.

7.1step 5.2step 6.1F1F8F9algebra∎

If $A_{l} u = E u$ , the one-block family $u δ_{(l, 0)}$ belongs to $D (\hat{A})$ ; step 6.1 gives $ψ = U^{- 1} (u δ_{(l, 0)}) \in H^{2}$ , $C ψ = E ψ$ , and $ψ = u (r) Y_{l 0} (Ω) / r$ almost everywhere. Its $H^{2}$ Fourier transform lies in $L^{1}$ by Cauchy–Schwarz with $(1 + ∣ ξ ∣^{2})^{- 1} \in L^{2}$ , so the absolutely convergent inverse defines a bounded continuous representative by dominated convergence, agreeing with the $L^{2}$ inverse by truncation. Step 5.2 and continuity of $Y_{l 0}$ make the polar expression continuous away from zero; two continuous functions agreeing almost everywhere on that open set agree everywhere, since any nonzero difference persists on a positive-measure neighborhood. Choose $Ω_{0}$ with $Y_{l 0} (Ω_{0}) \neq 0$ ; then $∣ u (r) / r ∣ \leq ∥ ψ ∥_{\infty} / ∣ Y_{l 0} (Ω_{0}) ∣$ for $0 < r \leq 1$ . This proves (e) with genuine operator-domain transport and justifies evaluation on a fixed ray. All clauses (a)–(e) follow.

Remarks

Form-domain lemma: for a self-adjoint operator $A \geq c$ with quadratic form $q_{A} [x] = \int λ d E_{x}$ , the domain $D (A)$ is dense in the form domain $Q (A) = D ((A - c I)^{1 / 2})$ in the norm $∥ x ∥_{Q}^{2} = ∥ x ∥^{2} + ∥ (A - c I)^{1 / 2} x ∥^{2}$ , and $q_{A} [x] = ⟨ A x, x ⟩$ for $x \in D (A)$ (Spectral form domain and core of a semibounded operator, Spectral theorem for unbounded self-adjoint operators (PVM form)).

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Kummer recurrence and termination

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , $l \geq 0$ , $k > 0$ , let $A_{l}$ be the radial operator of Transported radial endpoint realization, write $j : = \frac{γ}{2 k} - l - 1,$ and for $a = 2 l + 1$ and $n \in N_{0}$ let $L_{n}^{(a)} (x) : = \sum_{i = 0}^{n} (- 1)^{i} (\binom{n + a}{n - i}) \frac{x^{i}}{i!}$ be the $n$ -th Laguerre polynomial of order $a$ . Then:

(a) substitution. A $C^{2}$ function $u$ on $(0, \infty)$ satisfies $- u^{''} + (l (l + 1) / r^{2} - γ / r) u = - k^{2} u$ if and only if $v (x) : = e^{x / 2} x^{- (l + 1)} u (x / 2 k)$ is $C^{2}$ and satisfies $x v^{''} + (2 l + 2 - x) v^{'} + j v = 0$ on $(0, \infty)$ ;

(b) regular series. With $v_{0} : = 1$ and $v_{i + 1} : = \frac{(i - j) v_{i}}{(i + 1) (i + 2 l + 2)} (i \geq 0),$ the power series $v (x) : = \sum_{i \geq 0} v_{i} x^{i}$ converges for every $x \in R$ and defines a solution of $x v^{''} + (2 l + 2 - x) v^{'} + j v = 0$ on $R$ ;

(c) regular solution and general solution. The function $u_{r e g} (r) : = r^{l + 1} e^{- k r} v (2 k r)$ is a $C^{\infty}$ solution of the radial equation on $(0, \infty)$ with $u_{r e g} (r) / r^{l + 1} \to 1$ as $r ↓ 0$ ; it is, up to a nonzero scalar multiple, the only solution with $u (r) / r^{l + 1}$ bounded as $r ↓ 0$ ; every solution of the radial equation on $(0, \infty)$ is $c_{1} u_{r e g} + c_{2} u_{2}$ with $u_{2}$ defined near zero by $u_{2} (r) : = u_{r e g} (r) \int_{r_{*}}^{r} u_{r e g} (s)^{- 2} d s$ , where $0 < r_{*} < δ$ and $u_{r e g}$ has no zero in $(0, δ)$ , and extended uniquely to $(0, \infty)$ by the regular initial-value equation, and every solution that is not a multiple of $u_{r e g}$ has $∣ u (r) ∣ / r$ unbounded as $r ↓ 0$ ;

(d) termination. $v$ is a polynomial if and only if $j \in N_{0}$ ; in that case $v$ is a nonzero constant multiple of $L_{j}^{(2 l + 1)}$ , and $u_{r e g} (r) = c r^{l + 1} e^{- k r} L_{j}^{(2 l + 1)} (2 k r) (c \neq 0)$ lies in $L^{2} (0, \infty; d r)$ ;

(e) eigenfunction. If $j \in N_{0}$ then $u_{r e g} \in D (A_{l})$ and $A_{l} u_{r e g} = - k^{2} u_{r e g}$ ; equivalently, for every $m$ the transported function $ψ (r Ω) : = u_{r e g} (r) Y_{l m} (Ω) / r$ lies in $H^{2} (R^{3}) \cap D (C)$ and satisfies $C ψ = - k^{2} ψ$ .

The assertions are statements about the ordinary differential equation and the operators $A_{l}, C$ ; no physical interpretation of $γ, k$ or of any spectral level is made.

Facts & Assumptions

Given: The constants, radial operator, angular labels and choice hypotheses of the Statement. Inner products are linear in the first argument. All ordinary solutions considered below are complex-valued $C^{2}$ functions on $(0, \infty)$ .

[F1]

The endpoint supplier Transported radial endpoint realization defines $V_{l}$ as the completion of $C_{c}^{\infty} (0, \infty)$ for the norm $∥ u ∥_{2}^{2} + ∥ u^{'} - u / r ∥_{2}^{2} + l (l + 1) ∥ u / r ∥_{2}^{2}$ , and defines $a_{l} (u, w) = \int [(u^{'} - u / r) \overline{(w^{'} - w / r)} + l (l + 1) u \overline{w} / r^{2} - γ u \overline{w} / r]$ . This is a bounded form on $V_{l}$ . Its actual domain is $D (A_{l}) = {u \in V_{l} : \exists f \in L^{2}, a_{l} (u, w) = ⟨ f, w ⟩ \forall w \in V_{l}}$ , with $A_{l} u = f$ . Its exact block identity transports a single $(l, m)$ entry in $D (A_{l})$ into $D (C) = H^{2} (R^{3})$ , for every allowed $m$ . These claims currently carry both CC and AC; neither hypothesis is removed here.

[F2]

For continuous coefficients on a compact real interval, the regular IVP supplier Regular interval IVP: existence, a priori bound and entire spectral-parameter dependence gives a unique $C^{2}$ solution with each specified pair of initial data. Translating $[a, b]$ to $[0, b - a]$ applies its Statement to every $[a, b] \subset (0, \infty)$ . The supplier assumes Countable Choice.

[F3]

The derivative rules for exponentials and integer powers are The exponential function is smooth and $(\exp)^{'} = \exp$ , The complex exponential by its power series, The derivative $f^{'} (c) = \lim_{x \to c} \frac{f (x) - f (c)}{x - c}$ of $f : A \to R$ at a point $c \in A$ that is a limit point of $A$ , and differentiability on a set, Continuity of $f : A \to R$ at a point of $A$ and on $A$ : the $ε$ - $δ$ condition, its agreement with $\lim_{x \to c} f (x) = f (c)$ at a limit point, and continuity at an isolated point. The angular labels $l, m$ are those of the polar decomposition in Angular radial unitary reduction; thus $l$ is a nonnegative integer whenever $Y_{l m}$ is used.

[F4]

For integers $N \geq q \geq 0$ , $(\binom{N}{q}) = N! / [q! (N - q)!]$ by $(\binom{n}{k}) k! (n - k)! = n!$ for $k \leq n$ ; hence $(\binom{n}{k}) k! = n^{\underline{k}}$ , the quotient $n! / (k! (n - k)!)$ is a natural number, and $(\binom{n}{k}) = (\binom{n}{n - k})$ ; the binomial is the subset count of The set $[A]^{k}$ of $k$ -element subsets and the binomial coefficient $(\binom{n}{k}) : = ∣ [n]^{k} ∣$ . Factorials and finite sums use The factorial $n!$ and the falling factorial $n^{\underline{k}}$ , defined by recursion in $N$ , A finite sum in a commutative monoid indexed by an arbitrary finite set. Cauchy–Schwarz is Cauchy-Schwarz inequality for $L^{2}$ .

[F5]

For a continuous scalar integrand on a compact interval, its primitive is continuously differentiable with derivative the integrand, and integration of a continuous derivative gives its endpoint difference (Fundamental theorem of calculus for Banach-valued continuous curves, applied to the Banach space $C$ ). This supplier assumes CC. Uniform convergence permits passage through the integral because the integral of the absolute error is bounded by interval length times the supremum error.

Proof

1.1F3givenalgebra

[Substitution without dividing by a solution] Put $h (r) = r^{l + 1} e^{- k r}$ and $u = h φ$ . The multiplier $h$ is smooth and nonzero for every $r > 0$ , so $u \leftrightarrow φ$ preserves $C^{2}$ . Its derivatives satisfy $h^{'} / h = (l + 1) / r - k$ and $h^{''} / h = l (l + 1) / r^{2} - 2 k (l + 1) / r + k^{2}$ . The product rule gives the identity $- u^{''} + [l (l + 1) / r^{2} - γ / r + k^{2}] u = - h [φ^{''} + (2 (l + 1) / r - 2 k) φ^{'} + (γ - 2 k (l + 1)) φ / r]$ . Therefore the radial equation is equivalent to $r φ^{''} + (2 l + 2 - 2 k r) φ^{'} + (γ - 2 k (l + 1)) φ = 0$ . Write $φ (r) = \tilde{v} (2 k r)$ : then $φ^{'} = 2 k {\tilde{v}}^{'}$ and $φ^{''} = (2 k)^{2} {\tilde{v}}^{''}$ , and division by the nonzero scalar $2 k$ gives $x {\tilde{v}}^{''} + (2 l + 2 - x) {\tilde{v}}^{'} + j \tilde{v} = 0$ . The Statement's $v (x) = e^{x / 2} x^{- (l + 1)} u (x / (2 k))$ equals $(2 k)^{- (l + 1)} \tilde{v} (x)$ ; multiplication by this constant preserves the homogeneous equation. All transformations are reversible, including at zeros of $u$ , and establish (a).

1.2F3F4F5givenalgebra

[Convergence with every differentiated series] Let $b = 2 l + 2$ and $B = \max {1, ∣ j ∣ / b}$ . The recurrence gives $∣ v_{i + 1} ∣ \leq B ∣ v_{i} ∣ / (i + 1)$ because $i + ∣ j ∣ \leq B (i + b)$ . Induction from $v_{0} = 1$ yields $∣ v_{i} ∣ \leq B^{i} / i!$ . For any derivative order $q \geq 0$ and $R > 0$ , the $q$ -th formal derivative on $∣ x ∣ \leq R$ has absolute sum bounded by $\sum_{i \geq q} B^{i} R^{i - q} / (i - q)! = B^{q} e^{B R}$ . Every differentiated series therefore converges uniformly on that compact interval. To justify differentiation explicitly, integrate the uniformly convergent first derivative series from $0$ to $x$ ; the integral of its finite partial sums is the corresponding original partial sum minus its value at zero, and uniform convergence allows passage to the limit. The fundamental theorem for a continuous integrand proves the derivative identity. Repeating this for successive derivative series proves $v \in C^{\infty} (R)$ with all asserted derivatives.

2.1step 1.2F4givenalgebra

[Coefficient equation] By step 1.2, substitution and collection of coefficients are legitimate. The coefficient of $x^{i}$ in $x v^{''} + b v^{'} - x v^{'} + j v$ is $(i + 1) (i + b) v_{i + 1} + (j - i) v_{i}$ for each $i \geq 0$ , including $i = 0$ because the $x v^{''}$ contribution then vanishes. The recurrence makes every coefficient zero; the equation thus holds for all real $x$ . Conversely this same coefficient comparison shows that an analytic solution with constant coefficient $1$ must have exactly this recurrence. This proves (b).

3.1step 1.2step 2.1F4givenalgebra

[Exact termination and polynomial normalization] If $j \notin N_{0}$ , every factor $i - j$ is nonzero, so all $v_{i}$ are nonzero. Such a series cannot be a polynomial: its $i$ -th derivative at zero is $i! v_{i}$ by step 1.2. If $j = n \in N_{0}$ , the coefficients through $v_{n}$ are nonzero, $v_{n + 1} = 0$ , and the recurrence forces all later coefficients to vanish. For the stated Laguerre polynomial, $c_{i} = (- 1)^{i} (\binom{n + a}{n - i}) / i!$ has $c_{0} \neq 0$ and $c_{i + 1} / c_{i} = (i - n) / [(i + 1) (a + i + 1)]$ for $0 \leq i < n$ , by the factorial identities in [F4]. Since $a + i + 1 = i + 2 l + 2$ , it has precisely the same recurrence and $v = L_{n}^{(2 l + 1)} / (\binom{n + 2 l + 1}{n})$ . Its leading coefficient is nonzero. Thus the polynomial and nonzero normalization claims in (d) follow.

3.2step 1.1step 1.2step 2.1F3givenalgebra

[The regular branch and a zero-free initial interval] The function $u_{1} = u_{r e g} = r^{l + 1} e^{- k r} v (2 k r)$ is smooth on $(0, \infty)$ and solves the radial equation by steps 1.1 and 2.1. Since $v (0) = 1$ and $v$ is smooth, $u_{1} = r^{l + 1} (1 + O (r))$ as $r ↓ 0$ . It is real valued because the recurrence has real coefficients. Choose $δ > 0$ so that $u_{1} (r) / r^{l + 1} > 1 / 2$ throughout $(0, δ)$ . Hence $u_{1}$ has no zero there and has the required normalized endpoint limit.

4.1step 3.2F3F5givenconstructalgebra

[Local second solution and its exact leading asymptotic] Fix $r_{*} \in (0, δ)$ , put $I (r) = \int_{r_{*}}^{r} u_{1} (s)^{- 2} d s$ , and define $u_{2} = u_{1} I$ only on $(0, δ)$ . Here $I^{'} = u_{1}^{- 2}$ , so $u_{2}^{'} = u_{1}^{'} I + u_{1}^{- 1}$ , $u_{2}^{''} = u_{1}^{''} I$ , and $W (u_{1}, u_{2}) : = u_{1} u_{2}^{'} - u_{1}^{'} u_{2} = 1$ . Thus $u_{2}$ solves the same equation and is independent of $u_{1}$ . Set $p = 2 l + 1 > 0$ . Since $u_{1} (s)^{- 2} / s^{- p - 1} \to 1$ , comparison between $(1 - η) s^{- p - 1}$ and $(1 + η) s^{- p - 1}$ below any sufficiently small fixed endpoint shows $r^{p} \int_{r}^{r_{*}} u_{1} (s)^{- 2} d s \to 1 / p$ : the integral above that endpoint is finite and disappears after multiplication by $r^{p}$ , while $r^{p} \int_{r}^{t} s^{- p - 1} d s = (1 - (r / t)^{p}) / p$ . Letting $η ↓ 0$ gives $I (r) \sim - r^{- p} / p$ , hence $u_{2} (r) \sim - r^{- l} / (2 l + 1)$ . In particular $∣ u_{2} ∣ / r \to \infty$ and $∣ u_{2} ∣ / r^{l + 1} \to \infty$ , also when $l = 0$ .

4.2step 3.1step 3.2F1F3givenconstructalgebra

[Integrability and actual form-domain approximation] Suppose $j \in N_{0}$ . By step 3.1, $u = u_{1}$ is a polynomial times $r^{l + 1} e^{- k r}$ . Near zero $u = O (r^{l + 1})$ , $u^{'} = O (r^{l})$ , $u / r = O (r^{l})$ ; at infinity each is bounded by an exponential times a polynomial. Their squared integrals are finite: on $(0, 1)$ compare with $r^{2 l}$ , while at infinity a polynomial times $e^{- 2 k r}$ is integrable, as follows by bounding each power using the exponential series or repeated integration by parts. Thus $u, u^{'}, u / r \in L^{2}$ , and $u$ has zero trace at zero. To show membership in the actual completion $V_{l}$ , take smooth cutoffs $η_{ε}$ vanishing below $ε$ , equal to one above $2 ε$ , with $∣ η_{ε}^{'} ∣ \leq K / ε$ , and $χ_{R}$ equal to one below $R$ , vanishing above $2 R$ , with $∣ χ_{R}^{'} ∣ \leq K / R$ . The products $u_{ε, R} = η_{ε} χ_{R} u$ lie in $C_{c}^{\infty} (0, \infty)$ . The omitted $L^{2}$ , derivative and $u / r$ tails tend to zero; the added derivative at zero obeys $∥ η_{ε}^{'} u ∥_{2}^{2} \leq 4 K^{2} \int_{0}^{2 ε} ∣ u / r ∣^{2} d r \to 0$ , and the added derivative at infinity obeys $∥ χ_{R}^{'} u ∥_{2}^{2} \leq K^{2} R^{- 2} \int_{R}^{2 R} ∣ u ∣^{2} d r \to 0$ . Hence $u_{ε, R} \to u$ in $V_{l}$ , for instance along $ε = 1 / n$ , $R = n$ . This proves (d)'s $L^{2}$ claim and establishes $u \in V_{l}$ .

5.1step 4.1F2givenconstructalgebra

[Global continuation across every node] Put $q (r) = l (l + 1) / r^{2} - γ / r + k^{2}$ , continuous on $(0, \infty)$ . For $n \geq 1$ let $J_{n} = [r_{*} / (n + 1), (n + 1) r_{*}]$ . Apply [F2] on each $J_{n}$ to $u^{''} = q u$ with initial data $u_{2} (r_{*}) = 0$ , $u_{2}^{'} (r_{*}) = u_{1} (r_{*})^{- 1}$ . The unique solutions agree on nested overlaps and with step 4.1 wherever the latter is defined, so their union is a $C^{2}$ solution on $⋃_{n} J_{n} = (0, \infty)$ . No blowup at a finite positive node is possible because every such node lies inside one regular compact IVP interval. Differentiating the Wronskian gives $W^{'} = u_{1} u_{2}^{''} - u_{1}^{''} u_{2} = 0$ , and its value at $r_{*}$ is $1$ , so $W = 1$ globally. If $u_{1} (t) = 0$ , then $- u_{1}^{'} (t) u_{2} (t) = 1$ ; this also shows that every such node is simple and the continued $u_{2}$ is finite and nonzero there. The reciprocal integral of step 4.1 is never used across a node.

5.2step 3.2step 4.2F1F4givenalgebra

[Form eigenidentity and the full operator domain] For $w \in C_{c}^{\infty} (0, \infty)$ expand the kinetic integrand: $(u^{'} - u / r) \overline{(w^{'} - w / r)} = u^{'} \overline{w^{'}} - u^{'} \overline{w} / r - u \overline{w^{'}} / r + u \overline{w} / r^{2}$ . The last three terms are $- (u \overline{w} / r)^{'}$ and have zero integral because $w$ is compactly supported away from the endpoints. Therefore $a_{l} (u, w) = \int [u^{'} \overline{w^{'}} + l (l + 1) u \overline{w} / r^{2} - γ u \overline{w} / r] = \int [- u^{''} + (l (l + 1) / r^{2} - γ / r) u] \overline{w} = - k^{2} ⟨ u, w ⟩$ . Compact tests are dense in $V_{l}$ by its definition, $a_{l}$ is bounded by [F1], and $w \mapsto ⟨ u, w ⟩$ is continuous in its norm. The identity extends to every $w \in V_{l}$ . The defining condition in [F1] gives $u \in D (A_{l})$ and $A_{l} u = - k^{2} u$ . For every allowed $m$ , the family with sole nonzero $(l, m)$ entry $u$ has finite graph norm; the exact block identity gives $ψ = u Y_{l m} / r \in D (C) = H^{2} (R^{3})$ and $C ψ = - k^{2} ψ$ . Conversely that single-block domain identity gives the stated equivalence. This establishes (e).

6.1step 3.2step 4.1step 5.1F2givenalgebra

[Arbitrary solutions and endpoint uniqueness] At $r_{*}$ the initial-data matrix with columns $(u_{1}, u_{1}^{'})$ and $(u_{2}, u_{2}^{'})$ has determinant $1$ by step 5.1. For any solution $u$ , solve the invertible two-by-two linear system for $c_{1}, c_{2} \in C$ matching $(u (r_{*}), u^{'} (r_{*}))$ . The IVP uniqueness of [F2] on the nested $J_{n}$ gives $u = c_{1} u_{1} + c_{2} u_{2}$ everywhere. If $c_{2} \neq 0$ , the asymptotics in steps 3.2 and 4.1 give $u (r) \sim - c_{2} r^{- l} / (2 l + 1)$ , since $u_{1} / u_{2} \to 0$ . Hence $∣ u ∣ / r \to \infty$ and $∣ u ∣ / r^{l + 1} \to \infty$ . If $c_{2} = 0$ , the latter ratio is bounded, and every nonzero such solution is a nonzero scalar multiple of $u_{1}$ . This proves all of (c); the zero solution is included in the general linear representation.

7.1step 1.1step 1.2step 2.1step 3.1step 3.2step 4.1step 5.1step 6.1step 4.2step 5.2F1F2∎

[Conclusion and foundational use] Step 1.1 proves (a); steps 1.2 and 2.1 prove (b); steps 3.2, 4.1, 5.1 and 6.1 prove (c); steps 3.1 and 4.2 prove (d); and step 5.2 proves (e). CC is used through the regular IVP supplier in [F2]; the compact-interval extensions in step 5.1 themselves are uniquely determined, and the cutoff approximants in step 4.2 are explicit. The actual Coulomb operator premise in [F1] carries AC in its contract; that inherited assumption is retained, without relabeling it as CC-only.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Laguerre bound energy exhaustion

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , and for every $l \geq 0$ let $A_{l}$ be the radial operator of Transported radial endpoint realization; let $C = - Δ - γ / ∣ \cdot ∣$ on $D (C) = H^{2} (R^{3}; C)$ be the self-adjoint operator of Coulomb infinitesimal kinetic bound (c), with essential spectrum $[0, \infty)$ (Coulomb essential spectrum), and let $U$ be the polar unitary of Transported radial endpoint realization (d). Then:

(a) if $k > 0$ and $j : = γ / (2 k) - l - 1 \notin N_{0}$ , then the regular solution $u_{r e g}$ of Kummer recurrence and termination is not square-integrable on $(0, \infty)$ , and $- k^{2}$ is not an eigenvalue of $A_{l}$ ;

(b) the negative eigenvalues of $A_{l}$ are exactly the numbers $- \frac{γ^{2}}{4 n^{2}}, n \geq l + 1,$ each with one-dimensional eigenspace spanned by the corresponding regular solution $u_{r e g}$ ; in particular every negative eigenvalue of $A_{l}$ is simple;

(d) every eigenfunction of $C$ with negative eigenvalue lies in the closed span of the transported regular solutions: if $C ψ = E ψ$ with $E < 0$ and $U ψ = (u_{l m})$ , then $E = - γ^{2} / (4 n^{2})$ for some $n \geq 1$ , and each nonzero $u_{l m}$ is a scalar multiple of the regular solution of $A_{l}$ at that energy.

This exhausts the negative spectrum of $A_{l}$ and of $C$ by the terminating Laguerre solutions; no claim is made here about completeness of these eigenfunctions in the negative spectral subspace, about the multiplicity of the levels of $C$ , or about any physical interpretation.

Facts & Assumptions

Given: the exact hypotheses and operators of the Statement; $l, m$ are the integer angular labels of $U$ and $n$ is an integer bound-family label; $j = γ / (2 k) - l - 1$ and the regular solution normalized by $u_{r e g} (r) / r^{l + 1} \to 1$ .

[F1]

Kummer recurrence and termination supplies the convergent series $v (x) = \sum_{i \geq 0} v_{i} x^{i}$ , $v_{0} = 1$ , $v_{i + 1} = (i - j) v_{i} / ((i + 1) (i + 2 l + 2))$ , with $u_{r e g} = r^{l + 1} e^{- k r} v (2 k r)$ . It terminates precisely for $j \in N_{0}$ and is then proportional to $L_{j}^{(2 l + 1)}$ . Its second solution is constructed only on a zero-free neighborhood of zero and continued across nodes by regular initial-value uniqueness. It has $u_{2} (r) \sim - r^{- l} / (2 l + 1)$ , and every solution is $c_{1} u_{r e g} + c_{2} u_{2}$ .

[F2]

Transported radial endpoint realization gives the form $a_{l}$ on $V_{l}$ , the exact form-identity definition of $D (A_{l})$ , self-adjointness of $A_{l}$ , the block/domain identity $U C U^{*} = ⨁_{l, m} A_{l}$ , and boundedness of $u (r) / r$ near zero for every radial eigenfunction. Compactly supported smooth tests are dense in $V_{l}$ , and $a_{l}$ is bounded there.

[F3]

Coulomb infinitesimal kinetic bound gives self-adjointness of $C$ on exactly $H^{2} (R^{3})$ . Coulomb essential spectrum gives that every negative spectral point of $C$ is an isolated finite-multiplicity eigenvalue and that its negative spectral subspace is the closed span of its negative eigenspaces.

[F4]

The spectral PVM and exact-domain calculus of Spectral theorem for unbounded self-adjoint operators (PVM form), Unbounded Borel functional calculus: domains, products, spectral mapping, Projection valued measure give: a reducing block carries the restricted PVM; the projection at ${λ}$ is the projection onto the $λ$ -eigenspace; and a bounded Borel multiplier $b$ with bounded $λ b (λ)$ maps into the operator domain and realizes the indicated inverse equations. These AC-bearing suppliers are used explicitly in step 5.1; the current public AC hypothesis is retained.

Proof

1.1F1givenalgebra

[The nonterminating tail] Suppose $j \notin N_{0}$ . Every $v_{i}$ is real and nonzero. Choose $ε \in (0, 1 / 2)$ and an integer $I > \max (j, 0)$ large enough that $(i - j) / (i + 2 l + 2) \geq 1 - ε$ for all $i \geq I$ . All coefficients from $I$ onward have one sign and $∣ v_{i + 1} ∣ \geq (1 - ε) ∣ v_{i} ∣ / (i + 1)$ ; induction gives $∣ v_{i} ∣ \geq c (1 - ε)^{i} / i!$ for $i \geq I$ , where $c = ∣ v_{I} ∣ I! (1 - ε)^{- I} > 0$ . Since the series is absolutely convergent, on $x \geq 0$ the absolute value of its same-sign tail is its sum of absolute values. The reverse triangle inequality gives $∣ v (x) ∣ \geq c e^{(1 - ε) x} - P (x)$ , where one may take the nonnegative polynomial $P (x) = \sum_{i < I} ∣ v_{i} ∣ x^{i} + c \sum_{i < I} (1 - ε)^{i} x^{i} / i!$ . Thus for sufficiently large $r$ , $∣ u_{r e g} (r) ∣ \geq (c / 2) r^{l + 1} e^{k (1 - 2 ε) r}$ . Its squared integral diverges, since the exponent is positive; hence $u_{r e g} \notin L^{2} (0, \infty)$ .

1.2F1F2givenconstructalgebra

[Terminating solutions belong to the form domain] Suppose $j \in N_{0}$ and write $u = r^{l + 1} e^{- k r} v (2 k r)$ . Near zero, $u = O (r^{l + 1})$ , $u^{'} = O (r^{l})$ and $u / r = O (r^{l})$ ; at infinity each decays exponentially times a polynomial. Therefore $u, u^{'}, u / r \in L^{2}$ , and the trace at zero is zero. To verify $u \in V_{l}$ directly, choose a smooth cutoff $χ_{δ}$ that vanishes below $δ$ , equals one above $2 δ$ and has $∣ χ_{δ}^{'} ∣ \leq C / δ$ . The additional derivative term obeys $∥ χ_{δ}^{'} u ∥_{2}^{2} \leq 4 C^{2} \int_{δ}^{2 δ} ∣ u / r ∣^{2} d r \to 0$ . The other omitted terms tend to zero by integrability; cutting off the exponential tail and then smoothing on compact subsets of $(0, \infty)$ gives compact smooth approximants in the $V_{l}$ norm. This is membership in the specified endpoint realization, rather than membership inferred from $L^{2}$ alone.

2.1F1F2step 1.1algebra

[The actual origin condition excludes the second solution] If $u \in D (A_{l})$ and $A_{l} u = - k^{2} u$ , then $u$ solves the classical radial equation by [F2]. By [F1] it is $c_{1} u_{r e g} + c_{2} u_{2}$ . The regular quotient $u_{r e g} / r = O (r^{l})$ is bounded, whereas $u_{2} / r \sim - r^{- l - 1} / (2 l + 1)$ . If $c_{2} \neq 0$ , the reverse triangle inequality therefore makes $∣ u (r) ∣ / r \to \infty$ , contrary to [F2]. Consequently $u = c_{1} u_{r e g}$ , and every negative eigenspace has dimension at most one. Together with step 1.1 this proves (a).

2.2step 1.2F1F2F3algebra

[The form identity gives the operator domain and $H^{2}$ transport] The substitution of [F1] shows that $u$ solves $- u^{''} + (l (l + 1) / r^{2} - γ / r) u = - k^{2} u$ on $(0, \infty)$ . Against $w \in C_{c}^{\infty} (0, \infty)$ , integration by parts gives $a_{l} (u, w) = \int (u^{'} \overline{w^{'}} + l (l + 1) u \overline{w} / r^{2} - γ u \overline{w} / r) d r = - k^{2} ⟨ u, w ⟩$ . The equality between the kinetic forms follows by integrating $(u \overline{w} / r)^{'}$ , with no boundary term because $w$ is compactly supported away from zero. Boundedness of $a_{l}$ and density in [F2] extend the identity to every $w \in V_{l}$ . The actual definition of $D (A_{l})$ now gives $u \in D (A_{l})$ and $A_{l} u = - k^{2} u$ . The vector with this one $(l, m)$ component belongs to the direct-sum graph domain, so the exact block identity gives $ψ = u (r) Y_{l m} (Ω) / r \in D (C) = H^{2} (R^{3})$ and $C ψ = - k^{2} ψ$ for every $m$ . This argument uses no assertion that $e^{- k ∣ x ∣}$ is smooth at $x = 0$ .

3.1step 1.1step 2.1step 2.2algebra

[All radial negative eigenvalues] Termination means $j = n - l - 1 \in N_{0}$ , so $n \geq l + 1$ and $k = γ / (2 n)$ . Step 2.2 supplies a nonzero eigenvector at each $E_{n} = - γ^{2} / (4 n^{2})$ , and step 2.1 shows its eigenspace is exactly $C u_{r e g}$ . All other negative energies are excluded by steps 1.1 and 2.1. This proves (b), including every integer $n \geq l + 1$ , with no finite truncation of the bound family.

4.1F2step 2.2step 3.1algebra

[All negative eigenvalues and eigenvectors of $C$ ] By the exact block/domain identity, $C ψ = E ψ$ is equivalent to $A_{l} u_{l m} = E u_{l m}$ in every component, where $U ψ = (u_{l m})$ . If $ψ \neq 0$ , at least one component is nonzero. Step 3.1 therefore forces $E = E_{n}$ for some $n \geq 1$ , and all components with $l \geq n$ vanish; the remaining finitely many components are scalar multiples of the regular solutions at $k = γ / (2 n)$ . Conversely each $E_{n}$ is a $C$ eigenvalue by taking the $l = m = 0$ vector from step 2.2. Thus (c) and (d) hold. In fact every such eigenfunction is a finite linear combination of the transported regular solutions, which implies the asserted closed-span membership.

5.1F2F3F4step 3.1step 4.1constructalgebra

[Entire negative spectrum, including radial blocks] By [F3], every negative spectral point of $C$ is an eigenvalue, so step 4.1 proves $σ (C) \cap (- \infty, 0) = {E_{n} : n \geq 1}$ . Fix $l$ and realize $A_{l}$ as the reducing $(l, 0)$ block of $U C U^{*}$ . If $λ < 0$ is outside $σ (C)$ , the restricted inverse of $C - λ$ supplies a bounded inverse for $A_{l} - λ$ . If $λ = E_{n}$ with $n < l + 1$ , choose an open interval $J = (λ - δ, λ + δ)$ whose intersection with $σ (C)$ is just ${λ}$ , as allowed by [F3]. The restricted projection on $J$ equals the restricted projection on ${λ}$ by [F4]; this is zero, since step 3.1 excludes a radial eigenvector at this energy. The restricted PVM is consequently supported outside $J$ . Define $b (t) = (t - λ)^{- 1}$ outside $J$ and $b (t) = 0$ inside $J$ . Both $b$ and $t b$ are bounded. By the exact-domain calculus [F4], its block multiplier maps into $D (A_{l})$ and gives both $(A_{l} - λ) b (A_{l}) = I$ and $b (A_{l}) (A_{l} - λ) = I$ on the respective exact domains. Thus $λ \notin σ (A_{l})$ . The remaining negative points are eigenvalues from step 3.1, proving $σ (A_{l}) \cap (- \infty, 0) = {E_{n} : n \geq l + 1}$ . This establishes the final exhaustion sentence for both operators, including absence of additional continuous negative radial spectrum.

6.1step 2.1step 3.1step 4.1step 5.1∎

[Conclusion] Step 2.1 proves (a), step 3.1 proves (b), step 4.1 proves (c) and (d), and step 5.1 proves the stated exhaustion of the entire negative spectra.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Negative-subspace completeness and degeneracy

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ and let $C = - Δ - γ / ∣ \cdot ∣$ on $D (C) = H^{2} (R^{3}; C)$ be the self-adjoint operator of Coulomb infinitesimal kinetic bound with essential spectrum $[0, \infty)$ . For $n \geq 1$ , $0 \leq l \leq n - 1$ and $- l \leq m \leq l$ let $u_{n l}$ be the regular solution of Kummer recurrence and termination at $k = γ / (2 n)$ (the polynomial-exponential solution, unique up to a nonzero scalar), and put $ψ_{n l m} (x) : = \frac{u_{n l} (r)}{r} Y_{l m} (Ω) (x = r Ω) .$ Then:

(a) each $ψ_{n l m}$ lies in $D (C)$ and $C ψ_{n l m} = - \frac{γ^{2}}{4 n^{2}} ψ_{n l m}$ ; the vectors $ψ_{n l m}$ with distinct labels are pairwise orthogonal, and for fixed $n$ the eigenspace of $C$ at $- γ^{2} / (4 n^{2})$ is exactly their span; it has dimension $\sum_{l = 0}^{n - 1} (2 l + 1) = n^{2}$ ;

(b) after normalization, the family $(ψ_{n l m} / ∥ u_{n l} ∥_{2})_{n \geq 1, 0 \leq l \leq n - 1, - l \leq m \leq l}$ is a complete orthonormal family of the negative spectral subspace $E_{C} ((- \infty, 0)) L^{2} (R^{3})$ ;

(c) the family is not complete in $L^{2} (R^{3})$ : the nonnegative spectral subspace $E_{C} ([0, \infty)) L^{2} (R^{3})$ is nonzero (indeed infinite dimensional), so the bound states alone do not span $L^{2}$ .

All claims concern the mathematical spectral decomposition of $C$ ; no physical interpretation is asserted.

Facts & Assumptions

Given: $γ > 0$ , the operator $C$ with essential spectrum $[0, \infty)$ and spectral measure $E$ ; the labels $n, l, m$ ; the functions $ψ_{n l m}$ .

[F1]

The negative eigenvalues of $C$ are exactly $- γ^{2} / (4 n^{2})$ ( $n \geq 1$ ), and every eigenfunction of $C$ with negative eigenvalue $E = - γ^{2} / (4 n^{2})$ has, in each block $(l, m)$ with $l \leq n - 1$ , a scalar multiple of $u_{n l}$ , while every block with $l \geq n$ vanishes (because $- γ^{2} / (4 n^{2})$ is not a negative eigenvalue of $A_{l}$ when $l \geq n$ ) (Laguerre bound energy exhaustion).

[F2]

$U$ is unitary and $∥ ψ ∥_{L^{2}}^{2} = \sum_{l, m} ∥ u_{l m} ∥_{2}^{2}$ for $U ψ = (u_{l m})$ ; in particular the inverse image of a vector supported in one block $(l, m)$ with entry $u$ has norm $∥ u ∥_{2}$ (Angular radial unitary reduction (a), (b)). If $u \in D (A_{l})$ and $A_{l} u = E u$ , that one-block vector belongs to the direct-sum operator domain since $∥ u ∥_{2}^{2} + ∥ E u ∥_{2}^{2} < \infty$ , so its inverse image $ψ (x) = u (r) Y_{l m} (Ω) / r$ lies in $D (C)$ and $C ψ = E ψ$ for every allowed $m$ (Transported radial endpoint realization (d)).

[F3]

The negative spectral subspace $E_{C} ((- \infty, 0)) L^{2}$ is the closed linear span of the eigenspaces of $C$ for its negative eigenvalues (Coulomb essential spectrum (c)).

[F4]

For the self-adjoint (possibly unbounded) $C$ , if $C u = E u$ and $C v = E^{'} v$ with $u, v \in D (C)$ and distinct real $E, E^{'}$ , then $E ⟨ u, v ⟩ = ⟨ C u, v ⟩ = ⟨ u, C v ⟩ = E^{'} ⟨ u, v ⟩$ , hence $⟨ u, v ⟩ = 0$ by symmetry on its domain; $E ({λ})$ is the orthogonal projection onto $\ker (C - λ)$ ; and $λ \in σ_{e s s} (C)$ iff $rank E ((λ - ε, λ + ε)) = \infty$ for every $ε > 0$ (Symmetric, self-adjoint and essentially self-adjoint operators, Discrete and essential spectrum of a self-adjoint operator, Projection valued measure).

[F5]

For a complete orthonormal family $(e_{i})$ of a closed subspace $S$ of a Hilbert space one has $x = \sum_{i} ⟨ x, e_{i} ⟩ e_{i}$ and $∥ x ∥^{2} = \sum_{i} ∣ ⟨ x, e_{i} ⟩ ∣^{2}$ for $x \in S$ (Parseval equivalences for an orthonormal family, Orthonormal families, complete orthonormal systems and Hilbert bases, Hilbert space).

Proof

1.1F1F2F4givenalgebra

[Eigenvectors, exact norms and orthogonality] For each label put $E_{n} = - γ^{2} / (4 n^{2})$ . The Kummer termination and Laguerre exhaustion suppliers give a nonzero $u_{n l} \in D (A_{l})$ with $A_{l} u_{n l} = E_{n} u_{n l}$ ; [F2] transports the vector supported in block $(l, m)$ to $ψ_{n l m} \in D (C)$ with $C ψ_{n l m} = E_{n} ψ_{n l m}$ and $∥ ψ_{n l m} ∥_{2} = ∥ u_{n l} ∥_{2} > 0$ . For different $n$ , the numbers $E_{n}$ are distinct and self-adjointness gives orthogonality by [F4]. For equal $n$ and different $(l, m)$ , the direct-sum vectors have disjoint supports, hence their inner product is zero; unitarity transfers that orthogonality to $L^{2} (R^{3})$ . Thus $e_{n l m} : = ψ_{n l m} / ∥ u_{n l} ∥_{2}$ is an orthonormal family.

2.1step 1.1F1F2algebra

[The entire eigenspace and its dimension] Let $C ψ = E_{n} ψ$ . The block-domain identity in [F2] and exhaustion in [F1] give $U ψ = (a_{l m} u_{n l})_{0 \leq l \leq n - 1, - l \leq m \leq l}$ with all other blocks zero. There are only $\sum_{l = 0}^{n - 1} (2 l + 1)$ such blocks, so inversion gives the finite sum $ψ = \sum_{l = 0}^{n - 1} \sum_{m = - l}^{l} a_{l m} ψ_{n l m}$ ; conversely each such sum belongs to the eigenspace by step 1.1. Its squared norm is $\sum_{l = 0}^{n - 1} \sum_{m = - l}^{l} ∣ a_{l m} ∣^{2} ∥ u_{n l} ∥_{2}^{2}$ by [F2]. If the sum is zero, every coefficient vanishes because each radial norm is positive; thus these vectors form a basis. Their number is $2 \sum_{l = 0}^{n - 1} l + n = n (n - 1) + n = n^{2}$ , proving (a), including full $C$ degeneracy rather than only radial simplicity.

3.1step 1.1step 2.1F1F3F5algebra

[The bound orthonormal basis of the negative subspace] Let $S = E_{C} ((- \infty, 0)) L^{2}$ . By [F3], $S$ is the closed span of all negative eigenspaces, and by [F1] and step 2.1 these are exactly the finite-dimensional spaces spanned by $ψ_{n l m}$ for $n \geq 1$ . Each normalization scalar in step 1.1 is finite and nonzero, so the normalized and unnormalized families have exactly the same finite linear span, and therefore the same closed span $S$ . The orthonormal family $(e_{n l m})$ is consequently complete in $S$ . For $f \in S$ , [F5] gives $f = \sum_{n, l, m} ⟨ f, e_{n l m} ⟩ e_{n l m}$ in $L^{2}$ and $∥ f ∥_{2}^{2} = \sum_{n, l, m} ∣ ⟨ f, e_{n l m} ⟩ ∣^{2}$ , with the declared first-entry-linear convention. This proves (b).

4.1step 3.1F4givenalgebra

[An infinite-dimensional complementary subspace] Since $1 / 4 \in σ_{e s s} (C) = [0, \infty)$ , [F4] with radius $1 / 4$ gives infinite rank for $E_{C} ((0, 1 / 2))$ . Its range is contained in $E_{C} ([0, \infty)) L^{2}$ , because $(0, 1 / 2) \subseteq [0, \infty)$ and nested spectral projections satisfy $E_{C} ([0, \infty)) E_{C} ((0, 1 / 2)) = E_{C} ((0, 1 / 2))$ . Thus the nonnegative subspace is infinite dimensional. Spectral projections for disjoint sets satisfy $E_{C} ((- \infty, 0)) E_{C} ([0, \infty)) = 0$ , so this nonzero subspace is orthogonal to $S$ . Step 3.1 identifies the closed span of the bound states with $S$ , proving that family cannot be complete in all of $L^{2}$ and establishing (c).

5.1step 1.1step 2.1step 3.1step 4.1∎

[Conclusion] Step 1.1 proves domain membership and orthogonality; step 2.1 proves the full eigenspace and $n^{2}$ degeneracy in (a); step 3.1 proves (b); step 4.1 proves (c).

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Ground radial probability and units

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , let $C = - Δ - γ / ∣ \cdot ∣$ on $D (C) = H^{2} (R^{3}; C)$ be as above, and put $a_{0} : = 2 / γ > 0$ . Define $ψ_{100} (x) : = (\frac{γ^{3}}{8 π})^{1 / 2} e^{- r / a_{0}} (x = r Ω, r = ∣ x ∣) .$ Then:

(a) $ψ_{100} \in D (C)$ , $∥ ψ_{100} ∥_{L^{2}} = 1$ , and $C ψ_{100} = - \frac{γ^{2}}{4} ψ_{100}$ ;

(b) for every Borel set $B \subseteq R^{3}$ , writing $d x$ for Lebesgue measure, $\int_{B} ∣ ψ_{100} (x) ∣^{2} d x = \int_{0}^{\infty} \frac{γ^{3}}{8 π} r^{2} e^{- 2 r / a_{0}} (\int_{S^{2}} 1_{B} (r Ω) d σ (Ω)) d r,$ so the radial probability density $p (r) : = \frac{γ^{3}}{2} r^{2} e^{- 2 r / a_{0}} = 4 r^{2} a_{0}^{- 3} e^{- 2 r / a_{0}}$ is a probability density on $(0, \infty)$ : $p \geq 0$ and $\int_{0}^{\infty} p (r) d r = 1$ ;

(c) the mean of $p$ is $\int_{0}^{\infty} r p (r) d r = 3 a_{0} / 2$ , and the unique maximum (mode) of $p$ is at $r = a_{0}$ ;

(d) the spatial density $∣ ψ_{100} (x) ∣^{2} = \frac{γ^{3}}{8 π} e^{- 2 r / a_{0}}$ is strictly decreasing in $r$ and maximal at $x = 0$ . For every $ε > 0$ there is a remainder $R_{ε}$ such that

\int_{∣ x ∣ < ε} ∣ ψ_{100} (x) ∣^{2} d x = \frac{1}{6} (\frac{2 ε}{a_{0}})^{3} + R_{ε}, ∣ R_{ε} ∣ \leq \frac{1}{8} (\frac{2 ε}{a_{0}})^{4} .

Thus this probability is positive and tends to $0$ as $ε ↓ 0$ , and the spatial probability measure has no point atoms, including at the density maximum $x = 0$ ;

(e) parametrisation and dimensions. If $ℏ, m, κ > 0$ and $γ : = 2 m κ / ℏ^{2}$ , then $a_{0} = ℏ^{2} / (m κ)$ and the operator

C_{ℏ, m, κ} : = - \frac{ℏ^{2}}{2 m} Δ - \frac{κ}{∣ \cdot ∣}, D (C_{ℏ, m, κ}) = H^{2} (R^{3}; C),

has eigenvalues $- \frac{m κ^{2}}{2 ℏ^{2} n^{2}} = \frac{ℏ^{2}}{2 m} (- \frac{γ^{2}}{4 n^{2}})$ ( $n \geq 1$ ); under the dimensional conventions $[ℏ] = J s$ , $[m] = k g$ , $[κ] = J m$ , the constants $a_{0}^{2}$ and $m κ^{2} / (2 ℏ^{2})$ have the dimensions of $m^{2}$ and $J$ respectively.

Clauses (a)–(d) are mathematical statements about the normalised ground vector and its distribution; clause (e) is a dimensional-consistency statement about the parametrisation and is not a physical claim, measurement or identification of any quantity with nature.

Facts & Assumptions

Given: $γ > 0$ , $a_{0} = 2 / γ$ , the operator $C$ and its eigenfunctions as in the previous items of this page.

[F1]

The regular solution at $k = γ / 2$ , $l = 0$ is, up to a nonzero scalar, $u (r) = r e^{- γ r / 2} = r e^{- r / a_{0}}$ , and $A_{0} u = - (γ^{2} / 4) u$ with transported function $ψ = e^{- r / a_{0}} Y_{00}$ ; the eigenvalue site is exactly the ground level of $A_{0}$ ; moreover the negative eigenvalues of $C$ are exactly $- γ^{2} / (4 n^{2})$ , $n \geq 1$ , with $n^{2}$ -dimensional eigenspaces spanned by the transported regular solutions (Kummer recurrence and termination (d), (e), Laguerre bound energy exhaustion (b), (c), Transported radial endpoint realization (e), Negative-subspace completeness and degeneracy (a)). Nonnegative radial eigenvalues are excluded by Coulomb continuum spectral density (d); the exact block-domain identity then excludes nonnegative eigenvalues of $C$ as well.

[F2]

$Y_{00} = 1 / \sqrt{4 π}$ and $σ (S^{2}) = 4 π$ ; the polar decomposition of Lebesgue measure holds (Sphere ONB, Sphere integrals in angle coordinates, Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma).

[F3]

For $a > 0$ and $n \in {0, 1, 2, 3}$ one has $\int_{0}^{\infty} r^{n} e^{- a r} d r = n! / a^{n + 1}$ ; this follows by induction from the derivative of $r^{n} e^{- a r}$ and the fundamental theorem, the boundary terms at $0$ and $+ \infty$ vanishing (The second fundamental theorem: if $G$ is differentiable on $[a, b]$ with $G^{'} = f$ and $f$ is integrable, then $\int_{a}^{b} f = G (b) - G (a)$ , The exponential function is smooth and $(\exp)^{'} = \exp$ , Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ ).

[F4]

The $L^{2}$ norm of a Borel function is computed by the polar decomposition, and $∣ Y_{00} ∣^{2} = 1 / (4 π)$ (The space $L^{p} (μ)$ as the quotient by null functions, Complex Lp classes and Euclidean test-function conventions, Hilbert space).

Proof

1.1F1F2F3F4algebra

[The normalized ground vector] By [F1] the ground eigenfunction of $A_{0}$ is $u (r) = c r e^{- r / a_{0}}$ and the corresponding $C$ -eigenvector is $c e^{- r / a_{0}} Y_{00}$ , up to the scalar $c \neq 0$ . By [F4] and [F2], $∥ e^{- r / a_{0}} Y_{00} ∥_{L^{2}}^{2} = \frac{1}{4 π} \int_{0}^{\infty} r^{2} e^{- 2 r / a_{0}} 4 π d r = \int_{0}^{\infty} r^{2} e^{- 2 r / a_{0}} d r = 2 / (2 / a_{0})^{3} = a_{0}^{3} / 4$ by [F3]; hence the normalized ground vector is $ψ_{100} = (4 / a_{0}^{3})^{1 / 2} e^{- r / a_{0}} Y_{00} = (γ^{3} / (8 π))^{1 / 2} e^{- r / a_{0}}$ , since $Y_{00} = 1 / \sqrt{4 π}$ and $a_{0} = 2 / γ$ . By [F1] it lies in $D (C)$ and $C ψ_{100} = - (γ^{2} / 4) ψ_{100}$ . This is (a).

2.1step 1.1F2F3F4algebra

[The radial law] For a Borel set $B$ , [F2] and [F4] give $\int_{B} ∣ ψ_{100} ∣^{2} d x = \frac{γ^{3}}{8 π} \int_{B} e^{- 2 r / a_{0}} d x = \frac{γ^{3}}{8 π} \int_{0}^{\infty} r^{2} e^{- 2 r / a_{0}} \int_{S^{2}} 1_{B} (r Ω) d σ (Ω) d r$ , which is the displayed identity of (b); the factor $4 π$ enters only when the angular indicator is identically one, giving the radial marginal; the density $p (r) = \frac{γ^{3}}{2} r^{2} e^{- 2 r / a_{0}}$ is nonnegative and integrates to $1$ by the value of the integral computed in step 1.1.

3.1step 2.1F3algebra

[Mean and mode] By [F3] together with [F2], $\int_{0}^{\infty} r p (r) d r = \frac{γ^{3}}{2} \int_{0}^{\infty} r^{3} e^{- 2 r / a_{0}} d r = \frac{γ^{3}}{2} \cdot \frac{6}{(2 / a_{0})^{4}} = \frac{γ^{3}}{2} \cdot \frac{6 a_{0}^{4}}{16} = \frac{3 γ^{3} a_{0}^{4}}{16} = \frac{3 a_{0}}{2}$ because $γ a_{0} = 2$ . For the mode, differentiate $q (r) : = r^{2} e^{- 2 r / a_{0}}$ using [F3]'s ingredients: $q^{'} (r) = 2 r e^{- 2 r / a_{0}} - (2 / a_{0}) r^{2} e^{- 2 r / a_{0}} = (2 / a_{0}) r (a_{0} - r) e^{- 2 r / a_{0}}$ , which is positive for $0 < r < a_{0}$ , zero at $r = a_{0}$ and negative for $r > a_{0}$ ; hence $q$ , and with it $p$ , has its unique maximum at $r = a_{0}$ . This is (c).

3.2step 2.1F3algebra

[Density versus point probability] Set $t = 2 ε / a_{0}$ . Step 2.1 gives $P (∣ x ∣ < ε) = \frac{1}{2} \int_{0}^{t} s^{2} e^{- s} d s = 1 - e^{- t} (1 + t + t^{2} / 2)$ by differentiation and the fundamental theorem. Since $0 \leq 1 - e^{- s} = \int_{0}^{s} e^{- v} d v \leq s$ , the difference from $t^{3} / 6$ has absolute value at most $t^{4} / 8$ . Taking $R_{ε} : = P (∣ x ∣ < ε) - t^{3} / 6$ gives the exact remainder bound in (d), with the dimensionless scale $t = 2 ε / a_{0}$ ; the probability is strictly positive for every $ε > 0$ and tends to zero. The density $(π a_{0}^{3})^{- 1} e^{- 2 r / a_{0}}$ is strictly decreasing in $r$ , while every singleton has zero probability because its Lebesgue measure is zero. This proves (d), with no point atom at the density maximum.

4.1step 1.1step 3.2F1algebra

[Parametrisation and dimensions] Setting $γ = 2 m κ / ℏ^{2}$ gives $a_{0} = 2 / γ = ℏ^{2} / (m κ) > 0$ , and the operator identity $C_{ℏ, m, κ} = \frac{ℏ^{2}}{2 m} C$ on $H^{2}$ holds because $\frac{ℏ^{2}}{2 m} \cdot \frac{2 m κ}{ℏ^{2}} = κ$ ; hence the eigenvalues of $C_{ℏ, m, κ}$ are $\frac{ℏ^{2}}{2 m} \cdot (- \frac{γ^{2}}{4 n^{2}}) = - \frac{m κ^{2}}{2 ℏ^{2} n^{2}}$ by [F1], including exclusion of every nonnegative eigenvalue. For the dimensional statement, substituting the dimensional conventions and the dependent unit identity $J = k g m^{2} s^{- 2}$ , one computes $[ℏ^{2} / (m κ)] = J^{2} s^{2} k g^{- 1} J^{- 1} m^{- 1} = J s^{2} k g^{- 1} m^{- 1} = m$ , so $a_{0}^{2}$ has dimension $m^{2}$ , and $[m κ^{2} / ℏ^{2}] = k g J^{2} m^{2} J^{- 2} s^{- 2} = k g m^{2} s^{- 2} = J$ . This is (e).

5.1step 1.1step 2.1step 3.1step 3.2step 4.1∎

[Conclusion] Clauses (a), (b), (c), (d), (e) are steps 1.1, 2.1, 3.1, 3.2 and 4.1.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passOpen item page →

Positive scaling of a self-adjoint operator: the spectral projection-valued measure transforms by the preimage of the scaling map

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $H \neq {0}$ be a complex Hilbert space, let $A : D (A) \subset H \to H$ be a densely defined self-adjoint operator with spectral projection valued measure $E$ on the Borel $σ$ -algebra of $R$ (Spectral theorem for unbounded self-adjoint operators (PVM form), Projection valued measure), and let $c > 0$ . Write $c A$ for the operator with $D (c A) : = D (A)$ and $(c A) x : = c (A x)$ , and for $m_{c} : R \to R$ , $m_{c} (λ) : = c λ$ , define $F (B) : = E (m_{c}^{- 1} (B)) = E (c^{- 1} B) (B \subseteq R Borel) .$ Then:

(a) $F$ is a regular projection valued measure on the Borel sets of $R$ ;

(b) for every bounded Borel $f : R \to C$ the bounded integrals satisfy $Φ_{F} (f) = Φ_{E} (f \circ m_{c})$ , and the operator $\int_{R} λ d F (λ)$ of the spectral theorem equals $c A$ with domain $D (A)$ ;

(c) consequently $F$ is the spectral projection valued measure $E_{c A}$ of $c A$ : for every Borel $B \subseteq R$ , $E_{c A} (B) = E (c^{- 1} B);$ in particular the negative spectral subspaces agree, $E_{c A} ((- \infty, 0)) H = E ((- \infty, 0)) H .$

The assertion transfers the spectral measure along positive scaling of the operator. It asserts no spectral type of a particular operator, no multiplicity statement and no physical interpretation; $c < 0$ is not covered, and no statement is made about unbounded Borel functions of $c A$ beyond the identity functional calculus stated in (b).

Facts & Assumptions

Given: A nonzero complex Hilbert space $H$ ; a densely defined self-adjoint operator $A$ with spectral PVM $E$ ; a real $c > 0$ ; the map $m_{c} (λ) = c λ$ and $F (B) = E (c^{- 1} B)$ ; the Axiom of Countable Choice and the Axiom of Choice.

[F1]

Spectral theorem. There is a unique regular PVM $E$ on the Borel sets of $R$ with $D (A) = {x : \int_{R} λ^{2} d E_{x} (λ) < \infty}$ and $A x = \int_{R} λ d E (λ) x$ for $x \in D (A)$ ; conversely, for every regular PVM $E^{'}$ on $R$ the operator $\int λ d E^{'}$ with that domain is self-adjoint and its spectral PVM is $E^{'}$ again (Spectral theorem for unbounded self-adjoint operators (PVM form)).

[F2]

Projection valued measures. A PVM $E$ on a measurable space assigns orthogonal projections with $E (\emptyset) = 0$ , $E (X) = I$ , $E (B \cap C) = E (B) E (C)$ , and strong countable additivity: for pairwise disjoint $B_{n}$ with union $B$ one has $E (B) x = \sum_{n} E (B_{n}) x$ in norm for every $x$ ; $E$ is regular when the underlying space is locally compact Hausdorff with its Borel $σ$ -algebra and every $E_{x} (B) = ⟨ E (B) x, x ⟩$ is a regular Borel measure (Projection valued measure).

[F3]

Scalar and complex measures from a PVM. $E_{x}$ is a positive measure with $E_{x} (X) = ∥ x ∥^{2}$ , and $E_{x, y} (B) = ⟨ E (B) x, y ⟩$ is a finite complex measure, linear in $x$ and conjugate-linear in $y$ , with $∣ E_{x, y} (B) ∣ \leq ∥ x ∥ ∥ y ∥$ (Scalar and complex measures from a pvm clauses 1-3).

[F4]

Bounded functional calculus. For bounded measurable $f$ there is a unique $Φ_{E} (f) \in B (H)$ with $⟨ Φ_{E} (f) x, y ⟩ = \int f d E_{x, y}$ ; if $(s_{n})$ are simple with $∥ f - s_{n} ∥_{\infty} \to 0$ then $Φ_{E} (s_{n}) \to Φ_{E} (f)$ in operator norm; $∥ Φ_{E} (f) x ∥^{2} = \int ∣ f ∣^{2} d E_{x}$ ; and $Φ_{E}$ is linear, unital and multiplicative (Bounded borel pvm integral clauses 1, 2, Pvm integral is a star homomorphism clauses 1, 2).

[F5]

Unbounded functional calculus. For measurable $f$ the operator $f (E)$ has domain $D (f (E)) = {x : \int ∣ f ∣^{2} d E_{x} < \infty}$ and is defined as the norm limit of $Φ_{E} (f 1_{{∣ f ∣ \leq n}}) x$ ; $∥ f (E) x ∥^{2} = \int ∣ f ∣^{2} d E_{x}$ ; and if bounded measurable $g_{n}$ satisfy $∣ g_{n} ∣ \leq C ∣ f ∣$ pointwise and $g_{n} \to f$ pointwise, then $g_{n} (E) x \to f (E) x$ for every $x \in D (f (E))$ (Integral of a measurable function against a projection-valued measure, The unbounded PVM integral is densely defined, closed and normal).

[F6]

Monotone convergence. For a measure $μ$ and nonnegative measurable $h_{n}$ increasing pointwise to $h$ one has $\int h_{n} d μ \to \int h d μ$ (Monotone convergence for the integral).

[F7]

Regularity in $R$ . $R$ is locally compact Hausdorff and $σ$ -compact ( $R^{n}$ is locally compact and $σ$ -compact); a regular Borel measure is a Radon measure that is also inner regular on Borel sets (Regular Borel measure on an LCH space, Radon measure on an LCH space).

[F8]

The scaling map. The map $m_{c} (λ) = c λ$ is continuous on $R$ with continuous inverse $m_{1 / c}$ , hence a homeomorphism of $R$ onto itself, and it is Borel measurable; the image of a compact set under a continuous map is compact (Sums, scalar multiples, products, absolute values, maxima, minima and quotients with nonvanishing denominator of continuous functions are continuous, as are constants, the identity and every polynomial function clause 5, Homeomorphism, open map, closed map, embedding, and what it means for a property to be topological, Continuous functions on Euclidean spaces are Borel measurable, A continuous image of a compact space is compact; a continuous real-valued map on a nonempty compact space attains a maximum and a minimum; and a continuous bijection from a compact space to a Hausdorff space is a homeomorphism).

Proof

technique · direct; show that the transported set function $F(B)=E(c^{-1}B)$ is a regular PVM computing the operator $cA$, then identify it with the spectral measure of $cA$ by the uniqueness clause of the spectral theorem

Given: The data of the statement and [F1]-[F8].

1.1F2F8

[ $F$ is a PVM] For Borel $B$ the set $c^{- 1} B$ is Borel because $m_{c}$ is Borel measurable [F8], so $F (B) = E (c^{- 1} B)$ is an orthogonal projection; $F (\emptyset) = E (\emptyset) = 0$ and $F (R) = E (R) = I$ ; $c^{- 1} (B \cap C) = c^{- 1} B \cap c^{- 1} C$ gives $F (B \cap C) = F (B) F (C)$ ; and for pairwise disjoint $B_{n}$ with union $B$ the sets $c^{- 1} B_{n}$ are pairwise disjoint with union $c^{- 1} B$ , so strong countable additivity of $E$ [F2] gives $\sum_{n} F (B_{n}) x = \sum_{n} E (c^{- 1} B_{n}) x = E (c^{- 1} B) x = F (B) x$ for every $x \in H$ . Thus $F$ is a PVM on the Borel sets of $R$ .

2.1step 1.1F3F7F8

[ $F$ is regular] Fix $x \in H$ and write $ν : = F_{x}$ , so that $ν (B) = E_{x} (c^{- 1} B)$ . By [F3], $E_{x}$ is a finite measure with $E_{x} (R) = ∥ x ∥^{2}$ , so $ν (R) = ∥ x ∥^{2} < + \infty$ . For inner regularity let $B$ be Borel and $ε > 0$ ; by regularity of $E_{x}$ [F7] there is compact $K \subseteq c^{- 1} B$ with $E_{x} (c^{- 1} B ∖ K) < ε$ . Then $K^{'} : = m_{c} (K)$ is compact [F8], $K^{'} \subseteq B$ , and by $c^{- 1} K^{'} = K$ one has $ν (K^{'}) = E_{x} (K) > E_{x} (c^{- 1} B) - ε = ν (B) - ε$ . For outer regularity apply the same statement to the complement: choose open $V \supseteq c^{- 1} B$ with $E_{x} (V ∖ c^{- 1} B) < ε$ ; then $V^{'} : = m_{c} (V)$ is open [F8], $V^{'} \supseteq B$ , and $ν (V^{'}) - ν (B) = E_{x} (V ∖ c^{- 1} B) < ε$ . Hence every $ν = F_{x}$ is a Radon measure that is inner regular on Borel sets, i.e. a regular Borel measure [F7], and $F$ is a regular PVM. This proves (a).

2.2step 1.1F4

[Bounded functional identity] Let $f$ be bounded Borel. The functional calculi of $E$ and of $F$ exist by [F4], since $F$ is a PVM by step 1.1. For $f = 1_{B}$ and all $x, y \in H$ , $⟨ Φ_{F} (1_{B}) x, y ⟩ = \int 1_{B} d F_{x, y} = F_{x, y} (B) = ⟨ F (B) x, y ⟩ = ⟨ E (c^{- 1} B) x, y ⟩ = E_{x, y} (c^{- 1} B) = \int 1_{c^{- 1} B} d E_{x, y} = \int (1_{B} \circ m_{c}) d E_{x, y} = ⟨ Φ_{E} (1_{B} \circ m_{c}) x, y ⟩$ , so the uniqueness clause of [F4] gives $Φ_{F} (1_{B}) = Φ_{E} (1_{B} \circ m_{c})$ . Both sides are linear in the bounded measurable function [F4], so the identity holds for every simple function. For bounded Borel $f$ choose simple $s_{n}$ with $∥ f - s_{n} ∥_{\infty} \to 0$ ; then $∥ f \circ m_{c} - s_{n} \circ m_{c} ∥_{\infty} = ∥ f - s_{n} ∥_{\infty} \to 0$ , and the norm-convergence clause of [F4] applied on both sides gives $Φ_{F} (f) = \lim Φ_{F} (s_{n}) = \lim Φ_{E} (s_{n} \circ m_{c}) = Φ_{E} (f \circ m_{c})$ .

3.1step 2.2F1F4F5F6algebra

[Domain and operator identity] Put $f (λ) : = λ$ , so that $f (E) = A$ and $f (E_{F}) = \int λ d F (λ)$ by [F1] and [F5], and put $g : = f \circ m_{c} = m_{c}$ , so that $g (E) = c A$ . For $n \geq 1$ the truncations $f_{n} : = f 1_{{∣ f ∣ \leq n}}$ are bounded Borel and step 2.2 with $f_{n}$ gives $Φ_{F} (f_{n}) = Φ_{E} (f_{n} \circ m_{c}) = Φ_{E} (g_{n})$ , where $g_{n} = f_{n} \circ m_{c} = g 1_{{∣ g ∣ \leq n}}$ . For every $x \in H$ the norm identities of [F4] and [F5] give $\int ∣ f_{n} ∣^{2} d F_{x} = ∥ Φ_{F} (f_{n}) x ∥^{2} = ∥ Φ_{E} (g_{n}) x ∥^{2} = \int ∣ g_{n} ∣^{2} d E_{x}$ ; since $∣ f_{n} ∣^{2} ↑ λ^{2}$ and $∣ g_{n} ∣^{2} ↑ c^{2} λ^{2}$ pointwise, monotone convergence [F6] gives $\int λ^{2} d F_{x} = c^{2} \int λ^{2} d E_{x}$ for every $x \in H$ . Hence $D (f (E_{F})) = {x : \int λ^{2} d F_{x} < \infty} = D (A)$ by [F1] and [F5]. For $x \in D (A)$ the convergence clause of [F5] applied to $g_{n} \to g$ with $∣ g_{n} ∣ \leq ∣ g ∣$ gives $f (E_{F}) x = \lim Φ_{F} (f_{n}) x = \lim Φ_{E} (g_{n}) x = g (E) x$ ; since $g_{n} / c \to f$ pointwise with $∣ g_{n} / c ∣ \leq ∣ f ∣$ , [F5] gives $Φ_{E} (g_{n} / c) x \to f (E) x = A x$ ; linearity [F4] then makes the same limit equal to $c \lim Φ_{E} (g_{n} / c) x = c A x$ . Therefore $f (E_{F}) = c A$ with domain $D (A)$ .

4.1step 1.1step 2.1step 3.1F1∎

[Identification of the spectral measure] By steps 1.1 and 2.1, $F$ is a regular PVM on the Borel sets of $R$ , and by step 3.1 the operator $\int λ d F (λ) = f (E_{F})$ equals $c A$ with domain $D (A)$ . The converse clause of the spectral theorem [F1] applied to the regular PVM $F$ shows that $\int λ d F (λ)$ is self-adjoint with spectral PVM $F$ ; by the uniqueness clause of [F1] applied to the self-adjoint operator $c A$ , its spectral PVM is $E_{c A} = F$ , that is, $E_{c A} (B) = E (c^{- 1} B)$ for every Borel $B$ . This is (c) for general $B$ ; at $B = (- \infty, 0)$ one has $c^{- 1} (- \infty, 0) = (- \infty, 0)$ , so the negative spectral subspaces agree.

Remarks

Why the preimage direction. The candidate set function must satisfy $F (B) = E (c^{- 1} B)$ rather than $E (c B)$ : the condition $m_{c} X = m_{c}$ in the spectral theorem forces $F$ to assign to $B$ the spectral mass that $E$ assigns to the preimage, as step 3.1 verifies by computing the transported quadratic forms.
Use in this physics pair. The lemma is used only to transfer the spectral resolution of the unscaled Coulomb operator $- Δ - γ / ∣ \cdot ∣$ to the model's energy operator $H_{κ} = (ℏ^{2} / 2 m) C_{γ}$ ; no further spectral geometry is claimed.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Two-body reduced mass

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $m_{1}, m_{2} > 0$ , put $M : = m_{1} + m_{2}$ , $μ : = m_{1} m_{2} / M$ , let $κ > 0$ , $ℏ > 0$ , and define the linear coordinate map $Φ : R^{6} \to R^{6}, Φ (x_{1}, x_{2}) : = (R, r), R : = \frac{m_{1} x_{1} + m_{2} x_{2}}{M}, r : = x_{1} - x_{2},$ with inverse $x_{1} = R + \frac{m_{2}}{M} r$ , $x_{2} = R - \frac{m_{1}}{M} r$ . Write $L^{2} = L^{2} (R^{6}; C)$ and for $f \in L^{2}$ put $(U_{Φ} f) (R, r) : = f (x_{1}, x_{2})$ with $x_{i}$ expressed through $(R, r)$ . Then:

(a) $Φ$ is a linear isomorphism with $∣ \det Φ ∣ = 1$ , and $U_{Φ}$ is a unitary operator on $L^{2}$ ;

(b) for every Schwartz function $f$ , $U_{Φ} (- (ℏ^{2} / 2 m_{1}) Δ_{x_{1}} - (ℏ^{2} / 2 m_{2}) Δ_{x_{2}}) U_{Φ}^{- 1} f = (- (ℏ^{2} / 2 M) Δ_{R} - (ℏ^{2} / 2 μ) Δ_{r}) f$ ; moreover both sides define the same self-adjoint operator on the common domain $H^{2} (R^{6}; C)$ ;

(c) the operator $H_{2} : = - (ℏ^{2} / 2 m_{1}) Δ_{x_{1}} - (ℏ^{2} / 2 m_{2}) Δ_{x_{2}} - κ / ∣ x_{1} - x_{2} ∣$ with $D (H_{2}) : = H^{2} (R^{6}; C)$ is self-adjoint, and $U_{Φ} H_{2} U_{Φ}^{- 1} = T_{R} \otimes I + I \otimes C_{μ}$ under the identification $L^{2} (R^{6}) ≅ L^{2} (R_{R}^{3}) \hat{\otimes} L^{2} (R_{r}^{3})$ of Euclidean L² product unitary: L²(R^a)⊗̂L²(R^b) ≅ L²(R^{a+b}), where $T_{R} : = - (ℏ^{2} / 2 M) Δ_{R}$ and $C_{μ} : = - (ℏ^{2} / 2 μ) Δ_{r} - κ / ∣ \cdot ∣$ ;

(d) $σ (H_{2}) = [- \frac{μ κ^{2}}{2 ℏ^{2}}, \infty)$ , and $H_{2}$ has no eigenvalues at all: the negative part of its spectrum is the interval $[- \frac{μ κ^{2}}{2 ℏ^{2}}, 0)$ , consisting entirely of continuous spectrum, and each relative bound-state energy $- \frac{μ κ^{2}}{2 ℏ^{2} n^{2}}$ of $C_{μ}$ is a threshold of the continuous spectrum of $H_{2}$ rather than an eigenvalue; in particular a relative bound state tensored with any centre-of-mass state is not an eigenvector of $H_{2}$ ;

(e) On $L^{2} (R_{R}^{3}; C)$ define the centre-of-mass momentum component

P_{R, j}^{c m} : = \frac{ℏ}{i} \partial_{R_{j}}, D (P_{R, j}^{c m}) : = {g \in L^{2} (R_{R}^{3}; C) : \partial_{R_{j}} g \in L^{2} as a weak derivative} .

On the full two-body space define its lift

{\tilde{P}}_{R, j} : = U_{Φ}^{- 1} (P_{R, j}^{c m} \otimes I) U_{Φ}, D ({\tilde{P}}_{R, j}) : = U_{Φ}^{- 1} {F \in L^{2} (R_{R}^{3} \times R_{r}^{3}; C) : \partial_{R_{j}} F \in L^{2} as a weak derivative} .

Each ${\tilde{P}}_{R, j}$ is self-adjoint and has no nonzero eigenvectors in $L^{2} (R^{6}; C)$ ; hence there is no normalisable state of definite total momentum (a simultaneous eigenvector of the three lifted components).

All statements are mathematical statements about the operators on $L^{2} (R^{6})$ and the coordinate map; no physical interpretation, no identification of $m_{1}, m_{2}, κ, ℏ$ with measured constants, and no claim about nature is made.

Facts & Assumptions

Given: $m_{1}, m_{2}, κ, ℏ > 0$ and the derived constants $M, μ$ ; the map $Φ$ and pullback $U_{Φ}$ ; Schwartz functions.

[F1]

Linear change of variables: for an invertible linear map $Φ$ with $∣ \det Φ ∣ = 1$ the pullback $U_{Φ} f : = f \circ Φ^{- 1}$ is unitary on $L^{2} (R^{6}; C)$ (The space $L^{p} (μ)$ as the quotient by null functions, Schwartz convolution and product laws, A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm).

[F2]

For $c > 0$ , $b \in R$ the operator $H_{0} = F_{2}^{*} M_{c ∣ ξ ∣^{2} + b} F_{2}$ with domain $W^{2, 2} (R^{d}; C)$ is self-adjoint and equals $- \frac{c}{4 π^{2}} Δ + b$ on Schwartz functions; in particular free kinetic operators with positive coefficients are self-adjoint on $W^{2, 2}$ (Real multiplication operators and free plus bounded-potential Schrödinger operators, Integer-order W^{k,2} and H^k agree with equivalent norms, Integer-order Sobolev spaces and their norms).

[F3]

Kato-Rellich: if $A$ is self-adjoint and $B$ is symmetric with $D (A) \subseteq D (B)$ and $A$ -bound less than one, then $A + B$ on $D (A)$ is self-adjoint (Kato-Rellich theorem, Relative boundedness with respect to an operator).

[F4]

Hardy bound and its infinitesimal form: for $g \in W^{1, 2} (R^{3}; C)$ , $∥ g / ∣ \cdot ∣ ∥_{2} \leq 2 ∥ \nabla g ∥_{2}$ , and for $g \in W^{2, 2} (R^{3})$ and every $η > 0$ , $∥ g / ∣ \cdot ∣ ∥_{2} \leq η ∥ Δ g ∥_{2} + η^{- 1} ∥ g ∥_{2}$ (Coulomb infinitesimal kinetic bound, Three-dimensional Hardy bound).

[F5]

The negative eigenvalues of $C_{μ}$ are $- \frac{μ κ^{2}}{2 ℏ^{2} n^{2}}$ ( $n \geq 1$ ), with eigenspaces spanned by the corresponding regular solutions, and each nonzero radial component of every negative eigenfunction is a multiple of its corresponding regular solution (Laguerre bound energy exhaustion, Negative-subspace completeness and degeneracy).

[F6]

Euclidean tensor identification: $L^{2} (R^{3}) \hat{\otimes} L^{2} (R^{3}) ≅ L^{2} (R^{6})$ isometrically, with elementary tensors mapping to products (Euclidean L² product unitary: L²(R^a)⊗̂L²(R^b) ≅ L²(R^{a+b}), Hilbert tensor construction: inner product, completion and product bases); the chain rule computes the second derivatives of a pullback along an invertible linear map (The chain rule for total derivatives: $D (g \circ f) (a) = D g (f (a)) \circ D f (a)$ , Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , Directional derivatives and partial derivatives of a map $U \subseteq R^{m} \to R^{n}$ ).

[F7]

$C_{μ}$ has no eigenvalue in $[0, \infty)$ : with $γ : = 2 μ κ / ℏ^{2}$ one has $C_{μ} = \frac{ℏ^{2}}{2 μ} C_{γ}$ for $C_{γ} : = - Δ - γ / ∣ \cdot ∣$ , whose block identity makes the eigenvalues of $C_{γ}$ the union of those of the radial operators $A_{l}$ , and no $A_{l}$ has eigenvalue $0$ or eigenvalue $k^{2} > 0$ (Coulomb continuum spectral density (d), Coulomb Jost Volterra solution (d), Transported radial endpoint realization (d)).

[F8]

Tonelli applies to the nonnegative squared norms of sections on the sigma-finite product $R_{R}^{3} \times R_{r}^{3}$ ; smooth compactly supported functions are dense in $W^{2, 2} (R^{6})$ (Tonelli's theorem for nonnegative measurable functions on a sigma-finite product, Compactly supported smooth functions are dense in W^{k,p}(R^n)). Real measurable multiplication is closed on its maximal $L^{2}$ domain (Real multiplication operators and free plus bounded-potential Schrödinger operators).

[F9]

For $γ > 0$ , $C_{γ} : = - Δ - γ / ∣ \cdot ∣$ has an onto unitary spectral representation $V_{γ}$ as the countable sum over spherical channels $(l, m)$ of bound copies with energies $- γ^{2} / (4 n^{2})$ , $n \geq l + 1$ , and continuum copies $L^{2} ((0, \infty), d λ)$ with maximal multiplier $M_{λ}$ , with no zero eigenvector (Onto radial continuum transform (d), Negative-subspace completeness and degeneracy). For a self-adjoint $A$ and $c > 0$ , its scaled PVM satisfies $E_{c A} (B) = E_{A} (c^{- 1} B)$ for every Borel $B$ (Positive scaling of a self-adjoint operator: the spectral projection-valued measure transforms by the preimage of the scaling map (b), (c)). The change-of-variables formula applies to the positive linear dilation of the continuum variable (A C^1 diffeomorphism satisfies the change-of-variables formula for nonnegative Lebesgue measurable functions).

[F10]

The partial Plancherel transform $F_{R} \otimes I$ is unitary on $L^{2} (R_{R}^{3} \times R_{r}^{3}; C)$ ; it maps the maximal weak-derivative domain ${F : \partial_{R_{j}} F \in L^{2}}$ onto ${\hat{F} : ξ_{j} \hat{F} \in L^{2}}$ and sends $\frac{ℏ}{i} \partial_{R_{j}}$ to multiplication by $2 π ℏ ξ_{j}$ . This follows from Plancherel theorem, Distributional derivatives are polynomial Fourier multipliers, and Euclidean L² product unitary: L²(R^a)⊗̂L²(R^b) ≅ L²(R^{a+b}). A real finite-valued multiplier on its maximal $L^{2}$ domain is self-adjoint (Real multiplication operators and free plus bounded-potential Schrödinger operators (a)); Tonelli applies to the product variables (Tonelli's theorem for nonnegative measurable functions on a sigma-finite product).

Proof

1.1F1F6givenalgebra

[The coordinate map] In the three coordinates the matrix of $Φ$ relative to the splitting $(x_{1}, x_{2}) \mapsto (R, r)$ is block diagonal with the $2 \times 2$ block $(\begin{array}{cc} m_{1} / M & m_{2} / M \\ 1 & - 1 \end{array})$ repeated three times; its determinant is $- m_{2} / M - m_{1} / M = - 1$ , so $∣ \det Φ ∣ = 1$ and $Φ$ is invertible with the displayed inverse. By the linear change-of-variables theorem the pullback $U_{Φ} f = f \circ Φ^{- 1}$ is unitary on $L^{2} (R^{6}; C)$ , since it preserves the integral of $∣ f ∣^{2}$ ; this is (a).

1.2F6F8F10algebra

[Lifted total momentum and no eigenvector] By [F10], the partial Plancherel transform conjugates each lifted operator ${\tilde{P}}_{R, j}$ to multiplication by the real function $(ξ, r) \mapsto 2 π ℏ ξ_{j}$ on its maximal domain ${G : ξ_{j} G \in L^{2} (R^{6})}$ . The real-multiplier theorem therefore makes ${\tilde{P}}_{R, j}$ self-adjoint on exactly the weak-derivative domain in (e). If ${\tilde{P}}_{R, j} Ψ = p Ψ$ , its partial Fourier transform satisfies $(2 π ℏ ξ_{j} - p) \hat{F} (ξ, r) = 0$ for a.e. $(ξ, r)$ , so it is supported on the hyperplane $ξ_{j} = p / (2 π ℏ)$ . This hyperplane has Lebesgue measure zero in $R_{ξ}^{3}$ ; Tonelli over the remaining momentum coordinates and $r$ shows $\hat{F} = 0$ and hence $Ψ = 0$ . Thus no lifted component, and consequently no triple of lifted components, has a nonzero eigenvector.

2.1step 1.1F6givenalgebra

[Kinetic separation on Schwartz functions] Let $f$ be Schwartz. The chain rule gives $\partial_{x_{1}, j} = \frac{m_{1}}{M} \partial_{R_{j}} + \partial_{r_{j}}$ and $\partial_{x_{2}, j} = \frac{m_{2}}{M} \partial_{R_{j}} - \partial_{r_{j}}$ , hence $Δ_{x_{1}} = \frac{m_{1}^{2}}{M^{2}} Δ_{R} + 2 \frac{m_{1}}{M} \partial_{R} \cdot \partial_{r} + Δ_{r}$ and $Δ_{x_{2}} = \frac{m_{2}^{2}}{M^{2}} Δ_{R} - 2 \frac{m_{2}}{M} \partial_{R} \cdot \partial_{r} + Δ_{r}$ . Multiplying by $\frac{ℏ^{2}}{2 m_{i}}$ and summing, the mixed terms cancel because $\frac{m_{1}}{M} \cdot \frac{1}{m_{1}} - \frac{m_{2}}{M} \cdot \frac{1}{m_{2}} = 0$ , while the $Δ_{R}$ coefficient is $- \frac{ℏ^{2}}{2 M}$ and the $Δ_{r}$ coefficient is $- \frac{ℏ^{2}}{2} (\frac{1}{m_{1}} + \frac{1}{m_{2}}) = - \frac{ℏ^{2}}{2 μ}$ ; this is the displayed identity of (b) at the level of Schwartz functions.

3.1step 2.1F2givenalgebra

[The kinetic operators agree as self-adjoint operators] Let $K$ denote the self-adjoint operator on $W^{2, 2} (R^{6})$ given by $- \frac{ℏ^{2}}{2 m_{1}} Δ_{x_{1}} - \frac{ℏ^{2}}{2 m_{2}} Δ_{x_{2}}$ (existence by [F2], applied to the Fourier multiplier $4 π^{2} (\frac{ℏ^{2}}{2 m_{1}} ∣ ξ_{1} ∣^{2} + \frac{ℏ^{2}}{2 m_{2}} ∣ ξ_{2} ∣^{2})$ , which is comparable to $∣ ξ ∣^{2}$ ), and let $K^{'}$ be the self-adjoint operator on $W^{2, 2} (R^{6})$ given by $- \frac{ℏ^{2}}{2 M} Δ_{R} - \frac{ℏ^{2}}{2 μ} Δ_{r}$ . Both act on Schwartz functions, a common core of $W^{2, 2}$ -type Fourier-multiplier operators, and step 2.1 shows $U_{Φ} K U_{Φ}^{- 1} f = K^{'} f$ for Schwartz $f$ ; since $U_{Φ} K U_{Φ}^{- 1}$ and $K^{'}$ are self-adjoint operators agreeing on a common core, they are equal, and $U_{Φ} (D (K)) = D (K^{'}) = H^{2} (R^{6}; C)$ . This completes (b).

4.1step 3.1F2F3F4F6F8algebra

[Norm lift and two-body self-adjointness] Under $U_{Φ}$ , the original multiplier $- κ / ∣ x_{1} - x_{2} ∣$ becomes $- κ / ∣ r ∣$ . First let $w \in C_{c}^{\infty} (R_{R}^{3} \times R_{r}^{3})$ . For each $R$ , [F4] gives $∥ w (R, \cdot) / ∣ r ∣ ∥_{L_{r}^{2}} \leq η ∥ Δ_{r} w (R, \cdot) ∥_{L_{r}^{2}} + η^{- 1} ∥ w (R, \cdot) ∥_{L_{r}^{2}}$ . Taking $L_{R}^{2}$ norms and the triangle inequality, with [F8], yields $∥ w / ∣ r ∣ ∥_{2} \leq η ∥ Δ_{r} w ∥_{2} + η^{- 1} ∥ w ∥_{2}$ . Approximate arbitrary $w \in H^{2} (R^{6})$ in its Sobolev norm by [F8]. Applying this estimate to differences makes the quotients Cauchy in $L^{2}$ ; closedness of multiplication identifies their limit with $w / ∣ r ∣$ , so the estimate holds on all of $H^{2}$ . The Fourier symbols of $- Δ_{r}$ and $K^{'} = - (ℏ^{2} / (2 M)) Δ_{R} - (ℏ^{2} / (2 μ)) Δ_{r}$ are simultaneous nonnegative multipliers $4 π^{2} ∣ ξ_{r} ∣^{2}$ and $4 π^{2} (ℏ^{2} ∣ ξ_{R} ∣^{2} / (2 M) + ℏ^{2} ∣ ξ_{r} ∣^{2} / (2 μ))$ ; their pointwise domination gives $∥ Δ_{r} w ∥_{2} \leq (2 μ / ℏ^{2}) ∥ K^{'} w ∥_{2}$ by Plancherel. Therefore $∥ κ w / ∣ r ∣ ∥_{2} \leq (2 κ μ η / ℏ^{2}) ∥ K^{'} w ∥_{2} + (κ / η) ∥ w ∥_{2}$ , an arbitrarily small operator relative bound. Kato–Rellich [F3] makes $K^{'} - κ / ∣ r ∣$ self-adjoint on $H^{2}$ . Conjugating back gives exactly $H_{2} = U_{Φ}^{- 1} (K^{'} - κ / ∣ r ∣) U_{Φ}$ , self-adjoint on $H^{2}$ ; the separated differential expression, under [F6], is $T_{R} \otimes I + I \otimes C_{μ}$ . This denotes the self-adjoint commuting spectral sum: choose $c$ with $C_{μ} + c \geq 0$ . In its joint product representation the nonnegative sum has domain equal to the intersection of the two factor domains, because each nonnegative summand is bounded pointwise by their sum. The relative norm bound just proved makes the $C_{μ}$ graph norm equivalent to the relative Laplacian graph norm; Fourier comparison of the two nonnegative kinetic weights identifies that intersection with $H^{2} (R^{6})$ . Thus its graph closure is the same Kato–Rellich operator. This proves (c) with an actual fiber-norm lift and no form-to-norm inference.

5.1step 4.1F2F5F6F7F8F9algebra

[Full spectrum and no eigenvalue by the product representation] By step 4.1 the transformed operator is $T_{R} \otimes I + I \otimes C_{μ}$ . Put $γ : = 2 μ κ / ℏ^{2} > 0$ and $c : = ℏ^{2} / (2 μ) > 0$ , so $C_{μ} = c C_{γ}$ on their common domain $H^{2} (R^{3})$ . Under $V_{γ}$ of [F9], this is the diagonal operator with scaled bound values $c (- γ^{2} / (4 n^{2})) = - μ κ^{2} / (2 ℏ^{2} n^{2})$ and continuum multiplier $c λ$ . The PVM transfer in [F9] gives, on a continuum copy, $E_{C_{μ}} (B) = M_{1_{c^{- 1} B} (λ)}$ in this representation. Define $(D_{c} h) (E) : = c^{- 1 / 2} h (E / c)$ on each continuum copy, with inverse $(D_{c}^{- 1} g) (λ) = c^{1 / 2} g (c λ)$ . Changing variables $E = c λ$ gives $\int_{0}^{\infty} ∣ D_{c} h (E) ∣^{2} d E = \int_{0}^{\infty} ∣ h (λ) ∣^{2} d λ$ , so $D_{c}$ is onto unitary. The same substitution gives $\int_{0}^{\infty} E^{2} ∣ D_{c} h (E) ∣^{2} d E = c^{2} \int_{0}^{\infty} λ^{2} ∣ h (λ) ∣^{2} d λ$ ; hence it maps the exact maximal domain of $c M_{λ}$ onto that of $M_{E}$ and satisfies $D_{c} (c M_{λ}) D_{c}^{- 1} = M_{E}$ . On indicators it gives $D_{c} M_{1_{c^{- 1} B} (λ)} D_{c}^{- 1} = M_{1_{B} (E)}$ , explicitly transporting the continuum PVM. The direct sum of the identity on every bound copy and $D_{c}$ on every continuum copy is onto unitary: its norm identity and inverse hold in each coordinate and therefore in the countable square sum. Composing it with $V_{γ}$ gives the full relative spectral resolution of $C_{μ}$ , including its diagonal domain and no zero eigenchannel. Apply this unitary and the $2 π$ -normalized center Fourier transform to the tensor decomposition, using [F6] and Plancherel. The absence of nonnegative relative eigenvalues in [F7] agrees with that resolution. The operator becomes the countable sum of real multiplication operators: on each relative bound copy its multiplier is $c_{R} (ξ) + E_{n}$ , where $c_{R} (ξ) = 4 π^{2} ℏ^{2} ∣ ξ ∣^{2} / (2 M)$ and $E_{n} = - μ κ^{2} / (2 ℏ^{2} n^{2})$ ; on each continuum copy it is $c_{R} (ξ) + E^{'}$ for $E^{'} > 0$ . These multipliers are all at least $E_{1}$ , so the operator is bounded below by $E_{1}$ . For every $E \geq E_{1}$ choose $ξ_{0}$ with $c_{R} (ξ_{0}) = E - E_{1}$ and normalized smooth Fourier profiles supported in shrinking balls around $ξ_{0}$ . Their inverse transforms $φ_{δ}$ are Schwartz, and $∥ (T_{R} - (E - E_{1})) φ_{δ} ∥ \to 0$ by continuity of $c_{R}$ . The correctly ordered vectors $φ_{δ} \otimes u_{1}$ have norm one and belong to the operator domain, with residual $(T_{R} - (E - E_{1})) φ_{δ} \otimes u_{1}$ tending to zero; a bounded resolvent at $E$ would contradict this, so $E \in σ (H_{2})$ . Hence $σ (H_{2}) = [E_{1}, \infty)$ . For an eigenvalue equation at any real $E$ , each bound-component Fourier class must be supported on ${ξ : c_{R} (ξ) = E - E_{n}}$ , a sphere, a point or the empty set, all Lebesgue-null; each continuum component must be supported on ${(ξ, E^{'}) : E^{'} = E - c_{R} (ξ)}$ , which has zero product measure by Tonelli since every $E^{'}$ section is a singleton. The countable spectral sum therefore has no nonzero eigenvector. All spectral points, including the negative interval, are continuous spectrum because the operator is self-adjoint. Every $E_{n}$ is the lower edge of its relative bound-channel band $[E_{n}, \infty)$ , a threshold without a total eigenvector. This proves (d) with all masses and couplings retained.

6.1step 1.1step 1.2step 2.1step 3.1step 4.1step 5.1∎

[Conclusion] Steps 1.1, 2.1, 3.1, 4.1 and 5.1 establish (a)–(d), and step 1.2 establishes (e).

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Atomic dipole matrix-element domain

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ > 0$ , let $C = - Δ - γ / ∣ \cdot ∣$ on $H^{2} (R^{3}; C)$ be as above, let $(ψ_{n l m})$ be the bound eigenfunctions of Negative-subspace completeness and degeneracy, and for $i = 1, 2, 3$ define the coordinate multiplication operator $Q_{i} ψ : = (ξ \mapsto ξ_{i} ψ (ξ)), D (Q_{i}) : = {ψ \in L^{2} (R^{3}; C) : ξ_{i} ψ \in L^{2}} .$ Write also $Q_{0} : = x_{3}$ and $Q_{\pm} : = (x_{1} \pm i x_{2}) / \sqrt{2}$ . Then:

(a) each $Q_{i}$ is self-adjoint on $D (Q_{i})$ , the domain is a proper dense subspace of $L^{2}$ , and every bound state satisfies $ψ_{n l m} \in D (Q_{i})$ , with $∥ Q_{i} ψ_{n l m} ∥_{2}^{2} = \int_{0}^{\infty} r^{2} ∣ u_{n l} (r) ∣^{2} d r \cdot \int_{S^{2}} ∣ Ω_{i} Y_{l m} (Ω) ∣^{2} d σ (Ω) < \infty$ ;

(b) for all labels and each $q \in {- 1, 0, 1}$ the matrix element $⟨ ψ_{n^{'} l^{'} m^{'}}, Q_{q} ψ_{n l m} ⟩$ is absolutely convergent and factors as $⟨ ψ_{n^{'} l^{'} m^{'}}, Q_{q} ψ_{n l m} ⟩ = (\int_{0}^{\infty} r u_{n^{'} l^{'}} (r) \overline{u_{n l} (r)} d r) (\int_{S^{2}} Y_{l^{'} m^{'}} (Ω) \overline{Q_{q} (Ω) Y_{l m} (Ω)} d σ (Ω)),$ where $Q_{q} (Ω)$ denotes the restriction of the polynomial $Q_{q}$ to $S^{2}$ ;

(c) the angular factor vanishes unless $l^{'} = l + 1, m^{'} = m + q$ or $l \geq 1, l^{'} = l - 1, m^{'} = m + q$ ; these are necessary conditions only: no assertion is made that an allowed pair has nonzero radial integral, and no transition probability is asserted or computed;

(d) for the oscillator: fix real parameters $m > 0$ , $ω > 0$ , $ℏ > 0$ and put $ℓ : = \sqrt{ℏ / (m ω)} > 0$ . For the complete orthonormal family $(ψ_{n}^{o s c})$ of Oscillator scaling and exact realization and the position operator $X f (x) = x f (x)$ on $D (X) = {f \in L^{2} (R, d x) : x f \in L^{2}}$ (multiplication by $ℓ y$ after $x = ℓ y$ ), one has $X ψ_{n}^{o s c} = \frac{ℓ}{\sqrt{2}} (\sqrt{n} ψ_{n - 1}^{o s c} + \sqrt{n + 1} ψ_{n + 1}^{o s c}),$ so $⟨ ψ_{n^{'}}^{o s c}, X ψ_{n}^{o s c} ⟩$ vanishes unless $n^{'} = n \pm 1$ , with the displayed coefficients; in particular the oscillator position matrix elements are supported on a single step $∣ Δ n ∣ = 1$ , while the Coulomb selection rules constrain only $Δ l$ and $Δ m$ and leave the radial integral free.

This item concerns matrix elements of multiplication operators on $L^{2}$ and their finiteness and selection rules; it asserts no absorption, emission or transition rate and no physical interpretation.

Facts & Assumptions

Given: $γ > 0$ ; the integer bound/angular labels of the stated family; maximal domains $D (Q_{q}) = {f : Q_{q} f \in L^{2}}$ , so $D (Q_{0}) = D (Q_{3})$ and $D (Q_{\pm}) = D (Q_{1}) \cap D (Q_{2})$ because $∣ Q_{\pm} ∣^{2} = (x_{1}^{2} + x_{2}^{2}) / 2$ ; the bound states $ψ_{n l m} = u_{n l} (r) Y_{l m} (Ω) / r$ of the previous items; the coordinate multipliers $Q_{i}, Q_{q}$ ; the oscillator parameters $m, ω, ℏ > 0$ and $ℓ = \sqrt{ℏ / (m ω)} > 0$ .

[F1]

For a real finite-valued measurable $a$ the multiplier $M_{a} f = a f$ with $D (M_{a}) = {f : a f \in L^{2}}$ is self-adjoint (Real multiplication operators and free plus bounded-potential Schrödinger operators (a), Symmetric, self-adjoint and essentially self-adjoint operators); the coordinate functions $ξ_{i}$ are real finite-valued.

[F2]

Every bound state has $u_{n l} \in L^{2} (0, \infty)$ , $u_{n l} (r) = c r^{l + 1} e^{- k r} L_{n - l - 1}^{(2 l + 1)} (2 k r)$ up to a nonzero scalar with $k = γ / (2 n)$ , and $∥ ψ_{n l m} ∥_{2} = ∥ u_{n l} ∥_{2}$ (Negative-subspace completeness and degeneracy, Kummer recurrence and termination, Angular radial unitary reduction).

[F3]

Dipole angular selection rules: for the spherical components $Q_{0} = x_{3}$ , $Q_{\pm} = (x_{1} \pm i x_{2}) / \sqrt{2}$ and the family $(Y_{l m})$ , the angular matrix element $⟨ Y_{l^{'} m^{'}}, Q_{q} ∣_{S^{2}} Y_{l m} ⟩$ can be nonzero only if $l^{'} = l + 1, m^{'} = m + q$ or $l \geq 1, l^{'} = l - 1, m^{'} = m + q$ ; the rule is a necessary condition on the angular integrals only (degree and weight clauses (b), (c) of Dipole angular selection rules, Associated Legendre function normalization).

[F4]

Oscillator: for real $m, ω, ℏ > 0$ and $ℓ = \sqrt{ℏ / (m ω)} > 0$ , the scaled Hermite functions $ψ_{n}^{o s c} = ℓ^{- 1 / 2} g_{n} (x / ℓ)$ form a complete orthonormal family of $L^{2} (R, d x)$ with $x = ℓ y$ , and the Hermite recurrence gives $\sqrt{2} y g_{n} = \sqrt{n} g_{n - 1} + \sqrt{n + 1} g_{n + 1}$ with the lowering term omitted at $n = 0$ (equivalently define $g_{- 1} = 0$ and $ψ_{- 1}^{o s c} = 0$ ) (Oscillator scaling and exact realization, Hermite recurrence (d), (g), Orthonormal families, complete orthonormal systems and Hilbert bases, Parseval equivalences for an orthonormal family).

[F5]

Polar volume integration is $d x = r^{2} d r d σ$ (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma); nonnegative product integrals use Tonelli's theorem for nonnegative measurable functions on a sigma-finite product, and absolutely convergent product integrals use Tonelli and Fubini for the completed product, with only almost-everywhere section measurability. For $a > 0$ and integer $p \geq 0$ , integration by parts starting with $p = 0$ gives $\int_{0}^{\infty} r^{p} e^{- a r} d r = p! / a^{p + 1}$ ; the boundary terms vanish since a sufficiently high exponential-series term bounds every fixed polynomial power (The second fundamental theorem: if $G$ is differentiable on $[a, b]$ with $G^{'} = f$ and $f$ is integrable, then $\int_{a}^{b} f = G (b) - G (a)$ , The exponential function is smooth and $(\exp)^{'} = \exp$ , Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ ). Cauchy-Schwarz for the $L^{2}$ and $L^{2} (S^{2}, σ)$ pairings (Cauchy-Schwarz inequality for $L^{2}$ , Real and complex inner product spaces, with the inner product linear in the first argument, The space $L^{p} (μ)$ as the quotient by null functions, Complex Lp classes and Euclidean test-function conventions).

Proof

1.1F1F5givenalgebra

[Coordinate domains] Each real $Q_{i}$ is self-adjoint by [F1]. For $f \in L^{2}$ , $f_{N} = 1_{∣ x_{i} ∣ \leq N} f$ lies in $D (Q_{i})$ and converges in $L^{2}$ , proving density. Put $φ (t) = (1 + ∣ t ∣)^{- 1}$ ; it belongs to $L^{2} (R)$ whereas $t^{2} ∣ φ (t) ∣^{2} \geq 1 / 4$ for $∣ t ∣ \geq 1$ . For each $i$ , $f_{i} (x) = φ (x_{i}) \prod_{j \neq i} e^{- ∣ x_{j} ∣}$ belongs to $L^{2}$ but $x_{i} f_{i} \notin L^{2}$ by nonnegative product integration. Thus every domain is proper, including $i = 2, 3$ , not just the first-coordinate example.

2.1step 1.1F2F5algebra

[Bound-state domain membership] By [F2], $u_{n l}$ is polynomial-exponential with positive decay rate and is $O (r^{l + 1})$ at zero. Hence $\int_{0}^{\infty} r^{2} ∣ u_{n l} ∣^{2} d r < \infty$ : near zero the integrand is $O (r^{2 l + 4})$ , and all polynomial-exponential moments at infinity are finite by [F5]. Polar coordinates give exactly $∥ Q_{i} ψ_{n l m} ∥^{2} = \int r^{2} ∣ u_{n l} ∣^{2} d r \int ∣ Ω_{i} Y_{l m} ∣^{2} d σ$ , whose angular factor is at most one. Each bound state therefore lies in every $D (Q_{i})$ and in the stated complex component domains. This proves (a).

3.1step 2.1F2F5algebra

[Absolute convergence and factorization] Put $v = u_{n^{'} l^{'}}$ , $u = u_{n l}$ , $Y^{'} = Y_{l^{'} m^{'}}$ and $Y = Y_{l m}$ . Cauchy–Schwarz gives $\int_{0}^{\infty} r ∣ v \overline{u} ∣ d r \leq ∥ v ∥_{2} ∥ r u ∥_{2} < \infty$ and $\int_{S^{2}} ∣ Y^{'} \overline{Q_{q} Y} ∣ d σ \leq ∥ Y^{'} ∥_{2} ∥ Q_{q} Y ∥_{2} \leq 1$ . Tonelli and Fubini in [F5] justify absolute integration of the polar product and its factorization into $(\int_{0}^{\infty} r v \overline{u} d r) (\int_{S^{2}} Y^{'} \overline{Q_{q} Y} d σ)$ , exactly (b) for the linear-first pairing. Coordinate multiplication generally produces two angular blocks; no false single-block claim is used.

4.1step 3.1F3algebra

[All necessary angular branches] By [F3], $Q_{q} h_{l m} = R_{q} + r^{2} G_{q}$ has harmonic degrees $l + 1, l - 1$ , with $G_{q} = (\sum c_{q, i} \partial_{i} h_{l m}) / (2 l + 1)$ . At $l = 1$ its derivative term can be a nonzero constant in $V_{0}$ , so the downward branch $l^{'} = 0$ is retained. At $l = 0$ the derivatives vanish and only the upward branch remains. Degree orthogonality and the $L_{z}$ eigenvalue calculation give exactly $l^{'} = l + 1, m^{'} = m + q$ , or $l \geq 1, l^{'} = l - 1, m^{'} = m + q$ . These are necessary only; the radial factor is $\int_{0}^{\infty} r u_{n^{'} l^{'}} \overline{u_{n l}} d r$ and may vanish for an allowed angular pair. No transition probability is defined. This proves (c).

5.1step 4.1F4F5algebra

[The oscillator comparison] In the scaled variable $x = ℓ y$ multiplication by $x$ is $ℓ$ times multiplication by $y$ ; by [F4] the Hermite recurrence gives, on the complete orthonormal family $(ψ_{n}^{o s c})$ , $X ψ_{n}^{o s c} = ℓ y ψ_{n}^{o s c} = \frac{ℓ}{\sqrt{2}} (\sqrt{n} ψ_{n - 1}^{o s c} + \sqrt{n + 1} ψ_{n + 1}^{o s c})$ ; the lowering term being absent at $n = 0$ . The recurrence puts $X ψ_{n}^{o s c}$ in $L^{2}$ , proving membership in the maximal position domain; taking pairings with $ψ_{n^{'}}^{o s c}$ and using orthonormality [F4] and Cauchy-Schwarz [F5] gives $⟨ ψ_{n^{'}}^{o s c}, X ψ_{n}^{o s c} ⟩ = \frac{ℓ}{\sqrt{2}} (\sqrt{n} δ_{n^{'}, n - 1} + \sqrt{n + 1} δ_{n^{'}, n + 1})$ , so exactly the transitions $∣ Δ n ∣ = 1$ occur, with the displayed coefficients. This is (d).

6.1step 1.1step 2.1step 3.1step 4.1step 5.1∎

[Conclusion] Clauses (a), (b), (c), (d) are steps 1.1 and 2.1, 3.1, 4.1, and 5.1.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Coulomb Jost Volterra solution

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ \geq 0$ , $l \geq 0$ , and let $A_{l}$ be the radial operator of Transported radial endpoint realization with form domain $V_{l}$ . Let $D : = {k \in C : Re k > 0, Im k \geq 0}$ be the parameter domain for the Jost solution. For each $k_{\min} > 0$ also write $K : = {k \in D : Re k \geq k_{\min}}$ for its restriction. Then:

(a) regular solution. With $c_{0} : = 1$ , $c_{- 1} : = 0$ and $c_{n} : = - \frac{γ c_{n - 1} + k^{2} c_{n - 2}}{n (n + 2 l + 1)} (n \geq 1),$ the series $φ_{l} (k, r) : = r^{l + 1} \sum_{n \geq 0} c_{n} r^{n}$ converges for every $r > 0$ and every $k \in C$ , the function $(k, r) \mapsto φ_{l} (k, r)$ is holomorphic in $k$ and $C^{\infty}$ in $r > 0$ , and $φ_{l}$ solves $- u^{''} + (l (l + 1) / r^{2} - γ / r) u = k^{2} u$ with $φ_{l} (k, r) / r^{l + 1} \to 1$ as $r ↓ 0$ ;

(b) Jost-type solution. Put $a : = γ / (2 k)$ , $θ (r) : = k r + a \log r$ , $w (r) : = e^{i θ (r)}$ , $b : = a^{2} + l (l + 1) + i a$ and consider the integral equation $v (r) = 1 + \int_{r}^{\infty} \frac{b}{s^{2}} w (s)^{2} (\int_{r}^{s} w (y)^{- 2} d y) v (s) d s .$ For every $k \in D$ the equation has exactly one bounded solution $v (\cdot; k)$ on $[r_{0}, \infty)$ for some $r_{0} > 0$ depending on $k$ ; extend $f_{l}$ uniquely by its regular initial data to every $r > 0$ and put $v = f_{l} / w$ there. Constructions with different admissible tails agree by bounded-solution and initial-value uniqueness, so these functions are independent of the tail and of $k_{\min}$ . One has $v (r) = 1 + O (1 / r)$ , $v^{'} (r) = O (1 / r)$ ; the functions $f_{l} (k, r) : = w (r) v (r)$ and $v$ are jointly continuous on $D \times (0, \infty)$ , holomorphic in $k$ on ${Re k > 0, Im k > 0}$ , satisfy the radial equation $- \partial_{r}^{2} f_{l} + (l (l + 1) / r^{2} - γ / r) f_{l} = k^{2} f_{l}$ exactly, and obey the asymptotics $f_{l} = e^{i (k r + a \log r)} (1 + O (1 / r)), f_{l}^{'} = e^{i (k r + a \log r)} [i k + O (1 / r)] (r \to \infty);$

(c) Wronskians. For real $k > 0$ one has $W (\overline{f_{l}}, f_{l}) = 2 i k$ and $W_{l} (k) : = W (φ_{l}, f_{l}) \neq 0$ ; consequently every solution of the radial equation on $(0, \infty)$ is $A f_{l} + B \overline{f_{l}}$ with $A, B \in C$ , and $φ_{l}$ is a nonzero complex multiple combination of $f_{l}, \overline{f_{l}}$ ;

(d) no positive eigenvalue. For real $k > 0$ and $A, B \in C$ , if $g = A f_{l} + B \overline{f_{l}}$ then $\int_{R}^{L} ∣ g (r) ∣^{2} d r = (∣ A ∣^{2} + ∣ B ∣^{2}) [L - R - γ \log (L / R) / (2 k^{2})] + O (1)$ as $L \to \infty$ with $R > r_{0}$ fixed; hence $g \in L^{2} (r_{0}, \infty)$ if and only if $A = B = 0$ , and $k^{2} > 0$ is not an eigenvalue of $A_{l}$ .

All assertions concern the ordinary differential equation and the operator $A_{l}$ ; no scattering amplitude, cross-section or physical interpretation is asserted or computed.

Facts & Assumptions

Given: $γ \geq 0$ , $l \geq 0$ , $k_{\min} > 0$ , $K$ as above, and for $k \in D$ the quantities $a, θ, w, b$ ; the radial equation $L_{l} u = k^{2} u$ with $L_{l} u : = - u^{''} + (l (l + 1) / r^{2} - γ / r) u$ .

[F1]

On any compact interval $[a, b] \subset (0, \infty)$ the radial equation has a two-dimensional space of $C^{2}$ solutions, and for real $k$ complex conjugation preserves the solution space because both the potential and $k^{2}$ are real (Regular interval IVP: existence, a priori bound and entire spectral-parameter dependence, Continuity of $f : A \to R$ at a point of $A$ and on $A$ : the $ε$ - $δ$ condition, its agreement with $\lim_{x \to c} f (x) = f (c)$ at a limit point, and continuity at an isolated point).

[F2]

The Wronskian $W (u, v) = u v^{'} - u^{'} v$ of two solutions of $- (u^{''}) + q u = 0$ with $q$ continuous is constant (Regular interval IVP: existence, a priori bound and entire spectral-parameter dependence).

[F3]

Derivatives of exponential and logarithmic-type phases: $w^{'} = i (k + a / r) w$ , $w^{''} = - [(k + a / r)^{2} + i a / r^{2}] w$ (The complex exponential by its power series, The exponential function is smooth and $(\exp)^{'} = \exp$ , Complex sine, cosine, hyperbolic sine, and hyperbolic cosine are entire with their standard derivatives, Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ ).

[F4]

Banach fixed point and Gronwall: the standard existence-uniqueness and a priori bound arguments for Volterra integral equations; integrals of continuous functions against continuous kernels may be interchanged (A contraction of a nonempty complete metric space into itself has exactly one fixed point, the limit of the iterates from any starting point, Gronwall's integral inequality with variable and constant coefficients, Tonelli and Fubini for the completed product, with only almost-everywhere section measurability).

[F6]

Every eigenfunction $u \in D (A_{l})$ of $A_{l}$ with $A_{l} u = E u$ , $E$ constant, is a $C^{2}$ solution of the radial equation on $(0, \infty)$ (Transported radial endpoint realization (c)).

[F7]

A continuous function with zero contour integral on every triangle is holomorphic, and a locally uniformly convergent series of holomorphic functions is holomorphic with termwise derivatives (Morera's theorem: vanishing triangle integrals characterize holomorphy among continuous functions, A locally uniformly convergent series of holomorphic functions may be differentiated term by term).

Proof

1.1F7givenalgebra

[Regular solution] On a compact $k$ -set choose $C \geq \max (1, γ, \sup ∣ k ∣)$ . Induction gives $∣ c_{n} ∣ \leq C^{n} / n!$ , since the recurrence majorant divided by $C^{n} / n!$ is $[γ / C + ∣ k ∣^{2} (n - 1) / C^{2}] / (n + 2 l + 1) \leq 1$ ; at $n = 1$ the second term is absent. Each $c_{n}$ is a polynomial in $k^{2}$ . The exponential majorant also bounds every fixed $r$ -derivative on compact subintervals of $(0, \infty)$ , so the series is entire in $k$ , smooth in $r > 0$ , and termwise substitution gives the stated radial equation. Its leading coefficient is one, and the remaining series tends to zero as $r ↓ 0$ , uniformly on compact $k$ -sets. For real $k$ all coefficients and $φ_{l}$ are real.

1.2F3givenalgebra

[A common tail and the correct complex-power modulus] Perform the construction on $D : = {Re k > 0, Im k \geq 0}$ , then restrict to $K$ . Fix a compact $Q \subset D$ , and put $m : = \min_{Q} ∣ k ∣ > 0$ , $A : = \max_{Q} ∣ a ∣$ , $B : = \max_{Q} ∣ b ∣$ , $H : = 2 / m + 2 A / m^{2}$ , and $D : = B H$ . Choose $r_{0} \geq \max (1, γ / (2 m^{2}), 2 A / m, 2 D)$ . For $s \geq y \geq r_{0}$ , the exact modulus is $∣ w (s)^{2} w (y)^{- 2} ∣ = \exp [- 2 Im k (s - y)] (s / y)^{- 2 Im a}$ . Here $- 2 Im a = γ Im k / ∣ k ∣^{2}$ , and $\log (s / y) \leq (s - y) / y$ makes this modulus at most one. Write $p (y) = k + a / y$ and $d (y) = 1 / (- 2 i p (y))$ . Then $∣ p ∣ \geq m / 2$ , $∣ d ∣ \leq 1 / m$ and $∣ d^{'} ∣ \leq 2 A / (m^{2} y^{2})$ . Integration by parts on the finite interval gives $w (s)^{2} \int_{r}^{s} w (y)^{- 2} d y = d (s) - d (r) w (s)^{2} w (r)^{- 2} - \int_{r}^{s} d^{'} (y) w (s)^{2} w (y)^{- 2} d y$ , whose modulus is at most $H$ . Thus the actual kernel $R (k; r, s) : = b s^{- 2} w (s)^{2} \int_{r}^{s} w (y)^{- 2} d y$ satisfies $∣ R ∣ \leq D / s^{2}$ , uniformly for $k \in Q$ and $s \geq r \geq r_{0}$ .

2.1step 1.2F4algebra

[Contraction and bounded-solution uniqueness] On the complete space $C_{b} ([r_{0}, \infty); C)$ define $(V_{k} h) (r) = \int_{r}^{\infty} R (k; r, s) h (s) d s$ . The kernel estimate gives $∥ V_{k} ∥ \leq D / r_{0} \leq 1 / 2$ ; continuity follows by splitting the integral at a finite upper limit and using the uniformly small $s^{- 2}$ tail. Therefore $v = \sum_{n \geq 0} V_{k}^{n} 1$ is the unique fixed point of $h \mapsto 1 + V_{k} h$ , and $∥ v ∥_{\infty} \leq 2$ , $∣ v (r) - 1 ∣ \leq 2 D / r$ . Any bounded function for which the integral equation holds is continuous by the same kernel estimate and hence equals this fixed point. Constructions on different common tails agree on their overlap by uniqueness. This proves bounded-solution uniqueness on the chosen tail and the uniform $O (r^{- 1})$ estimate.

3.1step 1.2step 2.1F3algebra

[Differentiate the genuine double integral] In the integral equation the lower-limit term vanishes because $R (k; r, r) = 0$ , whereas $\partial_{r} R = - b s^{- 2} w (s)^{2} w (r)^{- 2}$ . The ratio bound in step 1.2 dominates this derivative times $v (s)$ by $2 B / s^{2}$ on each compact $r$ -interval in the tail. Truncated-integral differentiation followed by uniform convergence therefore gives $v^{'} (r) = - b w (r)^{- 2} I (r)$ , where $I (r) : = \int_{r}^{\infty} w (s)^{2} v (s) s^{- 2} d s$ ; this is a derivative identity, not a replacement of the original fixed-point equation by $v = 1 + b I$ . The integral for $I$ is absolutely convergent: for real $k$ , $∣ w ∣ = 1$ , and for $Im k > 0$ , $∣ w (s) ∣ = e^{- Im k s} s^{- Im a}$ decays exponentially times a power. Hence $I^{'} = - w (r)^{2} v (r) / r^{2}$ and $v^{''} = - 2 i (k + a / r) v^{'} + b v / r^{2}$ . The ratio estimate gives $∣ v^{'} ∣ \leq 2 B / r$ , uniformly on $Q$ . Using $w^{''} = - [(k + a / r)^{2} + i a / r^{2}] w$ and $2 k a = γ$ , the product rule for $f = w v$ gives $f^{''} = [- k^{2} - γ / r + l (l + 1) / r^{2}] f$ . Consequently $f = w [1 + O (r^{- 1})]$ and $f^{'} = w [i k + O (r^{- 1})]$ , with the factor $w$ retained and all errors uniform on $Q$ .

4.1step 1.2step 2.1step 3.1F3F4F7algebra

[Continuity and holomorphy on the common tail] Each Picard iterate is jointly continuous in $(k, r)$ on $Q \times [r_{0}, \infty)$ : on compact $r$ -intervals the finite-integral kernel is continuous, and the $D / s^{2}$ majorant makes the improper tail uniformly small. Induction on iterates followed by the uniform geometric bound $∥ V_{k}^{n} 1 ∥_{\infty} \leq 2^{- n}$ proves joint continuity of $v$ ; the derivative formula, dominated by $2 B / s^{2}$ , also proves joint continuity of $v^{'}$ . On each triangle in ${Re k > 0, Im k > 0}$ with a compact neighborhood, the finite-integral kernel is holomorphic because $1 / k$ and the complex exponential are holomorphic. If an iterate is holomorphic, interchange its $s$ -integral with the triangle contour integral using the uniform $s^{- 2}$ bound; the contour integral vanishes, so Morera proves that the next iterate is holomorphic. Uniform convergence and [F7] prove holomorphy of $v$ . Apply the same contour argument to $v^{'} = - \int_{r}^{\infty} b s^{- 2} w (s)^{2} w (r)^{- 2} v (s) d s$ to prove holomorphy of its initial data. Thus $f, f^{'}$ are continuous to real $k$ and holomorphic in the open first quadrant. The construction on $D$ also supplies an open complex neighborhood for points with $Re k = k_{\min}$ and $Im k > 0$ , which makes the asserted holomorphy at that boundary precise.

4.2step 2.1step 3.1F3algebra

[First integration by parts improves the derivative] Fix real $k > 0$ and write $Q (r) = 2 i θ^{'} (r) = 2 i (k + a / r)$ and $h (r) = v (r) / r^{2}$ . Since $θ^{'} \geq k$ , $Q^{- 1}$ is bounded and its first two derivatives are $O (r^{- 2})$ and $O (r^{- 3})$ . Initially $v = O (1)$ and $v^{'} = O (r^{- 1})$ , so $h = O (r^{- 2})$ , $h^{'} = O (r^{- 3})$ and $(h / Q)^{'} = O (r^{- 3})$ . Integration by parts, whose boundary term vanishes at infinity, gives $I (r) = - e^{2 i θ (r)} h (r) / Q (r) - \int_{r}^{\infty} e^{2 i θ (s)} (h / Q)^{'} (s) d s = O (r^{- 2})$ . The exact derivative identity then gives $v^{'} = O (r^{- 2})$ ; the ODE from step 3.1 gives $v^{''} = O (r^{- 2})$ .

5.1step 3.1step 4.1F1F3F7algebra

[Extend to every positive radius] Fix $ε > 0$ smaller than the common $r_{0}$ . The regular-IVP result gives a unique extension of $f$ from $(f (r_{0}), f^{'} (r_{0}))$ across $[ε, r_{0}]$ . To establish complex parameter dependence rather than invoking that supplier's real-parameter differentiability clause, write $Y = (f, f^{'})^{T}$ and $Y^{'} = M (k, r) Y$ , with $M = (\begin{matrix} 0 & 1 \\ l (l + 1) / r^{2} - γ / r - k^{2} & 0 \end{matrix})$ . Its initial-value integral series on $[ε, r_{0}]$ has $n$ th term bounded by $C_{0} [C_{M} (r_{0} - ε)]^{n} / n!$ , uniformly on compact $k$ -sets, since the $n$ fold ordered integration simplex has volume at most $(r_{0} - ε)^{n} / n!$ . Each term is continuous to real $k$ and holomorphic in the open first quadrant by finite contour integration. Thus the sum has both properties by [F7], agrees with the unique IVP solution, and supplies the asserted parameter properties for every $r > 0$ . Set $v = f / w$ on the extension; $w$ never vanishes, so the same properties hold for $v$ . Patching overlapping tails and intervals is unambiguous by IVP uniqueness. For fixed nonreal $k \in D$ , $f, f^{'}$ have the stated exponential-times-power decay.

5.2step 4.2step 2.1step 3.1F3algebra

[Second integration gives the amplitude coefficient] Now $h^{''} = v^{''} / r^{2} - 4 v^{'} / r^{3} + 6 v / r^{4} = O (r^{- 4})$ . Put $j = (h / Q)^{'}$ . Then $j = O (r^{- 3})$ , $j^{'} = O (r^{- 4})$ and $(j / Q)^{'} = O (r^{- 4})$ . A second integration by parts gives $\int_{r}^{\infty} e^{2 i θ} j = - e^{2 i θ (r)} j (r) / Q (r) - \int_{r}^{\infty} e^{2 i θ} (j / Q)^{'} = O (r^{- 3})$ . Therefore $I (r) = - e^{2 i θ (r)} v (r) / (2 i θ^{'} (r) r^{2}) + O (r^{- 3})$ , $v^{'} = b v / (2 i θ^{'} r^{2}) + O (r^{- 3}) = b / (2 i k r^{2}) + O (r^{- 3})$ , and integration to the known limit $v (\infty) = 1$ gives $v = 1 - b / (2 i k r) + O (r^{- 2})$ . Since $- b / (2 i k) = - a / (2 k) + i [a^{2} + l (l + 1)] / (2 k)$ and $∣ w ∣ = 1$ , the coefficient is $∣ f ∣^{2} = ∣ v ∣^{2} = 1 - a / (k r) + O (r^{- 2})$ . These estimates are uniform when real $k$ is restricted to a compact subset of $(0, \infty)$ , because all preceding constants and the lower bound for $θ^{'}$ can be chosen uniformly.

6.1step 1.1step 3.1step 5.1F1F2algebra

[Real-energy Wronskians] Any real $k > 0$ belongs to the public domain $D$ of the single patched family $f_{l}$ . By the asymptotic formulas, $W (\overline{f}, f) = 2 i (k + a / r) ∣ v ∣^{2} + 2 i Im (\overline{v} v^{'}) = 2 i k + O (r^{- 1})$ . It is constant by [F2], hence equals $2 i k$ . The pair $f, \overline{f}$ is therefore a basis of the two-dimensional solution space. If $W (φ_{l}, f) = 0$ , their initial data are proportional at any positive radius, hence IVP uniqueness implies $f = c φ_{l}$ ; neither solution is identically zero, by their normalizations. Since $φ_{l}$ is real, proportionality forces $W (\overline{f}, f) = 0$ , a contradiction. Thus $W_{l} (k) \neq 0$ , every solution is $A f + B \overline{f}$ , and the nonzero $φ_{l}$ has such a representation with coefficients not both zero.

7.1step 6.1step 5.2F3F6algebra

[The mixed term has a bounded integral] For fixed $R > r_{0}$ , integration by parts gives $\int_{R}^{L} f^{2} = [e^{2 i θ} v^{2} / Q]_{R}^{L} - \int_{R}^{L} e^{2 i θ} (v^{2} / Q)^{'}$ . The endpoint term is uniformly bounded and $(v^{2} / Q)^{'} = 2 v v^{'} / Q + v^{2} (Q^{- 1})^{'} = O (r^{- 2})$ is integrable, hence the whole integral is bounded independently of $L$ . Expanding $∣ A f + B \overline{f} ∣^{2} = (∣ A ∣^{2} + ∣ B ∣^{2}) ∣ f ∣^{2} + 2 Re (A \overline{B} f^{2})$ and integrating step 5.2 yields $\int_{R}^{L} ∣ A f + B \overline{f} ∣^{2} = (∣ A ∣^{2} + ∣ B ∣^{2}) [L - R - (a / k) \log (L / R)] + O (1)$ , with $a / k = γ / (2 k^{2})$ . If at least one coefficient is nonzero the integral diverges, since its linear term dominates the logarithm; if both vanish it is zero. Every positive-energy eigenfunction in $D (A_{l})$ is an $L^{2}$ classical solution by [F6], hence step 6.1 and this divergence exclude such an eigenfunction.

8.1step 1.1step 1.2step 2.1step 3.1step 4.1step 5.1step 6.1step 4.2step 5.2step 7.1given∎

[Conclusion and choice accounting] Steps 1.1; 1.2, 2.1, 3.1, 4.1 and 5.1; 6.1; and 4.2, 5.2 and 7.1 prove (a), (b), (c), and (d), respectively, for precisely the stated radial operator and parameter range. The explicit regular series, prescribed Picard iteration and unique IVP patching require no new arbitrary choice of solutions. The public Countable Choice assumption is retained for the invoked IVP and Hilbert-space suppliers; the public Axiom of Choice assumption is retained for the existing transported radial-operator supplier contract. No reduction of that supplier's choice hypothesis is inferred from this local argument.

Remarks

Cauchy-Schwarz in the forms $∣ \int f \overline{g} ∣ \leq ∥ f ∥_{2} ∥ g ∥_{2}$ and the finite-dimensional estimate on $C^{2}$ (Cauchy-Schwarz inequality for $L^{2}$ , Cauchy–Schwarz: $∣ ⟨ u, v ⟩ ∣ \leq ∥ u ∥ ∥ v ∥$ , with equality exactly for linearly dependent vectors, The space $L^{p} (μ)$ as the quotient by null functions).

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Coulomb continuum spectral density

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ \geq 0$ , $l \geq 0$ , let $A_{l}$ be the radial operator of Transported radial endpoint realization with spectral measure $E_{A_{l}}$ , and for $k > 0$ let $φ_{l} (k, \cdot)$ and $f_{l} (k, \cdot)$ be the solutions of Coulomb Jost Volterra solution, with $W_{l} (k) : = W (φ_{l} (k, \cdot), f_{l} (k, \cdot)) \neq 0$ . Write $ρ_{l} (E) : = \frac{\sqrt{E}}{π ∣ W_{l} (\sqrt{E}) ∣^{2}} (E > 0) .$

For $g \in L^{2} (0, \infty)$ write $μ_{g} (B) : = ⟨ E_{A_{l}} (B) g, g ⟩$ for its finite positive scalar spectral measure, where $B \subseteq R$ is Borel (Projection valued measure, Spectral theorem for unbounded self-adjoint operators (PVM form)). A finite positive Borel measure $μ$ is absolutely continuous with respect to Lebesgue measure $d E$ if $μ (N) = 0$ for every Borel set $N$ of Lebesgue measure zero; it is singular continuous if it is carried by a Borel Lebesgue-null set and has no atoms, meaning $μ ({E}) = 0$ for every $E \in R$ . Define the absolutely continuous subspace $H_{a c} (A_{l}) : = {g \in L^{2} (0, \infty) : μ_{g} is absolutely continuous with respect to d E} .$ It is a closed linear subspace because it equals $⋂_{N : ∣ N ∣ = 0} \ker E_{A_{l}} (N)$ , where the intersection is over Borel Lebesgue-null sets and $μ_{g} (N) = ∥ E_{A_{l}} (N) g ∥^{2}$ . For $I = (0, \infty)$ , put $H_{s c} (A_{l}; I) : = {g \in E_{A_{l}} (I) L^{2} (0, \infty) : μ_{g} is singular continuous} .$ The set $H_{s c} (A_{l}; I)$ is a closed linear subspace: linear combinations of such vectors have scalar measures supported on the union of their null carriers and have no atoms, and these two properties pass to limits of vectors by continuity of the PVM projections. Here $E_{A_{l}} (I) L^{2}$ is purely absolutely continuous when it is contained in $H_{a c} (A_{l})$ , and “ $A_{l}$ has no singular-continuous spectrum on $I$ ” means $H_{s c} (A_{l}; I) = {0}$ . Thus each operator spectral-type phrase in (d) is a statement about the scalar PVM measures, not an additional pointwise classification of energies.

Then:

(a) resolvent kernel. For $z = k^{2}$ with $Im k > 0$ and $k$ in the domain of the Jost construction, and for every $g \in L^{2} (0, \infty)$ with compact support, the resolvent is given by $((A_{l} - z)^{- 1} g) (r) = \int_{0}^{\infty} G_{l} (r, s) g (s) d s, G_{l} (r, s) : = - \frac{φ_{l} (k, \min (r, s)) f_{l} (k, \max (r, s))}{W_{l} (k)},$ and $G_{l}$ solves $(A_{l} - z) G_{l} = δ$ in the sense that $- G^{''} + (l (l + 1) / r^{2} - γ / r) G - z G = 0$ away from $r = s$ and the derivative of $G$ has the jump $- 1$ at $r = s$ ; moreover $∥ (A_{l} - z)^{- 1} ∥ \leq 1 / ∣ Im z ∣$ ;

(b) boundary value. For real $k > 0$ one has $φ_{l} (k, \cdot)$ real and, for all $r, s > 0$ , $Im G_{l} (k^{2} + i 0; r, s) = \frac{k φ_{l} (k, r) φ_{l} (k, s)}{∣ W_{l} (k) ∣^{2}};$

(c) spectral density. For every $g \in L^{2} (0, \infty)$ and every $0 < a < b$ , $⟨ E_{A_{l}} ((a, b)) g, g ⟩ = \lim_{ε ↓ 0} \frac{1}{π} \int_{a}^{b} Im ⟨ (A_{l} - E - i ε)^{- 1} g, g ⟩ d E,$ and if $g$ has compact support then $⟨ E_{A_{l}} ((a, b)) g, g ⟩ = \int_{a}^{b} ρ_{l} (E) ∣ \int_{0}^{\infty} φ_{l} (\sqrt{E}, r) g (r) d r ∣^{2} d E;$ consequently for compactly supported $g$ the measure $E_{A_{l}, g} (B) : = ⟨ E_{A_{l}} (B) g, g ⟩$ restricted to $(0, \infty)$ is absolutely continuous with density $ρ_{l} (E) ∣ \int φ_{l} g ∣^{2}$ ;

(d) absolutely continuous spectrum; no zero eigenvalue. For every $g \in L^{2} (0, \infty)$ the measure $E_{A_{l}, g}$ has no atom in $(0, \infty)$ ; the positive spectral subspace $E_{A_{l}} ((0, \infty)) L^{2} (0, \infty)$ is purely absolutely continuous and is the closed span of the positive spectral projections of compactly supported functions; it equals the absolutely continuous subspace of $A_{l}$ , and $A_{l}$ has no singular continuous spectrum on $(0, \infty)$ ; moreover $0$ is not an eigenvalue of $A_{l}$ .

No scattering phase, normalization of the continuum functions as a unitary transform, or physical cross-section is asserted; the transform itself is the subject of the companion item Onto radial continuum transform.

Facts & Assumptions

Given: the current AC+CC contract, $γ \geq 0$ , $l \geq 0$ , $q_{l} (r) = l (l + 1) / r^{2} - γ / r$ , the radial form realization $A_{l}$ , $φ_{l} / r^{l + 1} \to 1$ , $f_{l}$ with the stated Coulomb logarithmic phase, $W (u, v) = u v^{'} - u^{'} v$ , and $⟨ u, v ⟩ = \int u \overline{v}$ (linear in the first slot). Compact support means containment in a compact subset of the open half-line.

[F1]

The endpoint realization supplies $V_{l} = H_{0}^{1} (0, \infty)$ with equivalent norm, $u / r \in L^{2}$ for $u \in V_{l}$ , the continuous closed form $a_{l}$ , and the exact operator-domain criterion: $u \in D (A_{l})$ if and only if $u \in V_{l}$ and $a_{l} (u, v) = ⟨ h, v ⟩$ for every $v \in V_{l}$ and some $h \in L^{2}$ . It also supplies self-adjointness and $U C U^{*} = ⨁ A_{l}$ (Transported radial endpoint realization). The nonreal resolvent bound is Resolvent of a self-adjoint operator: nonreal resolvents and the estimate.

[F2]

The regular series and its derivatives converge uniformly on compact parameter and radial sets and have $φ_{l} = r^{l + 1} (1 + O (r))$ , $φ_{l}^{'} = O (r^{l})$ at zero. The Jost solution and its radial derivative are jointly continuous up to the real positive $k$ axis and decay exponentially times a power when $Im k > 0$ in its first-quadrant construction domain. For real $k > 0$ , $φ_{l}$ is real, $W ({\overline{f}}_{l}, f_{l}) = 2 i k$ , and $W_{l} (k) \neq 0$ (Coulomb Jost Volterra solution, especially its actual compact-parameter Volterra/IVP proof).

[F3]

The self-adjoint spectral theorem gives the finite positive scalar measure $μ_{g} (B) = ∥ E_{A_{l}} (B) g ∥^{2}$ , total mass $∥ g ∥^{2}$ , and $⟨ (A_{l} - z)^{- 1} g, g ⟩ = \int (λ - z)^{- 1} d μ_{g} (λ)$ . The projection at $λ$ is the projection onto $\ker (A_{l} - λ)$ (Spectral theorem for unbounded self-adjoint operators (PVM form), Unbounded Borel functional calculus: domains, products, spectral mapping, Projection valued measure).

[F4]

Cauchy–Schwarz, Fubini, dominated convergence, and continuity from below for finite measures are used in their scalar forms (Cauchy-Schwarz inequality for $L^{2}$ , Tonelli and Fubini for the completed product, with only almost-everywhere section measurability, Dominated convergence). No Stone interval formula, positive-energy absence of atoms, or onto continuum transform is an input.

[F5]

The form-core identity $∥ u^{'} - u / r ∥_{2} = ∥ u^{'} ∥_{2}$ and Hardy inequality $∥ u / r ∥_{2} \leq 2 ∥ u^{'} ∥_{2}$ extend to $V_{l}$ . Young's inequality yields $a_{l} (u, u) \geq \frac{1}{2} ∥ u^{'} ∥_{2}^{2} + l (l + 1) ∥ u / r ∥_{2}^{2} - 2 γ^{2} ∥ u ∥_{2}^{2}$ . This bound is uniform over $l \geq 0$ and follows from [F1]; no essential-spectrum theorem or negative bound-energy exhaustion is assumed below.

Proof

1.1F1F2F4givenalgebra

[Endpoint and form membership] Fix $z = k^{2}$ with $Im k > 0$ in the Jost construction domain; its positive real part makes $z$ nonreal. A function agreeing with a constant multiple of $φ_{l}$ near zero and a constant multiple of $f_{l}$ near infinity and lying in $H^{2}$ on intervening compact intervals has $u, u^{'}, u / r \in L^{2}$ and trace zero at zero by [F2]. It belongs to $V_{l}$ : a cutoff vanishing below $δ$ and equal to one above $2 δ$ adds derivative squared norm at most $C \int_{0}^{2 δ} ∣ u ∣^{2} / r^{2} \to 0$ ; a cutoff at infinity adds derivative squared norm at most $C R^{- 2} \int_{R}^{2 R} ∣ u ∣^{2} \to 0$ . The tails of all other norm terms vanish by integrability, and the resulting compact $H^{1}$ functions can be smoothed within the open half-line. This verifies actual form membership including $l = 0$ and $γ = 0$ .

1.2F1F3F4givenalgebra

[Radial form proof of zero exclusion] Define $S_{s} u (r) = e^{s / 2} u (e^{s} r)$ for real $s$ . Change of variables gives unitarity on $L^{2}$ , preservation of $V_{l}$ , and on the smooth form core the exact identity $a_{l} (S_{s} u, v) = e^{2 s} a_{l} (u, S_{- s} v) + γ (e^{2 s} - e^{s}) ⟨ S_{s} (u / r), v ⟩$ . Kinetic and angular terms scale by $e^{2 s}$ and the Coulomb term by $e^{s}$ . Continuity extends the identity to all $u, v \in V_{l}$ , since $u / r \in L^{2}$ . If $u \in D (A_{l})$ and $A_{l} u = 0$ , the form criterion gives $S_{s} u \in D (A_{l})$ and $A_{l} S_{s} u = γ (e^{2 s} - e^{s}) S_{s} (u / r)$ . Self-adjointness yields $0 = ⟨ A_{l} S_{s} u, u ⟩ = γ (e^{2 s} - e^{s}) ⟨ S_{s} (u / r), u ⟩$ . For $γ > 0$ and $s \neq 0$ divide by the nonzero coefficient. Strong continuity $S_{s} h \to h$ follows first for compact continuous $h$ by dominated convergence and then by density and unitarity for all $h \in L^{2}$ . Letting $s \to 0$ gives $\int ∣ u ∣^{2} / r = 0$ , hence $u = 0$ . For $γ = 0$ , $0 = ⟨ A_{l} u, u ⟩ = ∥ u^{'} - u / r ∥_{2}^{2} + l (l + 1) ∥ u / r ∥_{2}^{2}$ ; the identity $∥ u^{'} - u / r ∥_{2} = ∥ u^{'} ∥_{2}$ extends from the form core to $V_{l}$ , so $u^{'} = 0$ and its $L^{2}$ constant representative is zero. No unbounded dilation generator is differentiated.

1.3F1F4F5givenalgebra

[Compact weighted form embedding] The map $K : V_{l} \to L^{2} (0, \infty)$ , $K u = u / \sqrt{r}$ , is compact on bounded $H_{0}^{1}$ sets. To prove this directly, for $∥ u ∥_{2}, ∥ u^{'} ∥_{2} \leq C$ use $\int_{0}^{δ} ∣ u ∣^{2} / r \leq δ ∥ u / r ∥_{2}^{2} \leq 4 δ C^{2}$ and $\int_{R}^{\infty} ∣ u ∣^{2} / r \leq C^{2} / R$ . On $[δ, R]$ , partition into finitely many equal cells of length at most $h$ and replace $u$ by its average on each cell. The absolute continuity formula and Cauchy–Schwarz give $\int_{I} ∣ u - u_{I} ∣^{2} \leq h^{2} \int_{I} ∣ u^{'} ∣^{2}$ on each cell $I$ , so multiplication by $1 / \sqrt{r}$ gives total middle error at most $h^{2} C^{2} / δ$ . The resulting functions lie in a fixed finite-dimensional span of cell indicators divided by $\sqrt{r}$ , with bounded coefficients. Choosing $δ, R, h$ makes the error arbitrarily small, so the image has finite nets of every radius and is relatively compact in complete $L^{2}$ . This explicit finite-rank approximation works for real $l \geq 0$ as well as the angular integer labels.

2.1step 1.1F1F2algebra

[Nonreal Wronskian] If $W_{l} (k) = 0$ , regular-IVP uniqueness makes $φ_{l}$ and $f_{l}$ proportional, giving a nonzero common solution $u$ . Step 1.1 puts $u$ in $V_{l}$ . Integration by parts against compact smooth $v$ gives $a_{l} (u, v) = z ⟨ u, v ⟩$ , since the derivative form equals $\int u^{'} \overline{v^{'}}$ by cancellation of the derivative of $u \overline{v} / r$ . Form continuity extends this to every $v \in V_{l}$ . The exact criterion [F1] gives $u \in D (A_{l})$ and $A_{l} u = z u$ , contradicting self-adjointness. Thus the nonreal Wronskian is nonzero.

2.2step 1.3F1F3F5algebra

[Finite negative spectral bands] Assume $γ > 0$ , put $M = 2 γ^{2} + 1$ , and fix $ε > 0$ . The range $H_{ε} = E_{A_{l}} ([- M, - ε]) L^{2}$ lies in $D (A_{l})$ by bounded energy support; for $u \in H_{ε}$ the PVM gives $a_{l} (u, u) \leq - ε ∥ u ∥^{2}$ . The lower bound [F5] gives $∥ u^{'} ∥_{2}^{2} \leq 4 γ^{2} ∥ u ∥_{2}^{2}$ , so the unit ball of $H_{ε}$ is $H_{0}^{1}$ -bounded. If this spectral range were infinite-dimensional, AC allows choosing an orthonormal sequence $u_{j}$ in it (successively choose a unit vector orthogonal to the preceding finite span). For $v = u_{j} - u_{i}$ , nonnegativity of the kinetic and angular terms and the negative-band bound give $γ ∥ K v ∥_{2}^{2} \geq ε ∥ v ∥_{2}^{2} = 2 ε$ . Thus the images $K u_{j}$ are separated by a fixed positive distance, contradicting the finite nets in step 1.3. Hence every negative band has finite-dimensional spectral range. This orthonormal-sequence selection uses the retained AC assumption; the proof does not claim a CC-only recursive choice argument.

3.1step 1.1step 2.1F1F2F4algebra

[Green sign, derivative jump and actual inverse] Write $φ = φ_{l} (k, \cdot)$ , $f = f_{l} (k, \cdot)$ , $W = W_{l} (k)$ . For $g$ supported in $[r_{1}, r_{2}] \subset (0, \infty)$ let $I (r) = \int_{0}^{r} φ (s) g (s) d s$ , $J (r) = \int_{r}^{\infty} f (s) g (s) d s$ , and $u = - (f I + φ J) / W$ . The integrals are absolutely continuous; cross terms cancel in $u^{'} = - (f^{'} I + φ^{'} J) / W$ , and $u^{''} = - (f^{''} I + φ^{''} J) / W - g$ . Therefore $- u^{''} + (q_{l} - z) u = g$ almost everywhere. Below $r_{1}$ and above $r_{2}$ , $u$ is respectively a constant multiple of $φ$ and of $f$ , so step 1.1 gives $u \in V_{l}$ . Integration by parts against compact tests and form continuity give $a_{l} (u, v) = ⟨ z u + g, v ⟩$ for every $v \in V_{l}$ . Thus $u \in D (A_{l})$ and $u = (A_{l} - z)^{- 1} g$ . The kernel $G = - φ (r_{<}) f (r_{>}) / W$ is continuous and its derivative jump is $- [φ (s) f^{'} (s) - φ^{'} (s) f (s)] / W = - 1$ , giving $(L_{l} - z) G = δ_{s}$ distributionally. The stated inverse norm estimate is [F1].

3.2step 2.2F1F3F5algebra

[Negative scalar measures are countably atomic] Each band projection in step 2.2 has finite rank $N$ . Choose a finite orthonormal basis $e_{1}, \dots, e_{N}$ for its range and define $ν = \sum_{j} μ_{e_{j}}$ , supported in the band. Its support has at most $N$ points: otherwise $N + 1$ distinct support points admit disjoint open neighborhoods of positive $ν$ mass, hence $N + 1$ nonzero mutually orthogonal spectral projection ranges inside this $N$ -dimensional space. The complement of this support is the union of the rational open intervals of zero $ν$ mass, so it has zero $ν$ mass by countable subadditivity. Every scalar measure of a vector in the band is likewise supported on this finite set, since $ν (B) = 0$ makes $E (B)$ annihilate the basis and hence the entire band range. Exhausting negative energies by $[- M, - 1 / n]$ shows every radial negative scalar measure is supported on one countable union of finite sets. The lower bound [F5] excludes energies below $- M$ . For $γ = 0$ , the form is nonnegative and the negative projection is zero. This proves the negative pure-point fact locally for the whole stated real $l \geq 0$ range.

4.1step 3.1F2F4algebra

[Boundary algebra for complex sources] For real $k > 0$ , $Im (\overline{f} W) = φ Im (\overline{f} f^{'}) = k φ$ . Since $- f / W = - f \overline{W} / ∣ W ∣^{2}$ , conjugating $\overline{f} W$ gives $Im (- f / W) = k φ / ∣ W ∣^{2}$ , proving (b). For a complex source the kernel symmetry $G (r, s) = G (s, r)$ matters: the adjoint kernel on the compact support is $\overline{G (s, r)} = \overline{G (r, s)}$ , so $(R - R^{*}) / (2 i)$ has kernel $Im G$ . Consequently Fubini gives $Im ⟨ R (k^{2} + i 0) g, g ⟩ = k ∣ \int φ_{l} (k, r) g (r) d r ∣^{2} / ∣ W_{l} (k) ∣^{2}$ . This does not incorrectly distribute imaginary parts across complex source products.

5.1step 3.1step 4.1F2F4algebra

[Uniform compact-source limit] Fix $0 < a < b$ and compact source support $J \subset (0, \infty)$ . Choose $k_{\min} < \sqrt{a}$ . The set of first-quadrant roots ${\sqrt{E + i ε} : E \in [a, b], 0 \leq ε \leq ε_{0}}$ is compact inside the Jost parameter domain. The jointly continuous $φ, f, φ^{'}, f^{'}$ give continuous $W$ computed at any fixed radial point. Its real-axis modulus has a positive minimum on $[\sqrt{a}, \sqrt{b}]$ , hence a common positive lower bound for sufficiently small $ε_{0}$ . Uniform continuity therefore gives uniform convergence of $G (E + i ε; r, s)$ to its boundary value for $E \in [a, b]$ and $r, s \in J$ . Since $∥ g ∥_{1} \leq ∣ J ∣^{1 / 2} ∥ g ∥_{2}$ , the paired error is at most the supremum kernel error times $∥ g ∥_{1}^{2}$ . Pairings converge uniformly and step 4.1 gives their imaginary parts. No global limiting-absorption norm is assumed.

6.1step 5.1F3F4algebra

[Poisson approximation from the PVM] For continuous $h$ compactly supported in $(0, \infty)$ , extended by zero to $R$ , [F3] and Fubini give $π^{- 1} \int h (E) Im ⟨ (A_{l} - E - i ε)^{- 1} g, g ⟩ d E = \int (P_{ε} * h) (λ) d μ_{g} (λ)$ , where $P_{ε} (t) = ε / [π (t^{2} + ε^{2})]$ . Absolute integrability follows from $\int P_{ε} = 1$ and finite mass $∥ g ∥^{2}$ . Uniform continuity bounds $∣ h (x - t) - h (x) ∣$ by any prescribed $η > 0$ for $∣ t ∣ \leq δ$ ; the remainder is at most $2 ∥ h ∥_{\infty} \int_{∣ t ∣ > δ} P_{ε} \leq 4 ∥ h ∥_{\infty} ε / (π δ)$ . Hence $P_{ε} * h \to h$ uniformly. For compact-source $g$ , step 5.1 computes the other limit, yielding $\int h d μ_{g} = \int_{0}^{\infty} h (E) d_{g} (E) d E$ , with continuous nonnegative $d_{g} (E) = ρ_{l} (E) ∣ \int φ_{l} (\sqrt{E}, r) g (r) d r ∣^{2}$ .

7.1step 6.1F3F4algebra

[Borel density without a regularity prerequisite] For $0 < a < b$ , continuous $h_{n} (E) = \min (1, n (E - a)_{+}, n (b - E)_{+})$ increase to $1_{(a, b)}$ and vanish outside $[a, b]$ . Step 6.1 and dominated convergence give $μ_{g} ((a, b)) = \int_{a}^{b} d_{g} (E) d E$ . Every positive singleton has zero mass by shrinking intervals around it and local boundedness of $d_{g}$ . On a fixed compact interval $[c, d] \subset (0, \infty)$ , the two finite measures agree on all subintervals, including endpoints. Agreement on every relative Borel set follows by the elementary measure uniqueness argument: agreement sets form a Dynkin class (equal total mass, relative complements, disjoint countable unions), and intervals form an intersection-stable generator. For completeness, let $D$ be the smallest Dynkin class containing these intervals and the full compact interval; for a fixed generating interval $I$ , the sets $B$ with $I \cap B \in D$ form a Dynkin class containing the generators, hence contain $D$ . Repeating this with any fixed $B \in D$ shows that $D$ is closed under intersections of its members. A Dynkin class closed under intersections is a sigma algebra by disjointifying countable unions, so it contains the generated Borel sigma algebra. Exhaustion by $[1 / n, n]$ proves equality on every positive Borel set. The same exhaustion bounds the total density integral by $∥ g ∥^{2}$ . This proves density before either Stone's interval formula or an onto transform, without a separate finite-measure regularity assumption.

8.1step 7.1F3F4algebra

[Arbitrary vectors] For arbitrary $g \in L^{2}$ , take explicit $g_{n} = 1_{[1 / n, n]} g \to g$ . Projection contraction gives $∣ μ_{g} (B) - μ_{g_{n}} (B) ∣ \leq (∥ g ∥ + ∥ g_{n} ∥) ∥ g - g_{n} ∥$ uniformly over Borel $B$ . Every positive Lebesgue-null Borel set has zero $μ_{g_{n}}$ mass by step 7.1 and hence zero $μ_{g}$ mass. Thus every scalar spectral measure restricted to $(0, \infty)$ is absolutely continuous and every positive atom vanishes. These explicit cutoffs require no additional choice.

9.1step 8.1step 7.1F3F4algebra

[Exact interval Stone formula] PVM representation and Fubini give $π^{- 1} \int_{a}^{b} Im ⟨ (A_{l} - E - i ε)^{- 1} g, g ⟩ d E = \int F_{ε} (λ) d μ_{g} (λ)$ , with $F_{ε} (λ) = π^{- 1} [\arctan ((b - λ) / ε) - \arctan ((a - λ) / ε)]$ . It lies in $[0, 1]$ and tends to $1_{(a, b)} + \frac{1}{2} 1_{{a, b}}$ . Dominated convergence and the already proved zero endpoint atoms yield the exact interval formula for every $g$ . Step 7.1 gives its compact-source density integral.

9.2step 8.1F3algebra

[Positive subspace] Put $P_{+} = E_{A_{l}} ((0, \infty))$ . If $g = P_{+} g$ , its entire scalar measure is positive-energy supported and absolutely continuous by step 8.1. Thus this subspace has neither pure-point nor singular-continuous part. Also $P_{+} g_{n} \to P_{+} g$ , so positive spectral projections of compact-source vectors span it densely. This permits unprojected compact-source vectors to have negative-energy components.

10.1step 3.1step 4.1step 7.1step 9.1step 9.2step 1.2step 3.2F3∎

[Full AC subspace and conclusion] An absolutely continuous scalar measure gives zero mass to the countable negative support from step 3.2 and to ${0}$ . Hence its vector lies in $P_{+} L^{2}$ . Step 9.2 proves the reverse inclusion, so the full absolutely continuous subspace equals the positive spectral subspace, establishing corrected clause (d), whose zero exclusion is step 1.2. Clauses (a), (b), (c) follow from the Green inverse, boundary algebra, and PVM Poisson arguments above. The regular normalization and density $\sqrt{E} / (π ∣ W_{l} (\sqrt{E}) ∣^{2})$ are unchanged, and neither the onto transform nor any angular transport at noninteger $l$ was assumed.

TheoremStatement: AI-adaptedProof: AI-adaptedOpen item page →

Onto radial continuum transform

Statement

Assume the Axiom of Countable Choice and the Axiom of Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ ), The Axiom of Choice). Let $γ \geq 0$ , $l \geq 0$ , let $A_{l}$ be the radial operator of Transported radial endpoint realization with spectral measure $E_{A_{l}}$ and form domain $V_{l}$ , and for $E > 0$ let $φ_{l} (\sqrt{E}, \cdot)$ and $ρ_{l} (E) = \frac{\sqrt{E}}{π ∣ W_{l} (\sqrt{E}) ∣^{2}}$ be as in Coulomb continuum spectral density. Define, for $g$ in the dense subspace $C_{c}$ of compactly supported square-integrable functions, $(T_{l} g) (E) : = \sqrt{ρ_{l} (E)} \int_{0}^{\infty} φ_{l} (\sqrt{E}, r) g (r) d r (E > 0) .$ Then:

(a) $T_{l} g \in L^{2} ((0, \infty), d E)$ and $∥ T_{l} g ∥^{2} = ⟨ E_{A_{l}} ((0, \infty)) g, g ⟩$ ; consequently $T_{l}$ extends uniquely to a linear contraction $T : H \to L^{2} ((0, \infty), d E)$ , where $H : = L^{2} ((0, \infty), d r)$ and $P_{+} : = E_{A_{l}} ((0, \infty))$ , with $∥ T g ∥ = ∥ P_{+} g ∥$ for every $g \in H$ . Its restriction is a linear isometry ${\hat{T}}_{l} : E_{A_{l}} ((0, \infty)) L^{2} (0, \infty) ⟶ L^{2} ((0, \infty), d E)$ of the positive spectral subspace of $A_{l}$ into $L^{2} ((0, \infty), d E)$ , satisfying $T = {\hat{T}}_{l} P_{+}$ and $T_{l} g = {\hat{T}}_{l} (P_{+} g)$ for $g \in C_{c}$ . Equivalently, $P_{+} g \mapsto T_{l} g$ is a well-defined isometry on the dense projected core $P_{+} C_{c} \subseteq P_{+} H$ and ${\hat{T}}_{l}$ is its unique isometric extension;

(b) ${\hat{T}}_{l}$ is onto, hence unitary; equivalently the range of ${\hat{T}}_{l}$ is all of $L^{2} ((0, \infty), d E)$ and the adjoint ${\hat{T}}_{l}^{*}$ is an isometry inverse to ${\hat{T}}_{l}$ ;

(c) ${\hat{T}}_{l} A_{l} {\hat{T}}_{l}^{*}$ is multiplication by the independent variable on the domain ${h \in L^{2} ((0, \infty), d E) : \int E^{2} ∣ h (E) ∣^{2} d E < \infty}$ , and ${\hat{T}}_{l} h (A_{l}) {\hat{T}}_{l}^{*}$ is multiplication by $h$ for every bounded Borel function $h$ ;

(d) complete Coulomb spectral resolution. If $γ > 0$ , the operator $C = - Δ - γ / ∣ \cdot ∣$ on $H^{2} (R^{3}; C)$ is unitarily equivalent to the diagonal operator $⨁_{(l, m)} [(⨁_{n \geq l + 1} C) \oplus L^{2} ((0, \infty), d E)]$ with the constant $- γ^{2} / (4 n^{2})$ on the $(n, l, m)$ -copy of $C$ and multiplication by $E$ on each continuum copy; equivalently there is a unitary $V$ from $L^{2} (R^{3}; C)$ onto the displayed direct sum with $V C V^{*}$ equal to that diagonal operator, and the negative part of the spectrum is exhausted by the bound states of Negative-subspace completeness and degeneracy while the positive part is purely absolutely continuous. If $γ = 0$ , the representation is $⨁_{(l, m)} L^{2} ((0, \infty), d E)$ with multiplication by $E$ and no discrete copies, because the free Fourier multiplier has no nonpositive spectral projection.

The transform is a mathematical spectral representation of $A_{l}$ and $C$ ; no scattering amplitude, phase shift or physical interpretation is asserted.

Facts & Assumptions

Given: $γ \geq 0$ , $l \geq 0$ , the operator $A_{l}$ , its spectral measure $E_{A_{l}}$ , the density $ρ_{l}$ , the solutions $φ_{l}$ , and the unitary $U$ of Angular radial unitary reduction.

[F1]

For compactly supported $g, h$ and every $0 < a < b$ : $⟨ E_{A_{l}} ((a, b)) g, h ⟩ = \int_{a}^{b} ρ_{l} (E) (\int φ_{l} (\sqrt{E}, r) g (r) d r) \overline{(\int φ_{l} (\sqrt{E}, r) h (r) d r)} d E$ ; the measures have no atoms in $(0, \infty)$ , $ρ_{l} (E) > 0$ , and $φ_{l} (\sqrt{E}, \cdot) \neq 0$ in $L_{l o c}^{2}$ for every $E > 0$ (Coulomb continuum spectral density, Coulomb Jost Volterra solution).

[F2]

Compactly supported functions are dense in $L^{2} (0, \infty)$ , and a bounded linear isometry defined on a dense subspace extends uniquely to the closure with the same norm (A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm, The space $L^{p} (μ)$ as the quotient by null functions, Hilbert space).

[F3]

Under the retained public AC hypothesis, Spectral theorem for unbounded self-adjoint operators (PVM form) and Unbounded Borel functional calculus: domains, products, spectral mapping give the unique regular PVM, bounded Borel calculus, nonreal resolvent multiplier and exact operator-domain norm formula. Here $H = L^{2} ((0, \infty), d r)$ , $K = L^{2} ((0, \infty), d E)$ and $H_{+} = P_{+} H$ , $P_{+} = E_{A_{l}} ((0, \infty))$ . Projections on disjoint spectral sets are orthogonal, bounded Borel pairings integrate mixed measures, and the restricted $A_{+} = A_{l} ∣_{D (A_{l}) \cap H_{+}}$ is self-adjoint with the exact restricted spectral domain: its nonreal resolvent is the restriction of the full multiplier resolvent, so both nonreal ranges are onto. The inner product is linear in the first argument. Resolvent of a self-adjoint operator: nonreal resolvents and the estimate supplies $(A_{+} - z)^{- 1}$ with norm at most $∣ Im z ∣^{- 1}$ . This is an honest AC-bearing prerequisite use, retaining the stated AC and CC hypotheses on the later source40 CC lemma.

[F4]

A projection on $K$ commuting with all indicator multipliers is a scalar indicator multiplier, as proved locally in step 5.1. Orthogonal decomposition by a closed subspace supplies the required orthogonal projection under CC; dominated convergence and Cauchy-Schwarz are supplied by Dominated convergence and Cauchy-Schwarz inequality for $L^{2}$ .

[F5]

Fix the countable family $g_{j}$ of all finite rational-complex linear combinations of indicators $1_{(a, b)}$ , with $0 < a < b < \infty$ rational. Each is a compact source, and on each $(1 / k, k)$ , $k \geq 2$ , the members supported there are dense in $L^{2} (1 / k, k)$ by simple approximation and regularity. The regular solution is real for real $k$ and has locally uniform parameter bounds, including $φ_{l} (k, r) = r^{l + 1} [1 + O (r)]$ near zero, by Coulomb Jost Volterra solution. Compact positive-energy intervals have bounded continuous $ρ_{l}$ and bounded regular solutions on compact radial intervals. Tonelli and Fubini for the completed product, with only almost-everywhere section measurability justifies the compact double integrals below. The free Fourier multiplier and its zero-set property use Plancherel theorem and the transported endpoint realization.

Proof

1.1F1F2givenalgebra

[Projected norm identity on compact sources] For compactly supported $g$ , [F1] with $a ↓ 0$ , $b ↑ \infty$ (or the monotone convergence of the positive measures on the increasing intervals $(1 / k, k)$ ) gives $∥ T_{l} g ∥_{L^{2} ((0, \infty), d E)}^{2} = \int_{0}^{\infty} ρ_{l} (E) ∣ \int φ_{l} (\sqrt{E}, r) g (r) d r ∣^{2} d E = ⟨ E_{A_{l}} ((0, \infty)) g, g ⟩$ ; the right side is finite because it is the mass of the finite measure $E_{A_{l}, g}$ on $(0, \infty)$ . This is the norm identity of (a) on the dense compact-source domain $C_{c}$ .

2.1step 1.1F2F3algebra

[Extension through the positive projection] Step 1.1 gives $∥ T_{l} g ∥ \leq ∥ g ∥$ on compact sources, so extend $T_{l}$ first to a contraction $T$ on all $L^{2}$ . Passing to compact-source approximations yields $∥ T g ∥^{2} = ⟨ P_{+} g, g ⟩$ for every $g$ , where $P_{+} = E_{A_{l}} ((0, \infty))$ . Thus $T (I - P_{+}) = 0$ and $\hat{T} = T ∣_{P_{+} H}$ is an isometry, with $T = \hat{T} P_{+}$ . If $P_{+} g = P_{+} h$ for compact sources, the norm identity gives $T_{l} (g - h) = 0$ , so the induced map $P_{+} g \mapsto T_{l} g$ is well-defined and isometric. The projected core $P_{+} C_{c}$ is dense in $P_{+} H$ : project compact-source approximations to any vector in $P_{+} H$ and use $∥ P_{+} ∥ \leq 1$ . This proves the unique-extension and factorization assertions of (a). Its range is closed because Cauchy sequences in that range pull back to Cauchy sequences in the complete positive subspace. In particular compact-source images are dense in its range, without incorrectly calling $T_{l}$ isometric on all compact sources.

3.1step 2.1F1F3F4algebra

[Compressed resolvents and the missing norm equality] Write $W = \hat{T} : H_{+} \to K$ , $M = ran W$ and $Q$ for the orthogonal projection onto this closed range. Define $W^{*} = W^{- 1} Q$ : uniqueness of orthogonal decomposition gives directly $⟨ W u, v ⟩ = ⟨ u, W^{*} v ⟩$ , $W^{*} W = I_{H_{+}}$ and $W W^{*} = Q$ , without invoking another representation theorem. Polarize [F1] on compact sources; the mixed measures are finite, and equality on positive intervals extends to all positive Borel sets by the interval-generated finite-measure uniqueness argument. Integration of bounded simple functions and bounded approximation then gives $⟨ M_{b} T g, T h ⟩ = ⟨ b_{+} (A_{l}) g, h ⟩$ , where $b_{+}$ is $b$ on the positive half-line and zero elsewhere. Boundedness extends this identity from compact sources to all $g, h \in H$ . Thus $W^{*} M_{b} W = b (A_{+})$ , initially only a compression. For $z = x + i y$ , $y \neq 0$ , put $m_{z} (E) = (E - z)^{- 1}$ and $R_{z} = (A_{+} - z)^{- 1}$ . The identity gives $Q M_{m_{z}} W u = W R_{z} u$ . Pointwise $Im m_{z} = y ∣ m_{z} ∣^{2}$ , so $Im ⟨ M_{m_{z}} W u, W u ⟩ = y ∥ M_{m_{z}} W u ∥^{2}$ . Setting $v = R_{z} u$ and $u = (A_{+} - z) v$ , self-adjointness and the linear-first convention give $Im ⟨ R_{z} u, u ⟩ = Im ⟨ v, (A_{+} - z) v ⟩ = y ∥ R_{z} u ∥^{2}$ . The compressed pairing identifies these imaginary parts. Dividing by the nonzero real $y$ proves $∥ M_{m_{z}} W u ∥^{2} = ∥ R_{z} u ∥^{2} = ∥ W R_{z} u ∥^{2} = ∥ Q M_{m_{z}} W u ∥^{2}$ . Pythagoras forces $(I - Q) M_{m_{z}} W u = 0$ . Apply this to both $z$ and $\overline{z}$ , whose multipliers are adjoints: $M$ and $M^{⊥}$ reduce these multipliers, $Q$ commutes with them, and $M_{m_{z}} W = W R_{z}$ . This is the required compression-to-reduction proof before any onto claim.

4.1step 3.1F4algebra

[Resolvent reduction implies indicator reduction] For $0 < a < b$ and $ε > 0$ , define $p_{ε, a, b} (E) = π^{- 1} \int_{a}^{b} ε / [(E - t)^{2} + ε^{2}] d t$ . Its multiplier commutes with $Q$ : it is the operator-norm Riemann-integral limit of the imaginary parts of $M_{m_{t + i ε}}$ , since $∥ M_{m_{t + i ε}} - M_{m_{s + i ε}} ∥ \leq ∣ t - s ∣ / ε^{2}$ . One has $0 \leq p_{ε, a, b} \leq 1$ , and these functions tend as $ε = 1 / n ↓ 0$ to $1_{(a, b)}$ except at the two Lebesgue-null endpoints. Dominated convergence against $∣ h ∣^{2}$ for each $h \in K$ gives strong convergence of the multipliers, so $Q$ commutes with $M_{1_{(a, b)}}$ . The Borel sets whose multipliers commute with $Q$ form a sigma-algebra: complements use $I - M_{1_{B}}$ , intersections use products, and disjoint countable unions use strongly convergent finite sums; disjointification gives arbitrary countable unions. Positive intervals generate the Borel sigma-algebra, so all Borel indicators commute with $Q$ . Bounded simple approximation gives all bounded Borel multipliers; completed-measurable sets have the same multiplier as a Borel representative modulo null sets.

5.1step 4.1F4constructalgebra

[The reducing projection is a scalar indicator] Set $q (E) = e^{- E} > 0$ in $K$ , $ψ = Q q$ and $a = ψ / q$ using measurable representatives. Commutation gives $Q (b q) = b ψ$ for bounded measurable $b$ . For every measurable $B$ , contractivity gives $\int_{B} ∣ ψ ∣^{2} \leq \int_{B} q^{2}$ , hence $∣ a ∣ \leq 1$ a.e.; otherwise intersecting ${∣ a ∣ > 1 + 1 / n}$ with a bounded positive interval gives a contradiction to this inequality. Vectors $b q$ are dense: for $h \in K$ , $h_{n} = 1_{{1 / n < E < n, ∣ h / q ∣ \leq n}} h$ have that form and converge in $K$ by dominated convergence. Therefore $Q = M_{a}$ on $K$ . Self-adjointness and idempotence force $a = \overline{a} = a^{2}$ a.e., testing the multiplier identities on bounded-interval indicators. Thus $Q = M_{1_{S}}$ for a measurable $S$ , and $M = L^{2} (S)$ .

6.1step 5.1F1F4F5algebra

[Countable compact tests eliminate the complement] For the explicit family $g_{j}$ of [F5], each $T_{l} g_{j}$ lies in $M = L^{2} (S)$ and vanishes a.e. on $S^{c}$ . Remove the union of the countably many exceptional null sets. For every remaining $E \in S^{c}$ , positivity of $ρ_{l} (E)$ gives $\int φ_{l} (\sqrt{E}, r) g_{j} (r) d r = 0$ for every $j$ . On each $(1 / k, k)$ , $k \geq 2$ , this is a continuous linear functional by Cauchy-Schwarz and local square integrability of $φ_{l}$ . Its vanishing on the fixed dense family implies its vanishing on all of $L^{2} (1 / k, k)$ , including $\overline{φ_{l}} 1_{(1 / k, k)}$ . Hence $\int_{1 / k}^{k} ∣ φ_{l} ∣^{2} = 0$ . Exhausting $(0, \infty)$ contradicts the nonzero continuous regular solution of [F1]. The remaining part of $S^{c}$ is therefore empty, so $S^{c}$ is null, $Q = I_{K}$ , and $W$ is unitary with isometric inverse $W^{*}$ . Only countably many conull sets were intersected. This proves (b).

7.1step 3.1step 4.1step 6.1F3algebra

[Exact bounded and unbounded diagonalization] Steps 3.1, 4.1 and 6.1 give $\hat{T} E_{A_{l}} (B) ∣_{P_{+} H} = M_{1_{B}} \hat{T}$ for every positive Borel set. Linear combinations and bounded pointwise simple approximation therefore give $\hat{T} h (A_{l}) ∣_{P_{+} H} {\hat{T}}^{*} = M_{h ∣_{(0, \infty)}}$ for every bounded Borel $h$ . This uses the onto-ness of step 6.1 and does not incorrectly pair negative components through $T$ . For unbounded energy, spectral truncation and the norm formula give $\int E^{2} ∣ \hat{T} g (E) ∣^{2} d E = \int_{(0, \infty)} λ^{2} d E_{g} (λ)$ , so the exact generator domain is ${h : E h \in L^{2} (d E)}$ and $\hat{T} A_{l} ∣_{P_{+} H} {\hat{T}}^{*} = M_{E}$ there.

7.2step 2.1step 6.1F4F5algebra

[Inverse integral and exact scope of convergence] For $h \in K$ put $h_{n} = 1_{(1 / n, n)} h$ and $v_{n} (r) = \int_{1 / n}^{n} \sqrt{ρ_{l} (E)} φ_{l} (\sqrt{E}, r) h (E) d E$ . The compact parameter bounds of [F5] and Cauchy-Schwarz give $v_{n} \in L_{l o c}^{2} (0, \infty)$ . Fubini for compact-source $g$ gives $⟨ T g, h_{n} ⟩ = ⟨ g, v_{n} ⟩$ . Since $T = W P_{+}$ and $W^{*} h_{n} \in H_{+}$ , the left side also equals $⟨ g, W^{*} h_{n} ⟩$ . All compact-source tests therefore identify $v_{n}$ locally a.e. with $W^{*} h_{n}$ , so $v_{n}$ represents a global $L^{2}$ element of $H_{+}$ . Finally $h_{n} \to h$ in $K$ and $W^{*}$ is an isometry, giving $W^{*} h = \lim_{n \to \infty} v_{n}$ in $L^{2} (d r)$ . This explicitly realizes the inverse without asserting pointwise convergence at infinity. If compact source means bounded support that may approach zero, cutting it off below $1 / n$ and using the regular $r^{l + 1}$ bound of [F5] extends all compact-energy integral identities by Cauchy-Schwarz and $L^{2}$ continuity.

8.1step 7.1step 7.2F3F5givenalgebra

[Complete resolution of $C$ ] By the block identity of Transported radial endpoint realization and the angular unitary of Angular radial unitary reduction, $C$ is unitarily equivalent to the direct sum over $(l, m)$ of the operators $A_{l}$ , and by the previous items $A_{l}$ is unitarily equivalent to $M_{E}$ on $L^{2} ((0, \infty), d E)$ together with the discrete eigenvalues $- γ^{2} / (4 n^{2})$ , $n \geq l + 1$ , each simple, whose eigenspaces are spanned by the transported regular solutions (Negative-subspace completeness and degeneracy, Laguerre bound energy exhaustion); the essential spectrum is $[0, \infty)$ and the positive spectrum is purely absolutely continuous by Coulomb continuum spectral density. The absence of a zero eigenvector follows from Coulomb continuum spectral density. For $γ > 0$ , PVM transport from [F3] through the angular unitary and the exact block identity shows that the normalized global negative eigenfamily restricts to a complete radial negative family in each channel; otherwise placing a missing radial vector in one channel contradicts global completeness. Thus each radial unitary sends $u$ to $((⟨ u, u_{n l} / ∥ u_{n l} ∥ ⟩)_{n \geq l + 1}, W P_{+} u)$ , with inverse the $L^{2}$ bound-family sum plus step 7.2. Assembling these unitaries in the fixed countable channel order gives the stated onto unitary $V$ , since orthogonal norm sums give convergence and surjectivity. The exact block domain and step 7.1 identify its diagonal domain with square summability of all energy-multiplied discrete and continuum coordinates. For $γ = 0$ , use the transported free operator: its explicit Fourier multiplier $∣ p ∣^{2}$ has zero projection on $(- \infty, 0]$ , since ${p = 0}$ is null. PVM uniqueness and unitary transport in [F3] give $⨁_{l, m} E_{A_{l}} ((- \infty, 0]) = 0$ , so every individual nonpositive block projection is zero. Each free radial block is consequently wholly positive, and the same positive transforms assemble a continuum-only unitary. No zero-energy bound copies are introduced.

9.1step 1.1step 2.1step 3.1step 4.1step 5.1step 6.1step 7.1step 7.2step 8.1∎

[Conclusion] Clauses (a), (b), (c), (d) are steps 1.1 and 2.1; 3.1 through 6.1; 7.1; and 8.1. Step 7.2 explicitly realizes the isometric adjoint inverse.

5 · Examples, experiments and counterexamples

None yet.