The generator domain of the phase group is the multiplication domain #
Category: 1-Mathlib (CSD-free; staged for upstream).
BACKLOG #154 (out of #129).
PhaseGenerator.lean proved the forward half: if κ·f ∈ L² then the orbit t ↦ e^{−itκ} f is
differentiable with derivative −i κ f (hasDerivAt_phaseGroup_apply), and the phase group
preserves that domain (phaseGroup_mem_mulDomain). This file proves the converse: if the orbit
has a right derivative at 0 in L² — HasGeneratorAt (phaseGroup hκ) f w — then κ·f ∈ L².
The generator domain of the phase group is the multiplication domain.
- ★★
mem_mulDomain_of_hasGeneratorAt— the Fatou argument. Alonghₙ = 1/(n+1)the quotientsqₙ = hₙ⁻¹(e^{−ihₙκ} − 1) fconverge towinL², so‖qₙ‖² → ‖w‖²; pointwise their squared moduli converge to|κ f|²(tendsto_phaseDefect); Fatou's lemma (lintegral_liminf_le') gives∫ |κ f|² ≤ liminf ∫ |qₙ|² = ‖w‖² < ∞. - ★
hasGeneratorAt_phaseGroup_iff: the two readings of the domain agree, with the generator value−i κ f(L2.mulOp), uniqueness of the right derivative identifyingw. - The free Hamiltonian: ★★
mem_freePositionOp_domain_of_hasGeneratorAtcarries the statement across the Fourier transform — a vector whose free orbit has a right derivative at0lies infreePositionOp.domain. - ★★★
schrodinger_mem_freePositionOp_domain: the interacting orbit stays indom H₀, now in the operator sense; and ★★★hasDerivAt_schrodinger_orbit:i ∂ₜ U(t)ψ = (H₀ + V)(U(t)ψ)withH₀the self-adjoint operatorfreePositionOpacting on the orbit point. This closes the gap#129left: the mild solution is classical in the operator-theoretic sense.
Honest scope #
- The potential remains the bounded multiplier
potential VL; the identification withschrodingerOp'smulCLMfor a realVis acoeFnstatement not made here. - One dimension and
κ = freeSymbolfor the Schrödinger part, as throughout; the Fatou lemma is for any measurableκon anyE.
References: Reed–Simon I §VIII.4 (the generator of a multiplication group); PhaseGenerator.lean
(phaseDefect, tendsto_phaseDefect, coeFn_slope_phaseGroup, norm_sq_eq_integral),
PerturbedGenerator.lean (HasGeneratorAt), SchrodingerGenerator.lean
(hasGeneratorAt_freeGroup_schrodinger, schrodinger_classical); Mathlib
MeasureTheory.lintegral_liminf_le'; specs/BACKLOG.md #154, #129.
The quotients, pointwise #
The right quotient of the phase group at 0, pointwise: f · h⁻¹(e^{−ihκ} − 1).
★★ The Fatou argument #
★★ A vector whose phase-group orbit has a right derivative at 0 lies in the multiplication
domain: κ·f ∈ L². Fatou along hₙ = 1/(n+1).
★ The two readings of the domain agree. A vector is in the generator domain of the phase
group with value w iff it is in the multiplication domain and w = −i κ f.
The free Hamiltonian in position space #
The free generator statement transports across the Fourier transform to the phase group.
★★ A vector whose free orbit has a right derivative at 0 lies in dom H₀.
★★★ The interacting orbit stays in dom H₀, in the operator sense.
The free generator value at a point of dom H₀ is −i H₀ of it.
★★★ The interacting Schrödinger equation with H₀ acting on the orbit:
d/dt U(t)ψ = −i (H₀ + V)(U(t)ψ) at every t > 0, where H₀ = freePositionOp is applied to the
orbit point, which lies in its domain. The mild solution is classical in the operator sense.