The free Hamiltonian in position space is −½ d²/dx², on Schwartz functions #
Category: 1-Mathlib (CSD-free; staged for upstream).
BACKLOG #128 (out of #64(i)).
FreeHamiltonian.lean built the free Hamiltonian in the position representation,
freePositionOp, as the Fourier conjugate of multiplication by the symbol 2π²ξ², and proved it
self-adjoint with spectrum [0, ∞). The name says −½Δ; the construction says "conjugate of a
multiplier". This file makes the name honest on the Schwartz core: for every Schwartz function
φ, its L² class lies in the domain of freePositionOp, and the operator sends it to the L²
class of −½ φ''.
fourier_kineticOp: the transform of the Schwartz kinetic operatorkineticOp φ(SchrodingerSchwartz.lean, already identified with−½Δφthere) is the symbol times the transform ofφ, pointwise𝓕(H₀φ)(ξ) = 2π²ξ² · 𝓕φ(ξ)(fourier_kineticOp_apply);toLp_fourier_mem_freeSymbolOp_domain, ★freeSymbolOp_toLp_fourier: in the momentum representation the multiplier acts on(𝓕φ).toLpexactly as the transform ofH₀φ;toLp_mem_freePositionOp_domain: Schwartz functions lie in the position-space domain;- ★★
freePositionOp_toLp:freePositionOp (φ.toLp) = (kineticOp φ).toLp, hence ★★★freePositionOp_toLp_eq_laplacian:= ((−½) • Δφ).toLp, and pointwise almost everywhere ★★★freePositionOp_toLp_ae:= −½ · φ''(iteratedDeriv 2 φ, i.e.deriv (deriv φ),freePositionOp_toLp_ae_deriv). freeSymbol_eq_neg_half_mul_sqrecords the convention the row flagged: with Mathlib's𝓕 f (ξ) = ∫ e^{−2πixξ} f(x) dx, the symbol2π²ξ²is−½ (2πiξ)², which is why−½ d²/dx²has this symbol and not½ξ².
Every step moves across the transform by Mathlib's own bridges — SchwartzMap.toLp_fourier_eq,
toLp_fourierInv_eq (the L² Fourier transform agrees with the Schwartz one on Schwartz
classes) and the Schwartz Fourier inversion — so nothing about L² limits is re-proved here.
Honest scope #
- The Schwartz core only. The row asked for two things: (a) the Fourier–derivative identity on
the whole domain
𝓕⁻¹{(1 + ξ²) f ∈ L²}, and (b) the identification of that domain with the Sobolev spaceH²(ℝ). This file proves (a) on Schwartz functions and does not prove that Schwartz functions are a core forfreePositionOp(that its closure from the Schwartz restriction is the whole operator), nor (b): Mathlib has noH^sscale onLpat this pin, andH²is a definition before it is a theorem. What is proved is that the self-adjoint operator of #64(i) restricts to−½ d²/dx²on every Schwartz function, which is the sense in which an analyst reads the name. - One dimension, as in
FreeHamiltonian.lean; the Laplacian formfreePositionOp_toLp_eq_laplacianis the one that would survive toℝⁿ, and the second-derivative form is its reading onℝ(laplacian_eq_iteratedDeriv_real). - Nothing here is a dynamics statement:
PhaseGenerator.leanalready has the Schrödinger equation withfreePositionOpas generator; this file says what that generator is on smooth data.
References: Reed–Simon I §IX.7 (the free Hamiltonian on H²); FreeHamiltonian.lean
(freeSymbolOp, freePositionOp, freePositionOp_apply), SchrodingerSchwartz.lean
(kineticOp, kineticOp_eq_laplacian), PhaseGenerator.lean
(hasDerivAt_fourierGroup_freePositionOp); Mathlib SchwartzMap.toLp_fourier_eq,
SchwartzMap.laplacian_apply, InnerProductSpace.laplacian_eq_iteratedDeriv_real;
specs/BACKLOG.md #128, #64.
The convention #
The transform of the kinetic operator #
The transform of the Schwartz kinetic operator is the symbol times the transform.
Pointwise: 𝓕(H₀φ)(ξ) = 2π²ξ² · 𝓕φ(ξ).
The momentum representation on Schwartz classes #
The transform of a Schwartz function lies in the domain of the momentum-space operator.
The momentum-space operator, pointwise almost everywhere: multiplication by the symbol. Stated for a variable of the domain so that its instances at concrete elements need no unfolding.
★ The multiplier acts on a Schwartz transform as the transform of the kinetic operator, stated for any element of the domain whose coordinate is the transform's class.
★ The same at the canonical element.
The position representation on Schwartz classes #
The L² inverse transform of a Schwartz transform is the Schwartz class itself.
Schwartz functions lie in the position-space domain of the free Hamiltonian.
★★ On a Schwartz class the position-space free Hamiltonian is the kinetic operator.
★★★ The free Hamiltonian in position space is −½Δ on Schwartz functions.
★★★ Pointwise, almost everywhere: (H₀ φ)(x) = −½ φ''(x).
The same, written with two derivatives: (H₀ φ)(x) = −½ (φ')'(x).