Maximal norm discontinuity for a bounded-region continuum fermion interaction
A quantum observable can change continuously on every fixed vector and in every fixed normal state and still fail to change continuously in operator norm. This note gives one explicit continuum-fermion interaction where states with more and more particles witness the largest possible norm change, at times tending to zero.
Open the original PDF (5 pages) · Download PDF · Original files · BibTeX · DOI: 10.5281/zenodo.23383930
The contribution and earlier work
O. Siebert's earlier Discontinuity of Continuum Fermion Dynamics supplies the high-density packing and localized phase mechanism. Its Proposition 2.6 already gives the maximal norm limit two for a full-space constant-number reference. The contribution here is an explicit free-Slater comparison that transfers that limit to one fixed, bounded spatial interaction cutoff.
The earlier localized-phase argument also adapts to this bounded symmetric cutoff kernel. That cutoff adaptation is a deduction explained in the note, rather than a literal cutoff theorem stated in Siebert's preprint. The note makes no claim of first priority or that the question remained open when it was prepared.
The target is the existential fermionic question on printed page 371 of Oberwolfach Report 7/2025, read with the distinction between pointwise operator norm and strong-operator continuity on printed page 365. The preceding fixed-density Hartree discussion concerns a separate bosonic model.
Why the two kinds of continuity differ
Operator norm asks for a bound that works uniformly over all unit vectors. Here the potential, interaction region and observable stay fixed, but the vectors used to test the change depend on the particle number N. The particles are packed into a fixed ball, so the interaction phase can change by half a turn on a time scale of order 1/N.
A smooth cutoff estimate shows that the freely evolved packed states leak out of the ball slowly enough on this time scale. A comparison with an exactly computable number-dependent phase then gives the maximal norm change. This comparison uses free Slater states; it does not assume that interacting evolution preserves Slater determinants.
Every fixed vector still sees a change tending to zero as time tends to zero. For every fixed normal state ρ, the quadratic expectation Tr(ρDt*Dt) also tends to zero. The large values below use a changing sequence of normal states.
The precise result
The one-particle space is L²(ℝ³), with kinetic operator h=−Δ on all of space. The fermionic Fock space is the direct sum of the antisymmetric n-particle sectors. Fix Λ=B(0,4), let Ln=∑j=1n1Λ(xj), and take one fixed real, radial, smooth, compactly supported potential 0≤V≤1 that equals one on Λ−Λ.
For κ∈{0,1}, use the sector Hamiltonians and their self-adjoint maximal direct sum H(κ):
Hn(κ) = Tn + Ln(Ln−κ), Tn = ∑j=1n(−Δxⱼ).
Here κ=0 includes diagonal density pairs, while κ=1 gives normal ordering. Both use coefficient one in the ordered-pair sum, without a factor ½. The spatial cutoff is sharp and the fields have no ultraviolet smearing.
There is one fixed normalized f∈Cc∞(B(0,½)) and explicit unit Slaters ΨN and ΞN=a*(f)ΨN, for N≥1, such that, with
tN(κ) = π/(2N+1−κ), U(κ)(t) = eitH⁽κ⁾,
DN(κ) = U(κ)(tN(κ))a(f)U(κ)(tN(κ))* − a(f).
The following limits hold as N→∞:
‖DN(κ)ΞN + 2ΨN‖ → 0,
‖DN(κ)ΞN‖ → 2, ‖DN(κ)‖ → 2.
The creation-operator difference has the same operator-norm limit. For the changing normal vector states ωN(B)=⟨ΞN,BΞN⟩,
ωN((DN(κ))*DN(κ)) → 4.
Since ‖a(f)‖=1, the triangle inequality bounds every such operator difference by two. The limiting value is therefore maximal.
An explicit fixed choice
Let b(s)=e−1/s for s>0 and b(s)=0 otherwise. The potential and mode may be chosen as
V(x) = b(81−|x|²) / (b(81−|x|²)+b(|x|²−64)),
f(x) = cfb(1−9|x|²),
where cf makes f have norm one. Thus V equals one up to radius eight and vanishes outside radius nine; f is supported within radius one third. These choices do not change with N.
For the packed orbitals, take Q=2e₁+(−¼,¼)³ and a normalized smooth function
φ(x) = cφ∏r=13b(1−16(xr−½)²).
Subdivide Q into mN³ cubes, with mN=⌈N1/3⌉ and side length ℓN=1/(2mN). If cj,N are the first N lower corners in lexicographic order, set
ψj,N(x) = ℓN−3/2φ((x−cj,N)/ℓN),
ΨN = ψ1,N∧⋯∧ψN,N, ΞN = a*(f)ΨN.
Disjoint supports give orthonormal orbitals, separated from f, and normalized Slater determinants. The proof compares the N- and (N+1)-particle evolutions with the constant-number reference Tn+n(n−κ). At the specified times, both comparison errors are O(N−1/6+N−1/3), while the relative reference phase is exactly −1.
Scope and limitations
This is an existence result for the displayed potential and ball in dimension three. The witnesses have increasing particle number in a fixed region, with no fixed-density or bounded-energy restriction. The result does not cover every potential or region, settle dimensions one and two, or construct a thermodynamic limit.
The comparison takes place among bounded operators on Fock space. It establishes no invariance or non-invariance theorem for the CAR algebra. Siebert's earlier extended-algebra construction, with norm continuity after time averaging, concerns a different class of observables and is compatible with this unaveraged fixed-mode result.
Original manuscript abstract
For continuum fermions in three dimensions, we give one fixed smooth compactly supported pair potential and one fixed bounded interaction region for which the Heisenberg orbit of a fixed annihilation operator is discontinuous in operator norm. Explicit finite-particle Slater states and times tending to zero witness norm differences tending to the maximal value 2. The construction works both with diagonal density pairs and with normal ordering, without ultraviolet smearing. A fixed-cutoff commutator estimate, a Slater second-moment identity, and a two-sector comparison supply a self-contained proof. Continuity on every fixed vector and in every fixed normal state remains valid.
Original files and citation
The unchanged native files appear in the original Zenodo order.
- Manuscript PDF — bounded-region-fermions-v1.0.pdf · Zenodo original
- Source and verification — bounded-region-fermions-source-verification-v1.0.zip · Zenodo original
Kriebel, Alec (14 October 2026). Maximal norm discontinuity for a bounded-region continuum fermion interaction (Version 1.0) [Preprint]. Zenodo. 10.5281/zenodo.23383930. Download BibTeX · Native record.
The manuscript and accompanying native archive are provided under CC BY 4.0. Referenced third-party works retain their own rights.
Main references
- Mini-Workshop: New Directions in Correlated Quantum Systems, Oberwolfach Report 7/2025, DOI: 10.4171/owr/2025/7. Printed page 365 distinguishes the norm and strong topologies; printed page 371 states the bounded-region fermionic question.
- O. Siebert, Discontinuity of Continuum Fermion Dynamics, arXiv:2609.34402v1, submitted 28 September 2026. Earlier packing/localized-phase mechanism; Proposition 2.6 gives maximal two for the full-space constant-number reference.
- O. Siebert, On the thermodynamic limit of interacting fermions in the continuum, arXiv:2409.10495v2, 24 February 2025; Letters in Mathematical Physics 115, 33 (2025), DOI: 10.1007/s11005-025-01919-y. Extended-algebra and time-averaging construction.
Read the manuscript
Open PDF in a full window · Download PDF · Original PDF on Zenodo
Research status and contributions
AI tools were used extensively in developing the research, proof, checks and writing, including adversarial reviews by separate AI agents. Alec Kriebel is responsible for the claims and artifacts. This is an unrefereed preprint; no independent human peer review, journal acceptance or formal proof certificate is claimed. Finite computational checks supplement the written proof, and numerical diagnostics are not rigorous interval enclosures.
Researchers interested in independent verification, substantial mathematical extensions, or developing a revised manuscript for journal submission are welcome to contribute. Authorship and acknowledgments depend on the substantive work actually contributed; endorsement alone does not confer authorship.