Forward generation does not ensure GNS descent: an explicit continuous counterexample
Read the paper (PDF) · Download PDF · Supporting ZIP · Zenodo record
Plain-English explanation · Read on this page · Collaboration
Abstract
For a general conditional-expectation dilation of a Markov semigroup, generation by the observed algebra and its forward translates does not ensure descent through the expectation's fixed GNS representation. We give a unital separable, injective, point-norm continuous example for the two-state exponential relaxation semigroup. Its GNS image is M2(ℂ), and at every positive time an explicit nonzero positive element leaves the representation kernel. This refutes the concrete generation-only proposal in Skeide's 2009 module-approach problem. The clock/cocycle method is established: the negative phenomenon is also a short corollary of an earlier clock dilation of Burgarth, Facchi and Hillier. The note makes the connection explicit and supplies a small matrix realization with exact witnesses; it makes no claim of literature-first priority.
The idea in plain English
A representation can simplify a mathematical system by treating some quantities as zero. For time evolution to pass through that representation, those quantities must still count as zero after time has passed.
This example starts with a two-state relaxation process. A larger system is built entirely from the observed quantities and their future translates, yet information erased by its fixed GNS representation becomes visible later. For every positive time, the paper gives an explicit nonzero positive element that starts in the full representation kernel and leaves it at that time.
The conclusion concerns this fixed representation. It shows that forward generation alone does not guarantee the proposed passage to represented dynamics.
The precise result
Let D be the diagonal algebra in M2(ℂ), identified with ℂ2. The paper constructs a unital separable C*-algebra A, a point-norm continuous semigroup (θt)t≥0 of injective unital *-endomorphisms, a unital embedding i:D→A, and p:A→D such that ip is a conditional expectation and
A = C*(θr(i(D)): r ≥ 0), Tt = pθti,
Pt = ½ [[1+e−t, 1−e−t], [1−e−t, 1+e−t]].
For the left representation π of the fixed expectation-GNS correspondence E, π(A) is isomorphic to M2(ℂ). For every s>0, there is an explicit nonzero positive as in the full representation kernel ker π for which θs(as) is outside ker π. Hence no dynamics on π(A), or on all adjointable operators Ba(E), can intertwine π with θ in the proposed way.
Why the result holds
- Use a continuous matrix-valued clock and translated unitary cocycle, then restrict to the algebra generated by the observed diagonal algebra and its forward translates.
- Identify the actual GNS correspondence explicitly: its left representation is evaluation at zero followed by matrix left multiplication.
- For each positive time, construct the positive kernel element above and calculate its nonzero represented image after evolution.
Scope and limitations
The counterexample targets the concrete generation-only proposal for a general conditional-expectation C*-dilation in Skeide’s 2009 problem. It assumes pθti=Tt; the stronger equation pθt=Ttp on all of A is not assumed and fails in the example.
The expectation is nonfaithful. The example is neither a strong dilation nor a weak/corner dilation. No normal von Neumann version, no nonexistence of a different dilation of the same Markov semigroup, and no failure of every stronger or incomparable minimality condition is asserted. Extension after kernel invariance has been separately imposed remains outside the result. The kernel at issue is the full GNS representation kernel, not a general identification with one vector-state null space.
Background and prior work
The clock, evaluation, and translated-cocycle mechanism is established prior work of Burgarth, Facchi, and Hillier. After the endpoint-continuity and forward-generated-restriction checks supplied in this note, their earlier amplitude-damping clock also implies failure of kernel invariance at some time as a short corollary. This note makes that connection explicit and gives a small M2/ℂ2 example with a nonzero positive witness for every positive time. It claims neither a new general dilation method nor literature-first priority.
- Source of the concrete forward-generation-only descent proposal. Michael Skeide, Some Problems, in Mini-Workshop: Product Systems and Independence in Quantum Dynamics
- Established clock/evaluation/cocycle mechanism; the earlier explicit clock gives the some-time negative consequence after the restrictions and continuity checks explained in the note. Daniel Burgarth, Paolo Facchi, and Robin Hillier, Control of quantum noise: on the role of dilations
- Finite time-dependent unitary dilation and initial square-root behavior. Benjamin Dive, Florian Mintert, and Daniel Burgarth, Quantum simulations of dissipative dynamics: Time dependence instead of size
- Related bounded-total-Hamiltonian first-derivative obstruction. Frederik vom Ende, Quantum-Dynamical Semigroups and the Church of the Larger Hilbert Space
- Related fixed-action extension question, with the same scope limits. Michael Skeide and K. Sumesh, CP-H-extendable maps between Hilbert modules and CPH-semigroups
- Distinctions among general, strong, and weak/corner dilation categories; cited locators refer to arXiv v3. Orr Moshe Shalit and Michael Skeide, CP-Semigroups and Dilations, Subproduct Systems and Superproduct Systems: The Multi-Parameter Case and Beyond
Status and AI use
Unrefereed preprint. AI tools assisted in developing, writing and checking the work. Independent human review is welcome.
Read the paper
Open the published PDF · Download PDF · Original PDF on Zenodo
Supporting materials
Download gns_verification.zip · Original supporting ZIP on Zenodo
Collaborate on this result
Useful contributions include independent checking of the fixed GNS identification and every-time positivity argument, comparison with earlier dilation constructions, or substantial work on a revised manuscript for journal submission.
Authorship would reflect substantive contributions to the work.
Cite this paper
Kriebel, Alec (2026). Forward generation does not ensure GNS descent: an explicit continuous counterexample. Version 1.0. Zenodo. 10.5281/zenodo.23356540.
Licensing
The deposited work is available under Creative Commons Attribution 4.0.