Quantum State Geometry as a Non-Markovian Ledger: How Much Past Does a Particle Remember?
Barandes has reformulated quantum theory as indivisible stochastic processes: definite configurations evolving by stochastic laws that generically do not factorize over time. Hilbert-space ingredients -- wave function, density matrix, Hamiltonian -- are then shown to be gauge-dependent bookkeeping. What that program lacks is a quantitative, observer-relative measure of retained past. We supply it from Quantum State Geometry (QSG). Memory is retained distinguishability: for an observer $O=(\Sigma_O,\Delta_O,\sigma_O)$, $N=S/V_0$ counts distinguishable micro-histories in the resolution ball $B_{\Delta_O}$, $S_O=k_B\ln N$ is relational entropy, and $d\tau_O=dS_O/\sigma_O$ is relational time. Forgetting is Bures contraction under a completely positive trace-preserving (CPTP) map. A division event is not postulated: it is the operational point at which contraction has washed out path-dependence to within resolution $\Delta_O$. We show three consequences. (i) The Markovian limit is recovered when the future depends only on the present $\rho$, not on the path that produced $\rho$. Non-Markovianity is residual path-dependence, quantified by fidelity deficit between same-present, different-history preparations. (ii) Causal locality in QSG is a horizon cutoff on $B_{\Delta_O}$: histories outside the light cone contribute no distinguishable $N$. This gives a QSG route to Tsirelson bound $\langle \mathrm{score}\rangle \le 1/\sqrt{2}$, complementary to Barandes--Hasan--Kagan. (iii) The Janus point, $N=1$, $S_O=0$, is a zero-memory boundary. The arrow of time is $dS_O>0$ away from Janus in both directions. We propose a falsifiable discriminator: prepare two ensembles with the same present $\rho(t_1)$ but different histories, then compare their next-step statistics. Orthodox Markovian QM predicts no difference. QSG predicts a difference bounded by retained memory $\Delta S_O$. Chiral Quantum State Geometry (CQSG) extends the ledger to signed memory $N(A)=|A_R|^2-|A_L|^2$, where memory can cancel. *Author DisclaimerAll underlying concepts, frameworks, and core research directions are the original work of the author. Generative AI tools were used strictly for assistance in manuscript drafting, code generation, and figure preparation. All theoretical formulations and mathematical proofs included herein are unverified AI-generated drafts shared as preliminary proposals for discussion and community review.
Authors
- Peter HF Lau
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24269931
- Primary Topic
- Quantum Mechanics and Applications
- Type
- preprint