Observation architecture and the identifiability of temporal irreversibility

Temporal irreversibility in a multichannel recording reflects both latent dynamics and the architecture of observation. Here we establish the exact identifiability boundaries separating these two contributions in stationary Gaussian processes. For every continuously differentiable, positive-definite spectral matrix of a real stationary Gaussian process, we construct an exactly equivalent observation of a time-reversible latent process of twice the dimension using a fixed real mixing matrix and a one-sample differential delay, preserving finite covariance support for finite-order target spectra. Overcoming this observational equivalence requires either physical resource constraints or structural restrictions on the measurement operator. Quantifying the resource cost in a rotational spectral family, we derive the exact minimum hidden spectral-trace cost at a fixed frequency and its broadband optimum, distinguishing an inverse-square divergence at short delays for fixed asymmetry from a finite limit along natural low-frequency paths. Conversely, when independent temporal filters act on already defined channels, closed cross-spectral cycles preserve a filter-invariant component of time-reversal asymmetry, which we bound from below via an asymptotically sharp triangle inequality on a phase-relaxed reversal divergence. Bridging population invariants and finite recordings, we demonstrate how spectral tapering interacts with differential group delay to induce systematic T⁻² closure bias, and we construct simultaneous finite-sample confidence lower bounds—including a self-consistent spectral calibration that operates within a specified covariance-tail envelope without prior knowledge of the normalized spectral shape. Controlled coordinate transformations and spatial rereferencing in human electroencephalography confirm that sensor-level spectral-cycle statistics depend critically on the chosen observation basis. Together, these results identify the structural observation assumptions and quantitative calibration resources required to connect an observable arrow of time to latent nonequilibrium dynamics.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24306356
Primary Topic
Advanced Thermodynamics and Statistical Mechanics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Observation architecture and the identifiability of temporal irreversibility

Yanzhe Zhao
Zenodo (CERN European Organization for Nuclear Research)
Advanced Thermodynamics and Statistical Mechanics
preprint

Observation architecture and the identifiability of temporal irreversibility

Yanzhe Zhao
preprint en

Abstract

Temporal irreversibility in a multichannel recording reflects both latent dynamics and the architecture of observation. Here we establish the exact identifiability boundaries separating these two contributions in stationary Gaussian processes. For every continuously differentiable, positive-definite spectral matrix of a real stationary Gaussian process, we construct an exactly equivalent observation of a time-reversible latent process of twice the dimension using a fixed real mixing matrix and a one-sample differential delay, preserving finite covariance support for finite-order target spectra. Overcoming this observational equivalence requires either physical resource constraints or structural restrictions on the measurement operator. Quantifying the resource cost in a rotational spectral family, we derive the exact minimum hidden spectral-trace cost at a fixed frequency and its broadband optimum, distinguishing an inverse-square divergence at short delays for fixed asymmetry from a finite limit along natural low-frequency paths. Conversely, when independent temporal filters act on already defined channels, closed cross-spectral cycles preserve a filter-invariant component of time-reversal asymmetry, which we bound from below via an asymptotically sharp triangle inequality on a phase-relaxed reversal divergence. Bridging population invariants and finite recordings, we demonstrate how spectral tapering interacts with differential group delay to induce systematic T⁻² closure bias, and we construct simultaneous finite-sample confidence lower bounds—including a self-consistent spectral calibration that operates within a specified covariance-tail envelope without prior knowledge of the normalized spectral shape. Controlled coordinate transformations and spatial rereferencing in human electroencephalography confirm that sensor-level spectral-cycle statistics depend critically on the chosen observation basis. Together, these results identify the structural observation assumptions and quantitative calibration resources required to connect an observable arrow of time to latent nonequilibrium dynamics.

Zenodo (CERN European Organization for Nuclear Research)
Tianjin University (CN)
Advanced Thermodynamics and Statistical Mechanics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon