arXiv:2610.06661
Kramers' Theorem Leaves a Topological Footprint in Classical Records
Chon-Fai Kam, Kai-Wen Wong
Chon-Fai KamAbstract
Time reversal $Θ$ squares to $-1$ on a half-integer spin, so every state is orthogonal to its time reverse. This sign has been read only from quantum objects. The equilibrium state loses it, but the classical record keeps it. For a detector that resolves handedness, the probabilities that one stretch of record is followed by the time reverse of another measure the cohomology class of time reversal, $Θ^2=\pm1$. Their matrix carries this sign in its signature, a topological integer that survives equal emission and absorption losses. A programmed emitter on quantum processors reads it in about $10^3$ shots.
AI Summary
Can the sign $\Theta^2=\pm1$ be inferred from a monitored system’s classical measurement record? The paper considers stationary quantum instruments satisfying detailed balance with physical time reversal and constructs the reversed-future kernel $\mathcal K(x,y)=P(x\bar y)$, where $\bar y$ reverses the outcome order and exchanges emission with absorption. An overlap identity relates this kernel to conditional quantum states. For $\Theta^2=-1$, any stretch preparing a pure state cannot be followed by its own time reverse. More generally, if conditional states span all Hermitian operators on a $d$-dimensional time-reversal-invariant block, Sylvester’s law gives $\operatorname{rank}\mathcal K=d^2$ and signature $n_+-n_-=\Theta^2d$. A fully reached doublet therefore has inertia $(1,3)$ for Kramers time reversal versus $(3,1)$ for $\Theta^2=+1$. A microreversible classical realization requires at least six states for the Kramers-doublet record. Thermal spin models illustrate the construction: handedness-resolving detection can expose the signature, whereas linear-polarization detection leaves doublets unresolved. Simulated programmable qubit instruments distinguish the two doublet classes using 1,000 shots of 30 steps, including specified noise. Thermal emitters require approximately $10^6$–$10^{10}$ independent windows even with perfect detection. Inference requires reversal symmetry and sufficient state-space reachability; an inaccessible hidden spin can change a model’s $\Theta^2$ without changing its record. Symmetric detection losses preserve inertia but weaken eigenvalues; unequal emission/absorption efficiencies require testing and correction. No physical-emitter experiment is reported. Classical memory bounds apply only to microreversible models, and extension to other symmetry cohomology classes remains open.
Central claim may not hold: Handedness-resolving detection does not by itself reveal Θ²: the paper's circularly detected twofold-axis example has inertia (1,1), so extracting the sign additionally requires suitable operator-space accessibility. The probabilities determine a kernel invariant rather than the time-reversal class of an arbitrary emitter, because an inaccessible hidden spin can reverse Θ² without changing any probabilities. The abstract suggests an achieved quantum-processor readout, whereas the reported 1000-shot results are simulations of a proposed programmed emitter.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 8 pages, 3 figures, including 2 pages of End Matter