By exploiting the gap between left and right eigenstates, the new preprint turns an abstract spectral singularity into three concrete observables that existing quantum simulators can measure without rebuilding the Hamiltonian.
Exceptional points are mathematically clean and experimentally slippery. These spectral singularities, where two or more eigenstates of a non-Hermitian system coalesce into a single defective state, sit at the center of proposals for quantum sensing, unidirectional transport, and robust lasing, yet in a many-body setting they rarely announce themselves through a single dramatic signal. A new arXiv preprint from the non-Hermitian physics community takes a pragmatic approach: it converts a defining mathematical property of non-Hermitian systems, that left and right eigenvectors are no longer simple adjoints of each other, into a layered set of observables a lab can actually measure.
The authors build their protocol on a one-dimensional complex XY spin chain that can be tuned between rotation-time (RT) symmetric and parity-time (PT) symmetric regimes. That choice is deliberate. RT and PT symmetry are the two most studied organizing principles for non-Hermitian physics, and the preprint shows they leave qualitatively different fingerprints on the same observables. Reading the two regimes together is what turns a list of measurements into a usable framework.
The first layer is global. Define an operator-level difference between the Hamiltonian and its adjoint, H − H†, and track how that quantity behaves as the system is tuned. At an exceptional point the response becomes non-analytic, a kink, cusp, or branch point that the preprint identifies as a clean location marker. This is the cheapest test, and the right place to start when you are not sure whether your chain has an EP nearby.
The second layer is local and static. The paper evaluates spin correlation functions on the right eigenstates and on the left eigenstates separately, then compares the two. In a Hermitian system the comparison would be redundant, because the two sides are conjugates. In a non-Hermitian system the gap between them is the resource: a measurable asymmetry that grows in a characteristic way as the system approaches the exceptional point. The asymmetry can be read off local observables, which matters for experiments that do not have access to the full state.
The third layer is dynamical. After a sudden quench, the time-averaged right-left entanglement entropy behaves differently in the two symmetry regimes. In the RT regime the signal peaks sharply at the exceptional point, a pronounced feature against a low baseline. In the PT regime the same quantity acts more like an order parameter, switching between finite and vanishing values as the system crosses the EP. That contrast is the conceptual payoff. It is also where the preprint is most useful to experimentalists, because the protocol requires only time-averaged post-quench data, not the kind of full state tomography that would force a reconstruction of H.
The honest limitation is part of the story. Static bipartite entanglement measures, the entanglement entropy you would compute from a single right or left eigenstate, do not by themselves capture the left-right distinction. A researcher who reaches for the standard static tool first will miss the EP. The preprint treats that as a finding rather than a footnote: it is the reason the dynamical probe has to be in the toolkit at all.
The protocol is also a practical one. Groups running trapped-ion, superconducting, or Rydberg-atom simulators can implement the three layers on existing hardware. The global H − H† test is a post-selection on measured observables. The static correlation asymmetry needs only local correlators evaluated on the two eigenstate branches. The dynamical layer needs a quench and time-averaged entropy estimates, which are routine in current quantum simulation platforms. None of the steps require rebuilding the Hamiltonian from scratch, which is what has kept EP physics mostly theoretical until now.
The interesting open question is transfer. The paper works in a clean, exactly solvable spin chain. Real hardware will have decoherence, finite size, and imperfect non-Hermitian engineering, all of which smear the sharp signatures the preprint identifies. Whether the three-layer protocol remains a reliable diagnostic in that messier setting is the next test. For now, the preprint gives experimentalists a checklist that is both principled and implementable: start with the global non-analyticity, confirm with static left-right correlation asymmetry, and use the quench-driven entanglement signal to distinguish an RT-style peak from a PT-style order-parameter switch. The PT-versus-RT contrast is where the framework earns its keep, because it tells you not just that an exceptional point is there, but which symmetry regime you are in.