A theoretical route to detecting molecular entanglement with spectroscopy

Their study, submitted to arXiv on November 7, 2026, is a theoretical preprint. The supplied record does not establish peer review. It develops a framework and illustrates it with molecular models, rather than reporting a new laboratory demonstration of entanglement.

Testing explanations, not appearances

Entanglement is defined relative to a chosen division of a system: between two molecules, for example, or between two localized electronic orbitals. A separable, or non-entangled, state can be expressed as a statistical mixture of product states of those parts.

Showing that an entangled model reproduces a spectrum is not enough to establish entanglement. A separable state might reproduce the same spectrum. The authors instead ask whether any admissible separable state can explain all the retained measurements simultaneously.

Their starting point is a model of how molecular evolution and detection turn an unknown input state into an optical signal. Fixed linear operations on that signal—such as integrating selected frequency regions—produce a short list of coefficients. Each coefficient corresponds to the average value of an effective observable determined by the response model and measurement protocol, independently of the unknown state.

The researchers calculate the region of coefficient combinations available to all admissible states and the smaller region available to separable states. A physically consistent signal outside the separable region certifies entanglement under the model assumptions. A signal inside it is inconclusive: entangled and separable states can produce identical retained information.

This avoids reconstructing the full quantum state, although calculating separable bounds can still be demanding for larger systems.

Two spectral regions can outperform either alone

One illustration considers two coupled light-absorbing molecular units, or chromophores. The input state is assumed to occupy only the ground and single-excitation sectors, with no population in the doubly excited state. The probe response nevertheless includes transitions into that doubly excited level.

The authors extract two coefficients by weighting different regions of a calculated pump–probe spectrum. For an example containing an 80% ground-state contribution and a 20% contribution from the lower exciton state, the coefficient pair lies outside the separable region. Neither coefficient alone establishes entanglement; their combination does.

An analogous mixture involving the upper exciton is entangled but escapes detection. In this model, a separable state reproduces its entire absorptive spectrum, not merely the two selected integrals. Choosing different frequency windows therefore cannot recover information that the spectrum itself lacks.

These calculations demonstrate a measurement principle, not evidence that entanglement improves photosynthetic energy transfer. The model also omits population relaxation and static disorder.

Fluorescence supplies complementary constraints

A second example uses a model motivated by experiments on coupled dibenzoterrylene molecules. Here, both molecules may be excited simultaneously.

After an idealized instantaneous removal of the preparation laser, the initial fluorescence intensity, its initial decay slope and unnormalized photon coincidences determine three quantities: mean excitation number, a component of coherence between the molecules, and double-excitation probability.

Together, these quantities certify entanglement for two modeled preparations at the single-excitation resonances. A central preparation remains inconclusive despite being entangled in the calculation. Omitting the coincidence-derived double-excitation information also makes both resonance tests inconclusive.

The authors explicitly state that this illustration uses ideal signals and does not certify entanglement from the previously published experimental data.

Two further examples examine polarization-dependent two-photon signals for orbital and vibrational entanglement. They show that averaging over molecular orientations need not erase all useful information. However, two fitted coefficients do not necessarily provide two independent constraints: in some geometries they are proportional.

Calibration determines the strength of the claim

The framework is model-dependent. The response model must describe every admissible comparison state, not just fit the proposed entangled target. Assumptions restricting which molecular levels are occupied must also be justified.

Measurements combined in one test must refer to the same prepared state. Scanning the preparation laser frequency generally changes that state, so those measurements cannot simply be pooled as if they described one input.

Orientation-dependent preparation and static disorder can require an explicitly ensemble-resolved treatment rather than a single averaged state. Detector timing, backgrounds, efficiencies and finite laser switch-off can likewise affect the fluorescence test.

Ultimately, certification requires the experimental confidence region to exclude every separable explanation allowed by calibration and model uncertainties. The paper provides a theoretical route to that comparison—not yet an experimental certificate.