Quantum decay reversal is limited by an environment’s memory, preprint argues

Different elapsed times, combined coherently

Proper time is the time accumulated along a system’s own path. Motion and gravitational potential can make it differ between paths. The preprint considers a quantum system placed in a coherent superposition of paths with different clock rates. In each branch, the internal physics remains the same; only the duration of evolution changes.

The researchers then apply post-selection: they retain only runs with a particular final measurement outcome for the variable controlling those paths. Because this outcome does not identify an individual branch, the forward-evolved states can interfere.

Under restricted conditions, their combination approximates evolution for a duration outside the range experienced by any branch—even a negative duration. The relevant mathematical effect is superoscillation: a combination of bounded-frequency components can locally vary faster than any component. Here it appears in the energy-dependent phases governing time evolution.

The effective clock rate is a weak value, a quantity defined by both the initial preparation and the selected final outcome. It need not lie between the actual branch rates. But a negative weak value alone does not guarantee accurate backward evolution: the approximation must work across the system’s relevant energy distribution.

What would count as reversal?

The proposed test system is a single sealed cell containing excited atoms, their emitted radiation and absorbing walls. Everything must remain part of the same quantum object and be recombined together. Escaping radiation or heat could record which branch occurred, undermining the interference.

Freshly excited atoms are insufficient to establish reversal. Starting from maximum excitation, their survival probability decreases for evolution in either time direction. Instead, the authors propose allowing emission to begin before the cell enters the clock-rate superposition.

The proposed signature is then an excited fraction, among successfully post-selected runs, higher than the fraction at insertion. Within the decay regime considered, every ordinary forward branch lowers that fraction. An increase would distinguish the selected outcome from a statistical mixture of those branches.

The readout uses projective measurements, which classify atoms as excited or unexcited. The signature lies in the frequencies of these outcomes across selected runs, not in an individual atom displaying an unusual energy value.

An exact calculation exposes the limit

The authors express the conditional excited fraction exactly through the survival amplitude: the complex overlap between an initially excited atom with no emitted photon and its evolved state.

Their expression includes both the amplitudes contributing to excitation and the normalization of the selected state. That normalization matters because the cell’s internal contents evolve differently in different branches and therefore carry information about the branch.

For two branches, post-selection amplifies those small differences alongside the desired effect. The resulting limitation depends on the full energy spread of the atom and radiation field, not merely the linewidth associated with the usual decay rate.

The authors evaluate a model with an environmental bandwidth—a range of frequencies participating in the emission—and a corresponding memory time, approximately its inverse. Before this time, excitation initially falls quadratically and the atom remains coherently connected to the developing radiation. Later, the model approaches Markovian decay: approximately exponential evolution with little dependence on earlier history.

In their calculation, with the run lasting one-tenth of the preceding decay interval, the increased-excitation signature survives until that interval reaches about 1.2 environmental memory times. It fails beyond that range. For free-space optical emission, the paper estimates a memory time around a femtosecond. Thus, the two-branch proposal reverses only the coherent onset of emission, not an established irreversible decay.

Deeper reversal has a much higher cost

The paper argues that longer reversal requires many branches, a strictly bounded relevant energy spectrum and a success probability decreasing exponentially with the number of environmental oscillations being reversed. Its estimate for undoing a nanosecond of optical decay is roughly one million branches.

The physical implementation is also unresolved. Suggested controls include gravitational potentials, a rotating platform and different inertial velocities. Maintaining coherence of the complete cell, controlling branch-dependent disturbances and stabilizing phases remain demanding requirements. The authors explicitly present the scheme as a thought experiment.

Post-selection does not provide controllable travel into the past. The paper also allows an interpretation entirely in terms of interference between forward-evolved states. Averaging over all selection outcomes removes the conditional anomaly, so the proposal does not overturn ordinary ensemble decay or the second law. Its central conclusion is narrower: within this construction, environmental bandwidth limits how faithfully rare selected outcomes can reproduce earlier states.