Work

Neutron multiplicity separates yield-optimal from information-optimal reactions in superheavy element synthesis

Year: 2026
Type: preprint
Abstract: Synthesising a superheavy nucleus costs weeks to months of accelerator beam time per attempt, so choosing which reaction to run is an allocation problem whose answer depends on which objective is being served. This work scores 20,724 projectile-target-channel combinations with a gradient-boosted correction to a three-stage fusion-evaporation baseline fitted to 92,929 measured cross sections, removes 7,319 of them on target availability, target half-life and beam intensity, and solves the same binary knapsack twice under a twelve-unit budget: once for expected production rate, once for prediction disagreement across four shell-correction scalings weighted by detectability. Of that disagreement, 88 per cent is the boosted model's response to the scaled feature, so the measure is not calibrated mass-model uncertainty. The two selections share no reaction. Four reactions maximise production and six maximise disagreement, and both spend the budget exactly. Only 1,295 of the feasible channels lie above the 10 fb detectability floor, and on those the two objectives are close to independent, at a Kendall tau of +0.03 between predicted cross section and disagreement. Neutron multiplicity orients them. Across the feasible set mean disagreement rises from 0.34 to 1.03 decades between the 1n and 6n channels while the predicted cross section falls, and both gradients survive above the floor in weaker form, so the two objectives sit at opposite ends of one axis. Zero overlap therefore follows partly from how the disagreement measure is built, and is reported here as a mechanism and not as independent evidence. An ablation isolating two shell-model features computed through a Jordan-Wigner-mapped pairing Hamiltonian returns a null.
Language: English
Open Access status: green
OpenAlex ID: W7225921120
Cites:
Cited by:
Related to: 10