A Review of Possible Self-Replication Mechanisms in Cosmology and Their Test against DESI, Planck and Supernova Data

In brief. Of about thirty variants of "cosmic copying", only one works and makes a testable prediction of its own — the "epidemic" production of dark matter; in its sterile-neutrino version it predicts a hard-X-ray line that can be searched for. Dark-energy copying is not established: the rules for its DESI DR3 test were fixed in advance, but the decision will most likely come only from DESI-II and Euclid; our model V5 is refuted.

Self-reproduction is an observed natural phenomenon: chain reactions, lasers and life all work this way. The Earth is part of the Universe, so it is legitimate to ask whether such growth occurs on cosmic scales — and to test it with data. This paper is part of the SRDP (Self-Replication of Dark Particles) programme. By self-replication we mean autocatalytic growth fed by a limited resource: the more "copies", the faster new ones appear. We tested about thirty such variants by calculation, and the best against Planck PR4, DESI DR2, supernova and structure-growth data. The main filter is the energy budget: a component's density can grow only if its pressure is below minus its density, if it takes energy from another component, or if gravity is modified.

What works is the "epidemic" production of dark matter, χ + X → χ + χ (Bringmann et al., 2021). For a sterile neutrino (a hypothetical heavy relative of ordinary neutrinos), the main paper's calculation, which includes the cooling of the dark particles after production, gives a mass window of ≈ 25–140 keV (≈ 45–140 keV with lenses and satellite galaxies; version 2: 30–150 keV). The model's own prediction is a mandatory decay line at half the mass, no fainter than a calculated minimum (weaker mixing would need a coupling stronger than galaxy clusters allow); together with the INTEGRAL/SPI limits it sets the window's upper edge.

Our model V5 (dark matter partly turning into dark energy) is not supported: ln B ≈ −3.4 against ΛCDM (the data favour the standard model by a factor of about 30), and a sharp transition gives a strong late-time integrated Sachs–Wolfe effect (ISW; Δχ² ≈ +80 to +110). The main-paper variants "carrying capacity of the medium" and "fading replication" fit DESI DR2 together with CMB and supernovae better than Λ (Δχ² down to −11.2, on a par with w₀w_a) but are Bayesian-indistinguishable from Λ; their DESI DR3 test was preregistered before the data release (OSF, DOI 10.17605/OSF.IO/W328H); by our power calculation DR3 will most likely not separate them from Λ — DESI-II and Euclid will mainly decide. Status: the dark-matter mechanism works; dark-energy self-replication is not established; V5 is refuted.

We also distinguish four ways of copying and present the author's ideas on the act of copying itself (pinch, pulsation, circulation, densification) as starting points for hypotheses, not as results.