Can COSI detect gamma-ray lines from rare isotopes produced in the astrophysical intermediate neutron-capture process?

The paper investigates whether the COSI γ-ray telescope can detect γ-ray line emission from rare isotopes produced in the astrophysical intermediate neutron-capture process (i process). The i process is a neutron-capture nucleosynthesis regime intermediate in neutron density between the slower s process and the more rapid r process, occurring when convective mixing during a thermo-nuclear runaway leads to proton ingestion, producing neutron densities of Nn ∼ 10 13–10 15 cm-3.

The astrophysical sites of the i process remain uncertain, but two candidates with predicted rapid mass ejections at metallicities of stars in the solar neighborhood are post-asymptotic giant branch (post-AGB) stars, such as Sakurai’s object (V4334 Sagittarii), and rapidly-accreting white dwarfs (RAWDs). Detailed 1D and 3D simulations indicate that the convective-reactive astrophysical fluid dynamics responsible for i-process nucleosynthesis can lead to violent, non-radial outbursts that ultimately result in mass ejections of i-process products.

The authors calculate the ejected yields of rare isotopes whose radioactive decays may produce detectable γ-ray lines, particularly in the 0.5–2 MeV energy range. They focus on isotopes such as 22Na, 89Sr, and 95Zr, which are expected to generate long-lasting emissions potentially observable by the COSI γ-ray telescope. They estimate the formation rates of these sources and the likelihood of detecting their γ-ray emissions within 1000 parsecs of the Sun.

Key findings include:

• The probability of observing i-process emission lines during COSI’s operational period is up to ≈ 1%, but could rise to 11% for 89Sr if the event is observed within a few days.

• Due to the long lifetime and large production of 22Na from proton-capture reactions, its detection is more likely, with a probability of ≈ 5%.

• Future space missions could significantly enhance detection capabilities, potentially increasing the observation probability to several tens of percent.

• Detection of long-lived neutron-rich isotopes such as 137Cs would provide the first direct γ-ray signature of intermediate neutron-density nucleosynthesis, distinguishing the i process from classical s- and r-process pathways.

The paper identifies six unstable isotopes ejected by the models whose γ-ray line fluxes could be detectable by COSI: 22Na (1.275 MeV), 86Rb (1.077 MeV), 89Sr (0.909 MeV), 95Zr (0.757 MeV), 103Ru (0.497 MeV), and 123Sn (1.089 MeV). Additionally, 137Cs (0.662 MeV) is highlighted for its long lifetime and potential detection.

The post-AGB VLTP and RAWD models produce strong 22Na 1.275 MeV γ-line emissions detectable over several years. In all selected models, both 89Sr and 95Zr produce strong γ-line fluxes detectable for up to two years after ejection. In most models, the 0.497 MeV 103Ru and 1.077 MeV 86Rb lines, and in some the 1.089 MeV 123Sn line, can be detected over periods of 0.25 to 1.5 years.

The predicted 137Cs production in RAWD model D reflects a larger neutron exposure, and a confirmed γ-ray detection would provide a direct, time-integrated probe of nucleosynthesis in this intermediate neutron-density regime.

From the RAWD formation rate of binary population synthesis models and the He-shell flash recurrence period, the authors estimate that the Milky Way hosts approximately 1000 RAWDs, about 9 of them within 1 kpc of the Sun. The annual probability of a RAWD ejecting i-process products within 1 kpc is ≈ 0.018%, or ≈ 0.036% over COSI’s two-year prime mission. However, even at a distance of 5 kpc COSI could still detect γ-ray lines from 86Rb, 89Sr, 95Zr, and 103Ru recently ejected by a solar-metallicity RAWD, with an annual probability of ≈ 0.45%, corresponding to ≈ 0.9% over the prime mission. If the 0.909 MeV 89Sr line of RAWD model A is detected within a few days of ejection, the maximum detection distance increases to around 17.5 kpc, and the annual detection probability rises to ≈ 5–5.5%, corresponding to ≈ 10–11% over COSI’s two-year prime mission.

For Sakurai’s object-like VLTP events, the three events recorded within ≈ 5 kpc over the past 100 yr imply an annual probability of observing another similar event within 1 kpc of 0.12% (0.24% over the prime mission), while an independent estimate from the Galactic star formation rate gives 0.04% (0.08%).

The paper concludes that if such an event does occur, the strongest and most long-lasting emissions are expected from the radioactive decays of 22Na, 89Sr, and 95Zr, at energies of 1.275 MeV, 0.909 MeV, and 0.757 MeV, respectively. The combination of short half-lives and rapid ejection makes these γ-ray lines a characteristic signature of the i process engine. A detection of one of these lines, particularly accompanied by the 22Na line, would constitute direct evidence for an i-process event with rapid ejection.

Based on this paper, here are specific improvements for AI systems:

• Improvement: Train an AI model on the event rate estimation methodology (RAWD formation rates, VLTP frequencies, Galactic density distributions)

• Capability: Given any transient astrophysical phenomenon, automatically compute detection probabilities for specified telescopes, accounting for distance distributions, event rates, and mission lifetimes

• Improvement: Build a neural network trained on the flux calculation equation (Eq. 1) and COSI sensitivity curves

• Capability: Given isotope yields, half-lives, distances, and telescope specifications, instantly predict detectability windows, optimal integration times, and maximum detection distances for any radioactive decay line

• Improvement: Use the paper's yield tables (Table I) across metallicity models A–G to train a classifier

• Capability: Given observed multi-line γ-ray fluxes (e.g., 22Na, 89Sr, 95Zr, 137Cs), automatically identify the progenitor metallicity and distinguish between RAWD, post-AGB VLTP, s-process, and r-process scenarios

• Improvement: Implement the paper's time-averaged flux formalism and optimal integration time (T*iso = 1.82 × half-life)

• Capability: For any planned observation, automatically schedule the optimal integration window to maximize signal-to-noise for short-lived isotopes, accounting for mission phase and background buildup

• Improvement: Integrate the paper's optical counterpart predictions (VLTP brightness, Rubin/LSST detection rates) with γ-ray detection criteria

• Capability: Automatically cross-correlate optical transient alerts with predicted γ-ray line signatures, triggering follow-up observations within the critical few-day window for short-lived isotopes like 89Sr

• Improvement: Train on the Monte Carlo sensitivity analysis (Section C) showing which (n,γ) rates most affect 89Sr and 95Zr yields

• Capability: Given nuclear reaction rate uncertainties, propagate them through nucleosynthesis models to produce confidence intervals on γ-ray flux predictions, flagging which isotopes have robust vs. fragile predictions

• Improvement: Apply the paper's analysis of Sakurai's object (1994 eruption, 3–3.5 kpc distance) to model past events

• Capability: Given historical transient events with known epochs and distances, compute present-day expected fluxes, determine if past telescopes (INTEGRAL/SPI) should have detected them, and constrain ejection physics from non-detections

• Improvement: Use the paper's RAWD spatial distribution (9 within 1 kpc, 1000 total) and VLTP rates

• Capability: Simulate full-sky surveys over arbitrary mission durations, generating synthetic γ-ray line maps with realistic event injection, detection probabilities, and false-alarm rates for mission planning