Could Heavy-Element Dust Explain Kilonovae’s Lingering Infrared Glow?

When neutron stars collide, some of their neutron-rich material is flung into space. Radioactive decay of newly made nuclei heats this expanding debris, creating a short-lived glow called a kilonova. The event GW170817 and its kilonova, AT2017gfo, helped establish mergers as sites of rapid neutron capture, or the r-process, which makes many heavy elements. A later candidate, AT2023vfi, showed strong infrared emission at a temperature below 1,000 kelvin. Explaining that cool, bright continuum with atomic emission and absorption alone is difficult.

Nanae Domoto, Kenta Hotokezaka and Daniel Kasen ask whether dust could supply another source of infrared light. Their proposed grains would be made from refractory r-process elements—materials that can remain solid at high temperatures—including zirconium, tungsten and osmium. A grain absorbs energy and re-emits it as thermal radiation, potentially producing a smooth infrared glow.

Modeling how grains might form

The researchers first estimated whether the merger ejecta could contain enough refractory material to make an infrared-emitting dust mass. They then calculated how small clusters of atoms might grow as the ejecta expand and cool. Unlike a classical nucleation approach used in earlier work, their kinetic model tracks clusters of different sizes and includes their formation and breakup. Because reaction-rate data for these rare heavy-element species are sparse, they use tungsten-cluster rates as a proxy for other elements.

The calculation does not derive the ejecta’s temperature from a fully self-consistent thermal model. Instead, it adopts a simple power-law temperature history, fitted to observed kilonova temperatures. The team also ran radiation-transport simulations—calculations of how light moves through the debris—with and without dust, to see how grains would change the predicted spectrum.

What the models suggest

Under the adopted conditions, clusters can grow into dust, especially in denser, slower-moving ejecta. The modeled dust begins forming roughly 10 days after a merger. Faster, less dense material is less favorable, so the authors do not predict that all ejecta—or all kilonovae—will necessarily become dusty.

In the radiation simulations, dust in the inner ejecta provides a smooth, nearly blackbody infrared continuum. The model spectrum resembles important features of AT2023vfi, including its strong infrared component; dust-free outer layers can contribute a smaller feature near 2.2 micrometers. Without dust, the simulated late-time spectrum is much more dominated by emission lines and does not resemble the observed infrared continuum as well. This shows that dust is a plausible way to produce the signal within the model—not that the observations uniquely establish dust as its source.

A promising idea with major unknowns

The result rests on physical conditions that are not yet securely known. In particular, the temperature prescription is empirical, and the authors say it is uncertain whether the gas itself cools enough for grains to form. The model assumes neutral gas, simplifies the chemical network, and lacks measured reaction rates for many relevant heavy-element clusters. It also omits some possible grain-destruction effects. The authors specifically flag energetic particles from alpha decay and fission as a potential threat to grain survival.

The grain-growth calculations follow clusters only up to a limited size; estimated final dust mass is therefore treated as an upper limit, and later growth through collisions between clusters is not modeled. Meanwhile, the radiation calculation assumes a simplified, spherical ejecta structure. The authors say adjustments to ejecta parameters and geometry would be needed to match individual events in detail. They also caution that their late-time atomic spectra have uncertainties, including incomplete atomic data and the omission of non-local thermodynamic equilibrium effects.

The paper further notes that late-time observations of AT2017gfo are sparse: the cited infrared evidence consists of measurements in only two bands at one epoch, so it does not support detailed modeling. More time-resolved infrared spectra could help distinguish a thermal dust continuum from alternative emission processes and test whether different mergers produce different late-time signals.

This work is a preprint, not a settled observational identification of r-process dust. Its main contribution is a proposed formation pathway and a demonstration that dust can reproduce aspects of the infrared spectrum under stated assumptions. If future observations and improved chemistry and thermal models support the idea, infrared light could offer a new way to study heavy-element production in neutron-star mergers. For now, the grains remain a compelling possibility, with their formation and survival still open questions.