Luminosity predictions for tungsten, platinum and gold
Heavy elements are predicted to be produced in neutron star mergers in significant quantities, which we would expect to see in the kilonova remnant. However, to disentangle the dense complex spectra of a kilonova we require the appropriate atomic data. In this paper we present calculations for the strengths of lines in the collisionally dominated late-time regime for tungsten, platinum and gold. Using a Dirac atomic R-matrix approach. For doubly-ionised tungsten in particular a line at ~4.5 microns may be prominent, using observations of the AT2017gfo and AT2023vfi kilonovae the luminosity of the 4.5 micron feature can be measured and a theoretical estimate of the mass of tungsten required to achieve that luminosity proposed. Informed by merger models and nucleosynthesis calculations, we then show how constraints on tungsten can be used to infer characteristics of the kilonova, such as the yields of other 3rd r-process peak elements or the lanthanides and actinides, or how the theoretical shape of a line is affected by the velocity distribution of the species it is from.
Luminosity predictions for the first three ionization stages of W, Pt, and Au to probe potential sources of emission in kilonova, 2025. Read the full MNRAS paper here
Luminosity density predicted as a function of wavelength (nm) for neutral tungsten (W I) and its first and second ionization stages (W II and W III). Calculations of this sort allow us to identify where each ion is most likely to contribute to the observed spectrum, and comparisons with data allow us to estimate how much of the ion may be present. The figure shows spectra generated at temperatures of Te = 0.15/0.25 eV and electron density ne = 1 × 10^6 cm−3, assuming a total mass of 1 × 10−3 M⊙ for each ion.
Machine learning and gravitational waves
Artificial intelligence and machine learning revolutionize daily life. In a recently published paper we show that machine learning methods can also be successfully employed to model the complex gravitational-wave signals from neutron star mergers [Phys. Rev. D 111, 023002 (2025)]. This is important to actually find those signals because their discovery relies on the availability of models that can be searched for in the data from gravitational-wave detectors. This modelling should be fast as well as accurate, which is where the new algorithms pay off.
Gravitational-wave model for neutron star merger remnants with supervised learning, Physical Review, 2025. Read full paper
The gravitational-wave signal of a neutron star merger as function of time from a sophisticated hydrodynamical merger simulation on a supercomputer (black line). The red line displays the models generated by a machine learning algorithm. It accurately matches the signal but only took a few milliseconds to be computed.