Demonstrating the utility of improved lanthanide line transition information and generalized decay sources for spectral analysis of kilonovae

Slides: Ryan Wollaeger, Chris Fontes, Adithan Kathirgamaraju

Collaborators: Nick Vieira, Chris Fryer, Oleg Korobkin, Marko Ristic

Outline

• Motivation review

​

• Generalized r-process source study

​

• New Nd data calibration

​

• Summary

Motivation review

• Kilonovae (KNe) are the radioactively powered UV/optical/IR signals from binary neutron star (NS) and NS-black hole mergers (NSM).
• Rapid neutron capture (r-process) nucleosynthesis in KN ejecta produces heavy elements, but the fraction of the Universe abundance is an open question.

​

• Joint electromagnetic (AT2017gfo) and gravitational wave (GW170817) signals from a binary neutron star merger were observed in 2017 at 40 Mpc, providing a wealth of data.

​

• The recently launched James Webb Space Telescope (JWST) should be able to obtain mid-IR spectra, with unprecedented signal-to-noise at ~100-200 Mpc.

​

• Since 2017, with a few other more distant NSM detections, researchers have attempted to:
• infer NS properties (e.g. binary masses) from KN/GW signals,
• constrain total mass of KN ejecta (~10x uncertainty for AT2017gfo/GW170817),
• determine elemental abundances and (relatively recently) spectroscopically identify elements in KN ejecta.

E. Troja, L. Piro, H. van Earten et al (2017)

HST (optical/IR)

Chandra (X-ray)

Motivation review

• 2D two-component model KNe from SuperNu simulations fit NIR bands well at late time, but not optical.
• Optical deficiency across nearly all models.
• Possible reasons:
• ejecta mass morphology,
• radioactive decay model,
• elemental composition,
• numerical algorithms,
• opacity method,
• invalid LTE assumption.

​

​

E. Troja, L. Piro, H. van Earten et al (2017)

The solid lines are a SuperNu KN model simulation

Model optical bands too dim.

Outline

• Motivation review

​

• Generalized r-process source study

​

• New Nd data calibration

​

• Summary

Generalized r-process source

Implementing a Spatially Dependent Heating Rate

Rosswog & Korobkin (2024)

Combining Heating Rates from Multiple Components

• Mass fraction averaged heating rate of each component (similar to previous implementation, well-separated fluid parcels from each ejecta component):

Dynamical-Wind Ejecta Geometries and Light Curves

Wind

Dynamical Ejecta

Source Method Effect on Spectra

Torus-Sphere

Torus-Peanut

Outline

• Motivation review

​

• Generalized r-process source study

​

• New Nd data calibration

​

• Summary

New Nd data calibration

• At the 2023 KN meeting in Stockholm, Fontes presented a comparison of atomic data from Autostructure (Badnell 2011; Badnell et al. 2012; Badnell 2016), HULLAC (Bar-Shalom et al 2001), and the LANL suite of atomic physics codes (Fontes et al 2015), for a simple pure-Nd KN model, showing significant departures in spectra at late time for LTE calculations with SuperNu.
• Neutral Nd sets the late-time spectra, at least for LTE calculations, but atomic data for neutral Nd is very computationally challenging from first principles.
• A more sophisticated multi-element study was pursued by Brethauer et al (2024), showing the effect on observability along with other uncertainties.

​

• We’ve now considered the effect of NIST-guided level calibration of the otherwise ab initio calculations from the LANL suite of atomic physics codes.

Nd data comparison review

• Configuration lists are similar; some other details about the code / models (used for KN studies) at a high level are listed here:

• Not clear what (if any) aspect of these differences have a significant effect, but Fontes et al (2020) demonstrate importance of accurate Hamiltonian (fully-relativistic comparisons).

Nd data comparison review

• Some basic data metrics – number of levels and lines (max values in bold):

When all LANL lines (down to f_c=1e-6) are included: Nd I => (18,104 / 19,116,842), Nd II => (6,888 / 3,390,966), Nd III => (1,650 / 197,010), Nd IV => (241 / 5,276)

Nd data comparison review

• Some basic data metrics – geometric average of gf (max values in bold):

When all LANL lines (down to f_c=1e-6) are included: Nd I => -1.1, Nd II => -3.3, Nd III => -3.1, Nd IV => -2.5

Nd data comparison review

• Some basic data metrics – oscillator strength-averaged line wavelength (microns; max values in bold):

When all LANL lines (down to f_c=1e-6) are included: Nd I => 0.94, Nd II => 0.46, Nd III => 0.26, Nd IV => 0.20

Nd data comparison review

Nd data comparison review

• Electron temperature at early time looks similar among all three calculations.

​

• From electron temperature versus velocity at late time, we see SuperNu simulations with LANL Nd data cool more rapidly.

Nd data comparison review

• Ion fractions behave similarly among the data sets.

​

• Divergence in temperature corresponds to Nd I (neutral) fraction beginning to dominate.

​

• Nd I is where LANL data is also most different from the other data sets.

Nd data comparison review

• Bolometric luminosity ~1.5x discrepant between SuperNu LCs using LANL and other data.

​

• Broadband optical magnitudes decay more quickly with LANL data.

​

• Eliminating the neutral ion stage from the LANL data:
• brings the bolometric luminosity into agreement,
• makes magnitudes shallower.

​

• However, discrepancy persists, suggesting need for accurate Nd I

New Nd data calibration

• Our partial calibration technique replaces calculated level energies with NIST values for some levels, consistently changing oscillator strengths (Fontes et al 2026, submitted):
• This approach is distinct from minimizing energy level differences through the single-electron potentials, as in HULLAC.
• Problem 1: NIST ASD groups fine-structure levels into “multiplets” that are labeled by LS terms, but configuration interaction (CI) in the LANL calculation may preclude a dominant corresponding label for a particular energy level.
• We make an educated guess of the association by comparing energy spacing in the NIST ASD multiplet with energy spacing from the calculation.
• Problem 2: NIST ASD does not contain a complete set of energies for a multiplet.
• For multiplets that are only partially specified in the NIST ASD, we replace all ab initio energies in the multiplet that are specified, and shift the remaining ab initio energies to preserve the relative, ab initio spacing.
• Problem 3: the same LS label, corresponding to different multiplets, can arise multiple times from one configuration in the ab initio calculation, but fewer occurrences of this LS label appear in the NIST ASD.
• As usual, the energies in the multiplets that do exist in the NIST ASD are replaced; the energies in a missing multiplet are all shifted by the same amount in order to preserve the relative spacing between missing and existing multiplets with the same LS label, while preserving the (ab initio) spacing within a missing multiplet

Comparison between LANL and NIST energies�(level labels represent the usual quantum numbers: ^(2S+1)L_J)

Nd³⁺ (or Nd IV)

Nd⁰⁺ (or Nd I)

Comparison between LANL and NIST energies�(level labels represent the usual quantum numbers: ^(2S+1)L_J)

Nd³⁺ (or Nd IV)

Nd⁰⁺ (or Nd I)

(Note: LANL code does not predict correct ground level.)

Ab initio (old) LANL spectra (blue):

Mid-IR structure in LANL Nd opacity reflected in mid-IR spectra (e.g. Korobkin et al 2021)

Calibrated (new) LANL spectra (blue):

Mid-IR structure in LANL Nd opacity significantly impacted

Mid-IR line structure is further diminished

Outline

• Motivation review

​

• Generalized r-process source study

​

• New Nd data calibration

​

• Summary

Summary

• KN modeling requires accurate treatment of the decay power, thermalization, and opacity.
• Decay power can be interpolated in different ways, corresponding to different interpretations of Ye values in ejecta fluid parcels.
• The effect is exacerbated by particular morphologies, variability in local expansion velocity measure, and Ye ranges.
• Effects on the order of 50% in brightness for the most extreme models tested.
• Neutral Nd (I) is a significant contributor to optical->IR photon reprocessing in late time, assuming LTE.
• Nd I opacity is potentially significant: ab initio LANL calculations produce significant mid-IR structure.
• Our preliminary energy level calibration significantly impacts the mid-IR structure.
• Non-LTE ionization (e.g. Hotokezaka 2021, Pognan series, Brethauer 2025) significantly affects neutral fraction, but a small amount of Nd I may motivate the effort for accurate Nd I atomic data.