NASA Ames Research Center published Astrophysics Challenge Problems on November 18, 2026, setting out computational work on black-hole mergers, gravitational-wave signal analysis, and interacting stellar systems. KBR researchers Gianni Mossi and Erik Gustafson prepared the technical slides, which describe the calculations and desired outputs for each application.
Two of the subjects relate directly to the Laser Interferometer Space Antenna, or LISA: generating predicted gravitational-wave signals and separating overlapping sources in the observatory’s measurements. The third examines how faster simulations of many gravitating objects could help researchers infer the history of observed star systems.
The document presents research targets and benchmark problems. It connects mathematical inputs and computing methods with astronomical questions, giving particular attention to calculations that must be repeated many times to compare possible explanations of an observation.
Calculating the Signals From Merging Black Holes
Gravitational waves are changes in spacetime that propagate outward from accelerating masses. Two black holes orbiting each other produce a changing wave signal as they approach and merge. The detailed shape of that signal depends on properties including their masses, spins, and orbit.
Researchers calculate expected signals, called waveforms, so that observations can be compared with predictions. A waveform provides a time-dependent description of the distortion that a detector would measure. Matching such predictions with data helps infer the properties of the source.
The Ames slides focus on demanding combinations of unequal masses, elongated orbits, and spins tilted relative to the orbital motion. Their stated examples include mass ratios ranging from roughly one thousand to one million between the larger and smaller objects. The relevant calculations must represent physical behavior at substantially different scales.
Numerical relativity solves Einstein’s equations on a computer. The document discusses software including the Einstein Toolkit and SpEC, alongside approximation methods that simplify the calculation in particular regimes. The choice of method affects which systems can be modeled and the computing resources required.
The proposed output is a waveform calculated without instrument noise. The authors describe research goals involving improved waveform accuracy or an order-of-magnitude improvement in computing speed. Faster calculations would permit more combinations of source properties to be explored within a given allocation of computing time.
The inputs include starting conditions, black-hole spins and velocities, and the spatial grid used by the solver. Specifying those inputs makes a benchmark reproducible. Comparing the resulting waveform with a reference then provides a way to assess the proposed method.
Separating Overlapping LISA Signals
The second task, described as a LISA global fit, approaches the problem from the measurement side. A detector’s data can contain signals from many astronomical objects at the same time, together with instrument noise. The analysis seeks a collection of source models that jointly explains that data.
Fitting a single isolated source requires estimating its parameters. A global fit also considers how many detectable sources are present and which types of objects produced them. The document gives examples with approximately 10,000 resolvable sources and about 100,000 parameters, illustrating the scale of the proposed inference problem.
The analysis takes processed measurement streams and candidate source models as inputs. Its outputs include probability distributions for source properties and possible sky locations. A probability distribution records the range of values supported by the data instead of reducing every inferred property to one exact number.
The slides discuss Markov chain Monte Carlo methods, which explore possible parameter combinations by sampling them. Existing approaches divide the work into detecting candidate signals and refining the parameters of the resulting source models. The proposed computing work addresses the repeated searches and evaluations involved in that process.
Waveform calculation and global fitting serve related roles. The first predicts what a particular source would produce. The second tests combinations of those predictions against the measured mixture. Improvements to either computation can change how many candidate explanations researchers can evaluate.
Reconstructing the Evolution of Stellar Systems
The third subject is an N-body problem, in which a computer follows the motions of many objects interacting through gravity. In dense stellar systems, the cumulative interactions between stars affect the system’s evolution over long periods.
A forward simulation starts with assumed conditions and calculates how the system changes. The inference task reverses the question: researchers compare simulated outcomes with observations to estimate which starting conditions could have produced the present system. That requires many forward calculations rather than a single run.
The slides identify quantities such as initial masses, sizes, concentrations, and the proportion of stars in binaries as possible inputs. They discuss the use of these calculations to interpret observations relevant to Hubble and the Roman Space Telescope, including the structure and motion of stars in clusters.
The requested computational outputs are observable properties at specified times during a simulation. Those outputs give researchers the comparisons needed to estimate a stellar system’s earlier conditions from the structure and motions measured today.


