arXiv subjects

Marco Crisostomi

Publications and source records attributed to Marco Crisostomi.

At least 19 recordsLinked to original sources

arXiv2026

labrador: A domain-optimized machine-learning tool for gravitational wave inference

Fast and reliable inference of gravitational-wave source parameters is crucial for analyzing large catalogs that are reaching the size of hundreds of detections, and for identifying short-lived electromagnetic counterparts. Neural posterior estimation has emerged as a powerful inference method, where the model is trained on simulated gravitational-wave data at considerable computational cost, but thereafter enables extremely fast and inexpensive inference at test time. Here, we extend this approach by incorporating domain-specific physical insights and methods in the model architecture. These include compressing the data by heterodyning against a reference waveform chosen via approximate likelihood maximization, removing parameter degeneracies through tailored coordinate systems, and eliminating known multimodalities by folding the parameter space. As a result, the network is approximately equivariant to changes in the source parameters, and achieves a reduced training cost and improved model interpretability. Our implementation, called labrador, can be trained end-to-end on a 1-day timescale on $\sim 10^2$ CPU cores and a V100 GPU, achieving a median importance-sampling efficiency of 1% on quadrupolar, aligned-spin signals in a broad mass range (chirp mass $\mathcal{M} \in 1\text{-}50\,\mathrm{M}_\odot$, mass ratio $q > 0.1$). labrador is the first neural inference code to achieve extensive coverage of long-duration signals with secondary masses $m_2 < 10\,\mathrm{M}_\odot$, rendered possible by its equivariance property. Among our novel contributions is a numerically stable procedure that enables neural posterior estimation when the simulation and inference priors differ.

gr-qc↗

arXiv2026

Impact of sky localization uncertainty on ringdown inference

As gravitational-wave ringdown signals grow louder, quasinormal-mode inference depends increasingly on the treatment of extrinsic parameters. Standard analyses fix sky localization - and sometimes also polarization and inclination - to point estimates from a prior inspiral-merger-ringdown analysis, artificially breaking degeneracies and underestimating the true uncertainty of mode-amplitude values. We test two alternatives: uninformative priors on the extrinsic parameters, sampled jointly with the remnant mass, spin, mode amplitudes, and phases; and informed priors on sky position from the full signal posterior. The former yields wider marginal constraints on amplitude posteriors, and both avoid potential bias introduced by fixing the sky localization. In contrast, mode amplitude ratios remain consistent across approaches, making them a robust observable for Kerr spectroscopy. Our publicly available pipeline enables fast ringdown analyses capable of sampling all parameters, requiring tens of minutes on a laptop for a full inference. Applied to GW250114 and GW190521, our methods confirm the robust detection of the $(2,2,1)$ overtone in GW250114, and, for GW190521, find only mild evidence for the $(3,3,0)$ mode.

gr-qc↗

arXiv2025

Beyond diagonal approximations: improved covariance modeling for pulsar timing array data analysis

Pulsar Timing Array (PTA) searches for nHz gravitational-wave backgrounds (GWBs) typically model time-correlated noise by assuming a diagonal covariance in Fourier space, neglecting inter-frequency correlations introduced by the finite observation window. We show that this diagonal approximation can lead to biased estimates of spectral parameters, especially for the common red process that represents the GWB. To address these limitations, we present a method that (i) computes the time-domain autocorrelation on a coarse grid using a fast Fourier transform (FFT), (ii) interpolates it accurately to the unevenly sampled observation times, and (iii) incorporates it into a low-rank likelihood via the Sherman--Morrison--Woodbury identity. Using both analytic covariance comparisons and end-to-end simulations inspired by the NANOGrav 15-year dataset, we demonstrate that our method captures frequency correlations faithfully, avoids Gibbs ringing, and recovers unbiased spectral parameters with modest computational cost. As PTA datasets increase in sensitivity and complexity, our approach offers a practical and scalable path to fully accurate covariance modeling for current and future analyses.

astro-ph.IM↗

arXiv2024

Bayesian evidence estimation from posterior samples with normalizing flows

We propose a novel method ($floZ$), based on normalizing flows, to estimate the Bayesian evidence (and its numerical uncertainty) from a pre-existing set of samples drawn from the unnormalized posterior distribution. We validate it on distributions whose evidence is known analytically, up to 15 parameter space dimensions, and compare with two state-of-the-art techniques for estimating the evidence: nested sampling (which computes the evidence as its main target) and a $k$-nearest-neighbors technique that produces evidence estimates from posterior samples. Provided representative samples from the target posterior are available, our method is more robust to posterior distributions with sharp features, especially in higher dimensions. For a simple multivariate Gaussian, we demonstrate its accuracy for up to 200 dimensions with $10^5$ posterior samples. $floZ$ has wide applicability, e.g., to estimate evidence from variational inference, Markov Chain Monte Carlo samples, or any other method that delivers samples and their likelihood from the unnormalized posterior density. As a physical application, we use $floZ$ to compute the Bayes factor for the presence of the first overtone in the ringdown signal of the gravitational wave data of GW150914, finding good agreement with nested sampling.

stat.ML↗

arXiv2024

Fixing the dynamical evolution of self-interacting vector fields

Numerical simulations of the Cauchy problem for self-interacting massive vector fields often face instabilities and apparent pathologies. We explicitly demonstrate that these issues, previously reported in the literature, are actually due to the breakdown of the well-posedness of the initial-value problem. This is akin to shortcomings observed in scalar-tensor theories when derivative self-interactions are included. Building on previous work done for k-essence, we characterize the well-posedness breakdowns, differentiating between Tricomi and Keldysh-like behaviors. We show that these issues can be avoided by ``fixing the equations'', enabling stable numerical evolutions in spherical symmetry. Additionally, we show that for a class of vector self-interactions, no Tricomi-type breakdown takes place. Finally, we investigate initial configurations for the massive vector field which lead to gravitational collapse and the formation of black holes.

gr-qc↗

arXiv2024

Rapid parameter estimation for pulsar-timing-array datasets with variational inference and normalizing flows

In the gravitational-wave analysis of pulsar-timing-array datasets, parameter estimation is usually performed using Markov Chain Monte Carlo methods to explore posterior probability densities. We introduce an alternative procedure that relies instead on stochastic gradient-descent Bayesian variational inference, whereby we obtain the weights of a neural-network approximation of the posterior by minimizing the Kullback-Leibler divergence of the approximation from the exact posterior. This technique is distinct from simulation-based inference with normalizing flows, since we train the network for a single dataset, rather than the population of all possible datasets, and we require the computation of the data likelihood and its gradient. Unlike Markov Chain methods, our technique can transparently exploit highly parallel computing platforms. This makes it extremely fast on modern graphical processing units, where it can analyze the NANOGrav 15-yr dataset in few tens of minutes, depending on the probabilistic model, as opposed to hours or days with the analysis codes used until now. We expect that this speed will unlock new kinds of astrophysical and cosmological studies of pulsar-timing-array datasets. Furthermore, variational inference would be viable in other contexts of gravitational-wave data analysis as long as differentiable and parallelizable likelihoods are available.

gr-qc↗

arXiv2023

Implications of the pulsar timing array detections for massive black hole mergers in the LISA band

The recent evidence of a stochastic background of gravitational waves in the nHz band by pulsar-timing array (PTA) experiments has shed new light on the formation and evolution of massive black hole binaries with masses $\sim 10^8$--$10^9 M_\odot$. The PTA data are consistent with a population of such binaries merging efficiently after the coalescence of their galactic hosts, and presenting masses slightly larger than previously expected. This momentous discovery calls for investigating the prospects of detecting the smaller ($\sim 10^5$--$10^7 M_\odot$) massive black hole binaries targeted by the Laser Interferometer Space Antenna (LISA). By using semi-analytic models for the formation and evolution of massive black hole binaries calibrated against the PTA results, we find that LISA will observe at least a dozen and up to thousands of black hole binaries during its mission duration. The minimum number of detections rises to $\sim 70$ if one excludes models that only marginally reproduce the quasar luminosity function at $z=6$. We also assess LISA's parameter estimation capabilities with state-of-the-art waveforms including higher modes and realistic instrumental response, and find that the masses, sky position, and distance will typically be estimated to within respectively 1%, 10 square degrees, and 10% for the detected systems (assuming a 4-year mission).

astro-ph.GA↗

arXiv2023

Neural Posterior Estimation with guaranteed exact coverage: the ringdown of GW150914

We analyze the ringdown phase of the first detected black-hole merger, GW150914, using a simulation-based inference pipeline based on masked autoregressive flows. We obtain approximate marginal posterior distributions for the ringdown parameters, namely the mass, spin, and the amplitude and phases of the dominant mode and its first overtone. Thanks to the locally amortized nature of our method, we are able to calibrate our posteriors with injected simulations, producing posterior regions with guaranteed (i.e. exact) frequentist coverage of the true values. For GW150914, our calibrated posteriors provide only mild evidence (~ 2 sigma) for the presence of an overtone, even if the ringdown is assumed to start at the peak of the amplitude.

gr-qc↗

arXiv2023

Robustness of kinetic screening against matter coupling

We investigate neutron star solutions in scalar-tensor theories of gravity with first-order derivative self-interactions in the action and in the matter coupling. We assess the robustness of the kinetic screening mechanism present in these theories against general conformal couplings to matter. The latter include ones leading to the classical Damour-Esposito-Farèse scalarization, as well as ones depending on the kinetic term of the scalar field. We find that kinetic screening always prevails over scalarization, and that kinetic couplings with matter enhance the suppression of scalar gradients inside the star even more, without relying on the non-linear regime. Fine tuning the kinetic coupling with the derivative self-interactions in the action allows one to partially cancel the latter, resulting in a weakening of kinetic screening inside the star. This effect represents a novel way to break screening mechanisms inside matter sources, and provides new signatures that might be testable with astrophysical observations.

gr-qc↗

arXiv2022

The well-posedness of the Cauchy problem for self-interacting vector fields

We point out that the initial-value (Cauchy) problem for self-interacting vector fields presents the same well-posedness issues as for first-order derivative self-interacting scalar fields (often referred to as $k$-essence). For the latter, suitable strategies have been employed in the last few years to successfully evolve the Cauchy problem at the level of the infrared theory, without the need for an explicit ultraviolet completion. We argue that the very same techniques can also be applied to self-interacting vector fields, avoiding a number of issues and "pathologies" recently found in the literature.

gr-qc↗

arXiv2022

New Horizons for Fundamental Physics with LISA

The Laser Interferometer Space Antenna (LISA) has the potential to reveal wonders about the fundamental theory of nature at play in the extreme gravity regime, where the gravitational interaction is both strong and dynamical. In this white paper, the Fundamental Physics Working Group of the LISA Consortium summarizes the current topics in fundamental physics where LISA observations of GWs can be expected to provide key input. We provide the briefest of reviews to then delineate avenues for future research directions and to discuss connections between this working group, other working groups and the consortium work package teams. These connections must be developed for LISA to live up to its science potential in these areas.

gr-qc↗

arXiv2022

Cosmology with the Laser Interferometer Space Antenna

The Laser Interferometer Space Antenna (LISA) has two scientific objectives of cosmological focus: to probe the expansion rate of the universe, and to understand stochastic gravitational-wave backgrounds and their implications for early universe and particle physics, from the MeV to the Planck scale. However, the range of potential cosmological applications of gravitational wave observations extends well beyond these two objectives. This publication presents a summary of the state of the art in LISA cosmology, theory and methods, and identifies new opportunities to use gravitational wave observations by LISA to probe the universe.

astro-ph.CO↗

arXiv2022

No evidence of kinetic screening in simulations of merging binary neutron stars beyond general relativity

We have conducted fully relativistic simulations in a class of scalar-tensor theories with derivative self-interactions and screening of local scales. By using high-resolution shock-capturing methods and a non-vanishing shift vector, we have managed to avoid issues plaguing similar attempts in the past. We have first confirmed recent results by ourselves in spherical symmetry, obtained with an approximate approach and pointing at a partial breakdown of the screening in black-hole collapse. Then, we considered the late inspiral and merger of binary neutron stars. We found that screening tends to suppress the (subdominant) dipole scalar emission, but not the (dominant) quadrupole scalar mode. Our results point at quadrupole scalar signals as large as (or even larger than) in Fierz-Jordan-Brans-Dicke theories with the same conformal coupling, for strong-coupling scales in the MeV range that we can simulate.

gr-qc↗

arXiv2021

Kinetic screening in nonlinear stellar oscillations and gravitational collapse

We consider k-essence, a scalar-tensor theory with first-order derivative self-interactions that can screen local scales from scalar fifth forces, while allowing for sizeable deviations from General Relativity on cosmological scales. We construct fully nonlinear static stellar solutions that show the presence of this screening mechanism, and we use them as initial data for simulations of stellar oscillations and gravitational collapse in spherical symmetry. We find that for k-essence theories of relevance for cosmology, the screening mechanism works in the case of stellar oscillation and suppresses the monopole scalar emission to undetectable levels. In collapsing stars, we find that the Cauchy problem, although locally well posed, can lead to diverging characteristic speeds for the scalar field. By introducing a ''fixing equation'' in the spirit of J. Cayuso, N. Ortiz, and L. Lehner [Phys. Rev. D 96, 084043 (2017)], inspired in turn by dissipative relativistic hydrodynamics, we manage to evolve collapsing neutron stars past the divergence of the characteristic speeds. We show that, in these systems, the screening mechanism is less efficient than for oscillating and static stars, because the collapsing star must shed away all of its scalar hair before forming a black hole. For k-essence theories of relevance for cosmology, the characteristic frequency of the resulting scalar monopole signal is too low for terrestrial detectors, but we conjecture that space-borne interferometers such as LISA might detect it if a supernova explodes in the Galaxy.

gr-qc↗

arXiv2021

Degenerate Horava gravity

Horava gravity breaks Lorentz symmetry by introducing a dynamical timelike scalar field (the khronon), which can be used as a preferred time coordinate (thus selecting a preferred space-time foliation). Adopting the khronon as the time coordinate, the theory is invariant only under time reparametrizations and spatial diffeomorphisms. In the infrared limit, this theory is sometimes referred to as khronometric theory. Here, we explicitly construct a generalization of khronometric theory, which avoids the propagation of Ostrogradski modes as a result of a suitable degeneracy condition (although stability of the latter under radiative corrections remains an open question). While this new theory does not have a general-relativistic limit and does not yield a Friedmann-Robertson-Walker-like cosmology on large scales, it still passes, for suitable choices of its coupling constants, local tests on Earth and in the solar system, as well as gravitational-wave tests. We also comment on the possible usefulness of this theory as a toy model of quantum gravity, as it could be completed in the ultraviolet into a 'degenerate Horava gravity' theory that could be perturbatively renormalizable without imposing any projectability condition.

hep-th↗

arXiv2021

K-dynamics: well-posed 1+1 evolutions in K-essence

We study the vacuum Cauchy problem for K-essence, i.e. cosmologically relevant scalar-tensor theories that involve first-order derivative self-interactions, and which pass all existing gravitational wave bounds. We restrict to spherical symmetry and show that there exists a large class of theories for which no breakdown of the Cauchy problem occurs outside apparent black hole horizons, even in the presence of scalar shocks/caustics, except for a small set of initial data sufficiently close to critical black hole collapse. We characterise these problematic initial data, and show that they lead to large or even diverging (coordinate) speeds for the characteristic curves. We discuss the physical relevance of this problem and propose ways to overcome it.

gr-qc↗

arXiv2021

Dynamics of Screening in Modified Gravity

Gravitational theories differing from General Relativity may explain the accelerated expansion of the Universe without a cosmological constant. However, to pass local gravitational tests, a "screening mechanism" is needed to suppress, on small scales, the fifth force driving the cosmological acceleration. We consider the simplest of these theories, i.e. a scalar-tensor theory with first-order derivative self-interactions, and study isolated (static and spherically symmetric) non-relativistic and relativistic stars. We produce screened solutions and use them as initial data for non-linear numerical evolutions in spherical symmetry. We find that these solutions are stable under large initial perturbations, as long as they do not cause gravitational collapse. When gravitational collapse is triggered, the characteristic speeds of the scalar evolution equation diverge, even before apparent black-hole or sound horizons form. This casts doubts on whether the dynamical evolution of screened stars may be predicted in these effective field theories.

gr-qc↗

arXiv2020

Consistency relations for large-scale structure in modified gravity and the matter bispectrum

We study perturbation theory for large-scale structure in the most general scalar-tensor theories propagating a single scalar degree of freedom, which include Horndeski theories and beyond. We model the parameter space using the effective field theory of dark energy. For Horndeski theories, the gravitational field and fluid equations are invariant under a combination of time-dependent transformations of the coordinates and fields. This symmetry allows one to construct a physical adiabatic mode which fixes the perturbation-theory kernels in the squeezed limit and ensures that the well-known consistency relations for large-scale structure, originally derived in general relativity, hold in modified gravity as well. For theories beyond Horndeski, instead, one generally cannot construct such an adiabatic mode. Because of this, the perturbation-theory kernels are modified in the squeezed limit and the consistency relations for large-scale structure do not hold. We show, however, that the modification of the squeezed limit depends only on the linear theory. We investigate the observational consequences of this violation by computing the matter bispectrum. In the squeezed limit, the largest effect is expected when considering the cross-correlation between different tracers. Moreover, the individual contributions to the 1-loop matter power spectrum do not cancel in the infrared limit of the momentum integral, modifying the power spectrum on non-linear scales.

astro-ph.CO↗