Searcharxiv⌕ Search

arXiv subjects

Nolan Smyth

Publications and source records attributed to Nolan Smyth.

18 recordsLinked to original sources
arXiv2026

Microlensing Detection and Inference via Learned Bayes Factors

We present a unified framework for gravitational microlensing event detection and parameter inference. Traditional pipelines use deterministic hard cuts on photometric statistics, systematically missing low-magnification events in the finite-source regime. We instead frame detection as Bayesian model comparison using Evidence Networks, which learn calibrated Bayes factors from binary-labeled simulations, and combine this with Neural Posterior Estimation (NPE) for amortized parameter inference. Both share a transformer encoder that handles irregularly-sampled time series without imputation. On simulated Roman Space Telescope data, our Evidence Network achieves $99.9\%$ detection efficiency with a false-positive rate below $6\times10^{-4}$ on simulated data with augmentation and noise, but no astrophysical confounders. Gains are most dramatic in the extreme finite-source regime ($\rho \gtrsim 5$), where detection rates reach ${\sim}95\%$ versus ${\sim}65\%$ for hard cuts, precisely the short-duration free-floating planet events most constraining for formation scenarios. Our NPE provides calibrated posteriors, working towards real-time analysis at survey scale.

astro-ph.IM↗
arXiv2026

Efficient Safety Alignment of Language Models via Latent Personality Traits

Current safety methods for large language models are known to be vulnerable to adversarial attacks, motivating research into robust alternatives. Latent Adversarial Training (LAT) is among the most effective defenses, but can degrade utility and requires training on large datasets of harmful prompts. We introduce Latent Personality Alignment (LPA), which replaces explicit harm refusal with adversarial training on just 66 harm-agnostic statements drawn from psychometric personality literature. We hypothesize that personality-anchored representations share latent structure with harm avoidance, so adversarially stabilizing them implicitly constrains the subspace exploited by jailbreak attacks. LPA achieves near-zero attack success rates on HarmBench across direct requests and five jailbreak methods, despite never seeing harmful content during training and no loss of performance on standard benchmarks. Moreover, the training process is lightweight; the entire procedure completes in minutes on a single GPU and uses 75x fewer examples than standard LAT. Extensive ablations demonstrate the robustness, efficiency, and generalization of our method.

cs.LG↗
arXiv2025

Transformer Embeddings for Fast Microlensing Inference

The search for free-floating planets (FFPs) is a key science driver for upcoming microlensing surveys like the Nancy Grace Roman Galactic Exoplanet Survey. These rogue worlds are typically detected via short-duration microlensing events, the characterization of which often requires analyzing noisy, irregularly-sampled observations. We present a pipeline for this task using simulation-based inference. We use a Transformer encoder to learn a compressed summary representation of the raw time-series data, which in turn conditions a neural posterior estimator. We demonstrate that our method produces accurate and well-calibrated posteriors over three orders of magnitude faster than traditional methods. We also demonstrate its performance on KMT-BLG-2019-2073, a short-duration FFP candidate event.

astro-ph.IM↗
arXiv2025

Neural Deprojection of Galaxy Stellar Mass Profiles

We introduce a neural approach to dynamical modeling of galaxies that replaces traditional imaging-based deprojections with a differentiable mapping. Specifically, we train a neural network to translate Nuker profile parameters into analytically deprojectable Multi Gaussian Expansion components, enabling physically realistic stellar mass models without requiring optical observations. We integrate this model into SuperMAGE, a differentiable dynamical modelling pipeline for Bayesian inference of supermassive black hole masses. Applied to ALMA data, our approach finds results consistent with state-of-the-art models while extending applicability to dust-obscured and active galaxies where optical data analysis is challenging.

astro-ph.GA↗
arXiv2025

Primordial Black Holes and the First Stars

Primordial black holes (PBHs) constitute a compelling dark matter candidate whose gravitational effects could significantly influence early cosmic structure formation. We investigate the impact of PBHs on Population III star formation through detailed $N$-body and hydrodynamic simulations, extending beyond previous semi-analytical approaches. Our results reveal a mass-dependent dichotomy in PBH effects: massive PBHs ($M_{\rm PBH} \gtrsim 10^2 M_\odot$) with sufficient abundance can accelerate structure formation and shift Pop III formation to higher redshifts, potentially conflicting with observational constraints from high-redshift galaxy surveys. Conversely, lower-mass PBHs can induce tidal disruption of gas-rich minihalos, suppressing star formation and delaying the cosmic dawn depending on their abundance. We quantify these competing effects to derive new constraints on the PBH mass function and their contribution to the total dark matter density, with implications for forthcoming observations with the James Webb Space Telescope and 21-cm cosmology experiments.

astro-ph.CO↗
arXiv2024

Kinetic recoupling of dark matter

We study the possibility that dark matter re-enters kinetic equilibrium with a radiation bath after kinetic decoupling, a scenario we dub kinetic recoupling. This naturally occurs, for instance, with certain types of resonantly-enhanced interactions, or as the result of a phase transition. While late kinetic decoupling damps structure on small scales below a cutoff, kinetic recoupling produces more complex changes in the power spectrum that depend on the nature and extent of the recoupling period. We explore the features that kinetic recoupling imprints upon the matter power spectrum, and discuss how such features can be traced to dark matter microphysics with future observations.

hep-ph↗
arXiv2024

New Light on Dark Extended Lenses with the Roman Space Telescope

The Roman Space Telescope's Galactic Bulge Time Domain Survey will constitute the most sensitive microlensing survey of the Galactic Bulge to date, opening up new opportunities to search for dark matter (DM). Many extensions of the Standard Model predict the formation of extended DM substructures, such as DM subhalos, boson/axion stars, and halo-dressed primordial black holes. We demonstrate that for such targets, Roman will be sensitive to a broad parameter space up to four orders of magnitude below existing constraints. Our analysis can be readily applied to other extended DM configurations as well.

astro-ph.CO↗
arXiv2024

Searching for Free-Floating Planets with TESS: Results from Sectors 61-65

Though free-floating planets (FFPs) may outpopulate their bound counterparts in the terrestrial-mass range, they remain one of the least explored exoplanet demographics. Due to their negligible electromagnetic emission at all wavelengths, the only observational technique able to detect these worlds is gravitational microlensing. Microlensing by terrestrial-mass FFPs induces rare, short-duration magnifications of background stars, requiring high-cadence, wide-field surveys to detect these events. The Transiting Exoplanet Survey Satellite (TESS), though designed to detect close-bound exoplanets via transits, boasts a Full-Frame Image cadence as short as 200 seconds and has monitored hundreds of millions of stars, providing a unique dataset in which to search for rare short-duration transients. We have performed a preliminary search for FFP microlensing in 7.5 million light curves from TESS Sectors 61 - 65. We find one short-duration event with a light curve morphology consistent with expectations for a low-mass FFP, but in tension with the expected FFP abundance in this mass range. We consider possible false positive interpretations of this event such as stellar flares, hearbeat binaries, and centrifugal breakout. We find that all interpretations pose some challenges, and discuss the possibility that the event may constitute a first example of a new class of pernicious false positives that future space-based microlensing efforts will encounter. Our ongoing search through the TESS dataset will significantly support the upcoming hunt for rogue worlds with dedicated space-based microlensing surveys, and our results may be used alongside these surveys to place interesting constraints on the spatial distribution of FFPs in the Galaxy.

astro-ph.EP↗
arXiv2024

Constraints on Primordial Black Holes from $N$-body simulations of the Eridanus II Stellar Cluster

The evolution of old, compact stellar structures provides strong constraints on macroscopic dark matter candidates such as primordial black holes. In view of recent observational data for the Eridanus II dwarf galaxy, we perform the first $N$-body simulations of its central stellar cluster to model dynamical heating by PBHs. We find evidence that such candidates must be lighter than about one solar mass if they constitute the totality of the dark matter. We additionally derive constraints on the fraction of the dark matter in macroscopic objects as a function of mass, by modeling the remainder of the dark matter as standard, fluid-like cold dark matter.

astro-ph.CO↗
arXiv2024

Dark Black Holes in the Mass Gap

In the standard picture of stellar evolution, pair-instability -- the energy loss in stellar cores due to electron-positron pair production -- is predicted to prevent the collapse of massive stars into black holes with mass in the range between approximately 50 and 130 solar masses -- a range known as the "{\em black hole mass gap}". LIGO detection of black hole binary mergers containing one or both black holes with masses in this {\em mass gap} thus challenges the standard picture, possibly pointing to an unexpected merger history, unanticipated or poorly understood astrophysical mechanisms, or new physics. Here, we entertain the possibility that a "dark sector" exists, consisting of dark electrons, dark protons, and electromagnetic-like interactions, but no nuclear forces. Dark stars would inevitably form given such dark sector constituents, possibly collapsing into black holes with masses within the mass gap. We study in detail the cooling processes necessary for successful stellar collapse in the dark sector and show that for suitable choices of the particle masses, we indeed predict populating the mass gap with dark sector black holes. In particular, we numerically find that the heavier of the two dark sector massive particles cannot be lighter than, approximately, the visible sector proton for the resulting dark sector black holes to have masses within the mass gap. We discuss constraints on this scenario and how to test it with future, larger black hole merger statistics.

hep-ph↗
arXiv2023

Constraints on sub-terrestrial free-floating planets from Subaru microlensing observations

The abundance of protoplanetary bodies ejected from their parent star system is presently poorly-constrained. With only two existing optical observations of interstellar objects in the $10^{8} - 10^{10}$ kg mass range and a small number of robust microlensing observations of free-floating planets (FFPs) in the $10^{24} - 10^{25}$ kg mass range, there is a large range of masses for which there are no existing measurements of the unbound population. The three primary microlensing surveys currently searching for FFPs operate at a cadence greater than 15 minutes, which limits their ability to observe events associated with bodies with a mass much below an Earth mass. We demonstrate that existing high-cadence observations of M31 with the Subaru Hyper Suprime-Cam place constraints on the abundance of unbound objects at sub-terrestrial masses, with peak sensitivity at $10^{-4}~M_\oplus$ for Milky Way lenses and $10^{-1}~M_\oplus$ for lenses in M31. For a fiducial $\frac{dn}{dM}\propto M^{-2}$ mass distribution, we find that the abundance of unbound objects is constrained to $n_\text{unbound} < 1.4 \times 10^{7} ~\rm{pc}^{-3}$ for masses within 1 dex of $10^{-4}~M_\oplus$. Additionally, we compute limits on an artificial ``monochromatic'' distribution of unbound objects and compare to existing literature, demonstrating that the assumed spatial distribution of lenses has very significant consequences for the sensitivity of microlensing surveys. While the observations ultimately do not probe abundances suggested by current models of planetary formation, our limits place direct observational constraints on the unbound population in the sub-terrestrial mass range and motivate new observational strategies for microlensing surveys.

astro-ph.EP↗
arXiv2023

Rogue worlds meet the dark side: revealing terrestrial-mass primordial black holes with the Nancy Grace Roman Space Telescope

Gravitational microlensing is one of the strongest observational techniques to observe non-luminous astrophysical bodies. Existing microlensing observations provide tantalizing evidence of a population of low-mass objects whose origin is unknown. These events may be caused by terrestrial-mass free-floating planets or by exotic objects such as primordial black holes. However, the nature of these objects cannot be resolved on an event-by-event basis, as the induced light curve is degenerate for lensing bodies of identical mass. One must instead statistically compare \textit{distributions} of lensing events to determine the nature of the lensing population. While existing surveys lack the statistics required to identify multiple subpopulations of lenses, this will change with the launch of the Nancy Grace Roman Space Telescope. Roman's Galactic Bulge Time Domain Survey is expected to observe hundreds of low-mass microlensing events, enabling a robust statistical characterization of this population. In this paper, we show that by exploiting features in the distribution of lensing event durations, Roman will be sensitive to a subpopulation of primordial black holes hidden amongst a background of free-floating planets. Roman's reach will extend to primordial black hole dark matter fractions as low as $f_\text{PBH} = 10^{-4}$ at peak sensitivity, and will be able to conclusively determine the origin of existing ultrashort-timescale microlensing events. A positive detection would provide evidence that a significant fraction of the cosmological dark matter consists of macroscopic, non-luminous objects.

astro-ph.CO↗
arXiv2023

Fast Parameter Inference on Pulsar Timing Arrays with Normalizing Flows

Pulsar timing arrays (PTAs) perform Bayesian posterior inference with expensive MCMC methods. Given a dataset of ~10-100 pulsars and O(10^3) timing residuals each, producing a posterior distribution for the stochastic gravitational wave background (SGWB) can take days to a week. The computational bottleneck arises because the likelihood evaluation required for MCMC is extremely costly when considering the dimensionality of the search space. Fortunately, generating simulated data is fast, so modern simulation-based inference techniques can be brought to bear on the problem. In this paper, we demonstrate how conditional normalizing flows trained on simulated data can be used for extremely fast and accurate estimation of the SGWB posteriors, reducing the sampling time from weeks to a matter of seconds.

astro-ph.IM↗
arXiv2022

Simulation Based Inference for Efficient Theory Space Sampling: an Application to Supersymmetric Explanations of the Anomalous Muon (g-2)

For the purpose of minimizing the number of sample model evaluations, we propose and study algorithms that utilize (sequential) versions of likelihood-to-evidence ratio neural estimation.We apply our algorithms to a supersymmetric interpretation of the anomalous muon magnetic dipole moment in the context of a phenomenological minimal supersymmetric extension of the standard model, and recover non-trivial models in an experimentally-constrained theory space. Finally we summarize further potential possible uses of these algorithms in future studies.

hep-ph↗
arXiv2022

Gravitational Baryogenesis and Dark Matter from Light Black Holes

We study a scenario in which the baryon asymmetry is created through Hawking radiation from primordial black holes via a dynamically-generated chemical potential. This mechanism can also be used to generate the observed dark matter abundance, regardless of whether or not the black holes fully evaporate. In the case that evaporation ceases, the observed dark matter abundance is generically comprised of both relic black holes and an asymmetric dark matter component. We show that this two-component dark matter scenario can simultaneously account for the observed baryon asymmetry and the cosmological dark matter, a possibility which evades constraints on either individual candidate.

hep-ph↗
arXiv2021

Optimal observing strategies for velocity-suppressed dark matter annihilation

Numerous particle models for the cosmological dark matter feature a pair-annihilation rate that scales with powers of the relative velocity between the annihilating particles. As a result, the annihilation rate in the central regions of a dark matter halo can be significantly lower than at the halo's periphery for particular ambient gravitational potentials. While this might be offset by an increasing dark matter pair number density in the inner halo, it raises the question: what angular region for dark matter models with velocity-suppressed annihilation rates optimizes signal-to-noise? Here, we consider simplified background models for galactic and extragalactic targets and demonstrate that the optimal observing strategy varies greatly case-by-case. Generally, a bright central source warrants an annular region of interest, while a flatter background warrants as large as possible an angular region, possibly including the central regions.

astro-ph.CO↗
arXiv2020

Updated Constraints on Asteroid-Mass Primordial Black Holes as Dark Matter

Microlensing of stars places significant constraints on sub-planetary-mass compact objects, including primordial black holes, as dark matter candidates. As the lens' Einstein radius in the source plane becomes comparable to the size of the light source, however, source amplification is strongly suppressed, making it challenging to constrain lenses with a mass at or below $10^{-10}$ solar masses, i.e. asteroid-mass objects. Current constraints, using Subaru HSC observations of M31, assume a fixed source size of one solar radius. Here we point out that the actual stars in M31 bright enough to be used for microlensing are typically much larger. We correct the HSC constraints by constructing a source size distribution based on the M31 PHAT survey and on a synthetic stellar catalogue, and by correspondingly weighing the finite-size source effects. We find that the actual HSC constraints are weaker by up to almost three orders of magnitude in some cases, broadening the range of masses for which primordial black holes can be the totality of the cosmological dark matter by almost one order of magnitude.

astro-ph.CO↗