Search

There are expected to be millions of isolated black holes in the Galaxy resulting from the death of massive stars. Measuring the abundance and properties of this remnant population would shed light on the end stages of stellar evolution and the evolution paths of black hole systems. Detecting isolated black holes is currently only possible via gravitational microlensing which has so far yielded one definitive detection. The difficulty in finding microlensing black holes lies in having to choose a small subset of events based on characteristics of their lightcurves to allocate expensive and scarce follow-up resources to confirm the identity of the lens. Current methods either rely on simple cuts in parameter space without using the full distribution information or are only effective on a small subsets of events. In this paper we present a new lens classification method. The classifier takes in posterior constraints on lightcurve parameters and combines them with a Galactic simulation to estimate the lens class probability. This method is flexible and can be used with any set of microlensing lightcurve parameters making it applicable to large samples of events. We make this classification framework available via the popclass python package. We apply the classifier to $\sim10,000$ microlensing events from the OGLE survey and find $23$ high-probability black hole candidates. Our classifier also suggests that the only known isolated black hole is an observational outlier according to current Galactic models and allocation of astrometric follow-up on this event was a high-risk strategy., Comment: 19 pages, 10 figures. Submitted to AAS Journals

We introduce a population model to analyze the mixing between hypothesised power-law and $\sim 35 M_\odot$ Gaussian bump black hole populations in the latest gravitational wave catalog, GWTC-3, estimating their co-location and separation. We find a relatively low level of mixing, $3.1^{+5.0}_{-3.1}\%$, between the power-law and Gaussian populations, compared to the percentage of mergers containing two Gaussian bump black holes, $5.0^{+3.2}_{-1.7}\%$. Our analysis indicates that black holes within the Gaussian bump are generally separate from the power-law population, with only a minor fraction engaging in mixing and contributing to the $\mathcal{M} \sim 14 M_\odot$ peak in the chirp mass. This leads us to identify a distinct population of Binary Gaussian Black Holes (BGBHs) that arise from mergers within the Gaussian bump. We suggest that current theories for the formation of the massive $35 M_\odot$ Gaussian bump population may need to reevaluate the underlying mechanisms that drive the preference for BGBHs., Comment: 19 pages, 11 figures. Submitted to PRD

Recent advancements in ultra-high-resolution unpaired image-to-image translation have aimed to mitigate the constraints imposed by limited GPU memory through patch-wise inference. Nonetheless, existing methods often compromise between the reduction of noticeable tiling artifacts and the preservation of color and hue contrast, attributed to the reliance on global image- or patch-level statistics in the instance normalization layers. In this study, we introduce a Dense Normalization (DN) layer designed to estimate pixel-level statistical moments. This approach effectively diminishes tiling artifacts while concurrently preserving local color and hue contrasts. To address the computational demands of pixel-level estimation, we further propose an efficient interpolation algorithm. Moreover, we invent a parallelism strategy that enables the DN layer to operate in a single pass. Through extensive experiments, we demonstrate that our method surpasses all existing approaches in performance. Notably, our DN layer is hyperparameter-free and can be seamlessly integrated into most unpaired image-to-image translation frameworks without necessitating retraining. Overall, our work paves the way for future exploration in handling images of arbitrary resolutions within the realm of unpaired image-to-image translation. Code is available at: https://github.com/Kaminyou/Dense-Normalization., Comment: Accepted to ECCV 2024

Primordial Black Holes (PBHs) could explain some fraction of dark matter and shed light on many areas of early-universe physics. Despite over half a century of research interest, a PBH population has so far eluded detection. The most competitive constraints on the fraction of dark matter comprised of PBHs ($f_{\rm DM}$) in the $(10^{-9}-10)M_{\odot}$ mass-ranges come from photometric microlensing and bound $f_{\rm DM}\lesssim10^{-2}-10^{-1}$. With the advent of the Roman Space Telescope with its sub-milliarcsecond (mas) astrometric capabilities and its planned Galactic Bulge Time Domain Survey (GBTDS), detecting astrometric microlensing signatures will become routine. Compared with photometric microlensing, astrometric microlensing signals are sensitive to different lens masses-distance configurations and contains different information, making it a complimentary lensing probe. At sub-mas astrometric precision, astrometric microlensing signals are typically detectable at larger lens-source separations than photometric signals, suggesting a microlensing detection channel of pure astrometric events. We use a Galactic simulation to predict the number of detectable microlensing events during the GBTDS via this pure astrometric microlensing channel. Assuming an absolute astrometric precision floor for bright stars of 0.1 mas for the GBTDS, we find that the number of detectable events peaks at $\approx 10^{3} f_{\rm DM}$ for a population of $ 1 M_{\odot}$ PBHs and tapers to $\approx 10f_{\rm DM}$ and $\approx 100f_{\rm DM}$ at $10^{-4}M_{\odot}$ and $10^{3}M_{\odot}$, respectively. Accounting for the distinguishability of PBHs from Stellar lenses, we conclude the GBTDS will be sensitive to a PBH population at $f_{\rm DM}$ down to $\approx10^{-1}-10^{-3}$ for $(10^{-1}-10^{2})M_{\odot}$ likely yielding novel PBH constraints., Comment: 21 pages, 11 figures, accepted to AAS Journals

We make forecasts for the constraining power of the 1D Wavelet Scattering Transform (WST) when used with a Lyman-$\alpha$ forest cosmology survey. Using mock simulations and a Fisher matrix, we show that there is considerable cosmological information in the scattering transform coefficients not captured by the flux power spectrum. We estimate mock covariance matrices assuming uncorrelated Gaussian pixel noise for each quasar, at a level drawn from a simple lognormal model. The extra information comes from a smaller estimated covariance in the first-order wavelet power, and from second-order wavelet coefficients which probe non-Gaussian information in the forest. Forecast constraints on cosmological parameters from the WST are more than an order of magnitude tighter than for the power spectrum, shrinking a $4D$ parameter space by a factor of $10^6$. Should these improvements be realised with DESI, inflationary running would be constrained to test common inflationary models predicting $\alpha_s = - 6\times 10^{-4}$ and neutrino mass constraints would be improved enough for a $5-\sigma$ detection of the minimal neutrino mass.

From the formation mechanisms of stars and compact objects to nuclear physics, modern astronomy frequently leverages surveys to understand populations of objects to answer fundamental questions. The population of dark and isolated compact objects in the Galaxy contains critical information related to many of these topics, but is only practically accessible via gravitational microlensing. However, photometric microlensing observables are degenerate for different types of lenses, and one can seldom classify an event as involving either a compact object or stellar lens on its own. To address this difficulty, we apply a Bayesian framework that treats lens type probabilistically and jointly with a lens population model. This method allows lens population characteristics to be inferred despite intrinsic uncertainty in the lens-class of any single event. We investigate this method's effectiveness on a simulated ground-based photometric survey in the context of characterizing a hypothetical population of primordial black holes (PBHs) with an average mass of $30 M_{\odot}$. On simulated data, our method outperforms current black hole (BH) lens identification pipelines and characterizes different subpopulations of lenses while jointly constraining the PBH contribution to dark matter to ${\approx}25$\%. Key to robust inference, our method can marginalize over population model uncertainty. We find the lower mass cutoff for stellar origin BHs, a key observable in understanding the BH mass gap, particularly difficult to infer in our simulations. This work lays the foundation for cutting-edge PBH abundance constraints to be extracted from current photometric microlensing surveys., Comment: 31 pages, 18 figures, submitted to AAS

We present new cosmological parameter constraints from the eBOSS Lyman-$\alpha$ forest survey. We use a new theoretical model and likelihood based on the PRIYA simulation suite. PRIYA is the first suite to resolve the Lyman-$\alpha$ forest in a ($120$~Mpc/h~)$^3$ volume, using a multi-fidelity emulation technique. We use PRIYA to predict Lyman-$\alpha$ forest observables with $\lesssim 1\%$ interpolation error over an $11$ dimensional ($9$ simulated, $2$ in post-processing) parameter space. We identify an internal tension within the flux power spectrum data. Once the discrepant data is removed, we find the primeval scalar spectral index measured at a pivot scale of $k_0 = 0.78$ Mpc$^{-1}$ to be $n_P = 1.009^{+0.027}_{-0.018}$ at 68\% confidence. This measurement from the Lyman-$\alpha$ forest flux power spectrum alone is in reasonable agreement with Planck, and in tension with earlier eBOSS analyses. The amplitude of matter fluctuations is $\sigma_8 = 0.733^{+0.026}_{-0.029}$ at 68\% confidence, in agreement with Dark Energy Survey weak lensing measurements and other small-scale structure probes and in tension with CMB measurements from Planck and ACT. The effective optical depth to Lyman-$\alpha$ photons from our pipeline is in good agreement with earlier high resolution measurements. We find a linear power at $z=3$ and $k = 0.009$ s/km of $\Delta_L^2 = 0.302^{+0.024}_{-0.027}$ with a slope $n_\mathrm{eff} = -2.264^{+0.026}_{-0.018}$. Our flux power spectrum only chains prefer a low level of heating during helium reionization. When we add IGM temperature data we find $n_P = 0.983\pm 0.020$ and $\sigma_8 = 0.703^{+0.023}_{-0.027}$. Our chains prefer an early and long helium reionization event, as suggested by measurements from the helium Lyman-$\alpha$ forest. In the near future we will use our pipeline to infer cosmological parameters from the DESI Lyman-$\alpha$ data., Comment: 41 pages, 13 figures, published in JCAP https://iopscience.iop.org/article/10.1088/1475-7516/2024/07/029

We present the PRIYA suite of cosmological simulations, based on the code and hydrodynamic model of the ASTRID simulation, and designed for cosmological analyses of the Lyman-$\alpha$ forest. Our simulation suite spans a $9$-dimensional parameter space, including $4$ cosmological parameters and $5$ astrophysical/thermal parameters. We have run $48$ low fidelity simulations with $1536^3$ particles in a $120$ Mpc/h box and $3$ high fidelity simulations with $3072^3$ particles in a $120$ Mpc/h box. All our simulations include a full physics model for galaxy formation, including supernova and AGN feedback, and thus also contain a realistic population of DLAs. We advance on earlier simulations suites by larger particle loads, by incorporating new physical models for patchy hydrogen and helium reionization, and by self-consistently incorporating a model for AGN feedback. We show that patchy helium reionization imprints an excess in the 1D flux power spectrum on large scales, which may allow future measurements of helium reionization bubble sizes. Simulation parameters are chosen based on a Latin hypercube design and a Gaussian process is used to interpolate to arbitrary parameter combinations. We build a multi-fidelity emulator for the 1D flux power spectrum and the mean IGM temperature. We show that our final interpolation error is $< 1\%$ and that our simulations produce a flux power spectrum converged at the percent level for $z=5.4$ - $2.2$. Our simulation suite will be used to interpret Lyman-$\alpha$ forest 1D flux power spectra from SDSS and future DESI data releases., Comment: 25 pages, 11 figures, matches version published in JCAP

We introduce MF-Box, an extended version of MFEmulator, designed as a fast surrogate for power spectra, trained using N-body simulation suites from various box sizes and particle loads. To demonstrate MF-Box's effectiveness, we design simulation suites that include low-fidelity suites (L1 and L2) at $256 \,\mathrm{Mpc}/h$ and $100 \,\mathrm{Mpc}/h$, each with $128^3$ particles, and a high-fidelity suite (HF) with $512^3$ particles at $256 \,\mathrm{Mpc}/h$, representing a higher particle load compared to the low-fidelity suites. MF-Box acts as a probabilistic resolution correction function, learning most of the cosmological dependencies from L1 and L2 simulations and rectifying resolution differences with just 3 HF simulations using a Gaussian process. MF-Box successfully emulates power spectra from our HF testing set with a relative error of $< 3\%$ up to $k \simeq 7 \,h/\mathrm{Mpc}$ at $z \in [0, 3]$, while maintaining a cost similar to our previous multi-fidelity approach, which was accurate only up to $z = 1$. The addition of an extra low-fidelity node in a smaller box significantly improves emulation accuracy for MF-Box at $k > 2 \,h/\mathrm{Mpc}$, increasing it by a factor of $10$. We conduct an error analysis of MF-Box based on computational budget, providing guidance for optimizing budget allocation per fidelity node. Our proposed MF-Box enables future surveys to efficiently combine simulation suites of varying quality, effectively expanding the range of emulation capabilities while ensuring cost efficiency., Comment: 19 pages, 16 figures. Match the MNRAS version. Data available at: https://github.com/jibanCat/MFBoxData, Code and tutorials available at: https://github.com/jibanCat/matter_emu_mfbox

We assemble the largest CIV absorption line catalogue to date, leveraging machine learning, specifically Gaussian processes, to remove the need for visual inspection for detecting CIV absorbers. The catalogue contains probabilities classifying the reliability of the absorption system within a quasar spectrum. Our training set was a sub-sample of DR7 spectra that had no detectable CIV absorption in a large visually inspected catalogue. We used Bayesian model selection to decide between our continuum model and our absorption-line models. Using a random hold-out sample of 1301 spectra from all of the 26,030 investigated spectra in DR7 CIV catalogue, we validated our pipeline and obtained an 87% classification performance score. We found good purity and completeness values, both ~80%, when a probability of ~95% is used as the threshold. Our pipeline obtained similar CIV redshifts and rest equivalent widths to our training set. Applying our algorithm to 185,425 selected quasar spectra from SDSS DR12, we produce a catalogue of 113,775 CIV doublets with at least 95% confidence. Our catalogue provides maximum a posteriori values and credible intervals for CIV redshift, column density, and Doppler velocity dispersion. We detect CIV absorption systems with a redshift range of 1.37 $\!-\!$ 5.1, including 33 systems with a redshift larger than 5 and 549 absorbers systems with a rest equivalent width greater than 2 A at more than 95% confidence. Our catalogue can be used to investigate the physical properties of the circumgalactic and intergalactic media., Comment: 18 pages, 25 figures, 3 tables

While hematoxylin and eosin (H&E) is a standard staining procedure, immunohistochemistry (IHC) staining further serves as a diagnostic and prognostic method. However, acquiring special staining results requires substantial costs. Hence, we proposed a strategy for ultra-high-resolution unpaired image-to-image translation: Kernelized Instance Normalization (KIN), which preserves local information and successfully achieves seamless stain transformation with constant GPU memory usage. Given a patch, corresponding position, and a kernel, KIN computes local statistics using convolution operation. In addition, KIN can be easily plugged into most currently developed frameworks without re-training. We demonstrate that KIN achieves state-of-the-art stain transformation by replacing instance normalization (IN) layers with KIN layers in three popular frameworks and testing on two histopathological datasets. Furthermore, we manifest the generalizability of KIN with high-resolution natural images. Finally, human evaluation and several objective metrics are used to compare the performance of different approaches. Overall, this is the first successful study for the ultra-high-resolution unpaired image-to-image translation with constant space complexity. Code is available at: https://github.com/Kaminyou/URUST, Comment: Accepted to ECCV 2022

In this work we extend our recently developed multi-fidelity emulation technique to the simulated Lyman-$\alpha$ forest flux power spectrum. Multi-fidelity emulation allows interpolation of simulation outputs between cosmological parameters using many cheap low-fidelity simulations and a few expensive high-fidelity simulations. Using a test suite of small box (30 Mpc/h) simulations, we show that multi-fidelity emulation is able to reproduce the Lyman-$\alpha$ forest flux power spectrum well, achieving an average accuracy when compared to a test suite of $0.8\%$. We further show that it has a substantially increased accuracy over single-fidelity emulators, constructed using either the high or low-fidelity simulations only. In particular, it allows the extension of an existing simulation suite to smaller scales and higher redshifts., Comment: 12 pages, 10 figures

Airway segmentation is essential for chest CT image analysis. Different from natural image segmentation, which pursues high pixel-wise accuracy, airway segmentation focuses on topology. The task is challenging not only because of its complex tree-like structure but also the severe pixel imbalance among airway branches of different generations. To tackle the problems, we present a NaviAirway method which consists of a bronchiole-sensitive loss function for airway topology preservation and an iterative training strategy for accurate model learning across different airway generations. To supplement the features of airway branches learned by the model, we distill the knowledge from numerous unlabeled chest CT images in a teacher-student manner. Experimental results show that NaviAirway outperforms existing methods, particularly in the identification of higher-generation bronchioles and robustness to new CT scans. Moreover, NaviAirway is general enough to be combined with different backbone models to significantly improve their performance. NaviAirway can generate an airway roadmap for Navigation Bronchoscopy and can also be applied to other scenarios when segmenting fine and long tubular structures in biomedical images. The code is publicly available on https://github.com/AntonotnaWang/NaviAirway.