Boudewijn F. Roukema’s research while affiliated with Nicolaus Copernicus University and other places

Classification of spectra (1) and anomaly detection (2) are fundamental steps to guarantee the highest accuracy in redshift measurements (3) in modern all-sky spectroscopic surveys. We introduce a new Galaxy Spectra Neural Network (GaSNet-III) model that takes advantage of generative neural networks to perform these three tasks at once with very high efficiency. We use two different generative networks, an autoencoder-like network and U-Net, to reconstruct the rest-frame spectrum (after redshifting). The autoencoder-like network operates similarly to the classical PCA, learning templates (eigenspectra) from the training set and returning modeling parameters. The U-Net, in contrast, functions as an end-to-end model and shows an advantage in noise reduction. By reconstructing spectra, we can achieve classification, redshift estimation, and anomaly detection in the same framework. Each rest-frame reconstructed spectrum is extended to the UV and a small part of the infrared (covering the blueshift of stars). Owing to the high computational efficiency of deep learning, we scan the chi-squared value for the entire type and redshift space and find the best-fitting point. Our results show that generative networks can achieve accuracy comparable to the classical PCA methods in spectral modeling with higher efficiency, especially achieving an average of $>98\%$ classification across all classes ($>99.9\%$ for star), and $>99\%$ (stars), $>98\%$ (galaxies) and $>93\%$ (quasars) redshift accuracy under cosmology research requirements. By comparing different peaks of chi-squared curves, we define the ``robustness'' in the scanned space, offering a method to identify potential ``anomalous'' spectra. Our approach provides an accurate and high-efficiency spectrum modeling tool for handling the vast data volumes from future spectroscopic sky surveys.

The size and complexity reached by the large sky spectroscopic surveys require efficient, accurate, and flexible automated tools for data analysis and science exploitation. We present the Galaxy Spectra Network/GaSNet-II, a supervised multinetwork deep learning tool for spectra classification and redshift prediction. GaSNet-II can be trained to identify a customized number of classes and optimize the redshift predictions. Redshift errors are determined via an ensemble/pseudo-Monte Carlo test obtained by randomizing the weights of the network-of-networks structure. As a demonstration of the capability of GaSNet-II, we use 260k Sloan Digital Sky Survey spectra from Data Release 16, separated into 13 classes including 140k galactic, and 120k extragalactic objects. GaSNet-II achieves 92.4 per cent average classification accuracy over the 13 classes and mean redshift errors of approximately 0.23 per cent for galaxies and 2.1 per cent for quasars. We further train/test the pipeline on a sample of 200k 4MOST (4-metre Multi-Object Spectroscopic Telescope) mock spectra and 21k publicly released DESI (Dark Energy Spectroscopic Instrument) spectra. On 4MOST mock data, we reach 93.4 per cent accuracy in 10-class classification and mean redshift error of 0.55 per cent for galaxies and 0.3 per cent for active galactic nuclei. On DESI data, we reach 96 per cent accuracy in (star/galaxy/quasar only) classification and mean redshift error of 2.8 per cent for galaxies and 4.8 per cent for quasars, despite the small sample size available. GaSNet-II can process ∼40k spectra in less than one minute, on a normal Desktop GPU. This makes the pipeline particularly suitable for real-time analyses and feedback loops for optimization of Stage-IV survey observations.

The Rubin Observatory’s 10-year Legacy Survey of Space and Time will observe near to 20 billion galaxies. For each galaxy the properties can be inferred. Approximately 105 galaxies observed per year will contain Type Ia supernovae (SNe), allowing SN host-galaxy properties to be calculated on a large scale. Measuring the properties of SN host-galaxies serves two main purposes. The first is that there are known correlations between host-galaxy type and supernova type, which can be used to aid in the classification of SNe. Secondly, Type Ia SNe exhibit correlations between host-galaxy properties and the peak luminosities of the SNe, which has implications for their use as standardisable candles in cosmology. We have used simulations to quantify the improvement in host-galaxy stellar mass (M*) measurements when supplementing photometry from Rubin with spectroscopy from the 4-metre Multi-Object Spectroscopic Telescope (4MOST) instrument. We provide results in the form of expected uncertainties in M* for galaxies with 0.1 < z < 0.9 and 18 < rAB < 25. We show that for galaxies mag 22 and brighter, combining Rubin and 4MOST data reduces the uncertainty measurements of galaxy M* by more than a factor of 2 compared with Rubin data alone. This applies for elliptical and Sc type hosts. We demonstrate that the reduced uncertainties in M* lead to an improvement of 7 per cent in the precision of the ‘mass step’ correction. We expect our improved measurements of host-galaxy properties to aid in the photometric classification of SNe observed by Rubin.

The Rubin Observatory's 10-year Legacy Survey of Space and Time will observe near to 20 billion galaxies. For each galaxy the properties can be inferred. Approximately $10^5$ galaxies observed per year will contain Type Ia supernovae (SNe), allowing SN host-galaxy properties to be calculated on a large scale. Measuring the properties of SN host-galaxies serves two main purposes. The first is that there are known correlations between host-galaxy type and supernova type, which can be used to aid in the classification of SNe. Secondly, Type Ia SNe exhibit correlations between host-galaxy properties and the peak luminosities of the SNe, which has implications for their use as standardisable candles in cosmology. We have used simulations to quantify the improvement in host-galaxy stellar mass ($M_\ast$) measurements when supplementing photometry from Rubin with spectroscopy from the 4-metre Multi-Object Spectroscopic Telescope (4MOST) instrument. We provide results in the form of expected uncertainties in $M_\ast$ for galaxies with 0.1 < z < 0.9 and 18 < $r_{AB}$ < 25. We show that for galaxies mag 22 and brighter, combining Rubin and 4MOST data reduces the uncertainty measurements of galaxy $M_\ast$ by more than a factor of 2 compared with Rubin data alone. This applies for elliptical and Sc type hosts. We demonstrate that the reduced uncertainties in $M_\ast$ lead to an improvement of 7\% in the precision of the "mass step" correction. We expect our improved measurements of host-galaxy properties to aid in the photometric classification of SNe observed by Rubin.

Curved-spacetime geometric-optics maps derived from a deep photometric survey should contain information about the three-dimensional matter distribution and thus about cosmic voids in the survey, despite projection effects. We explore to what degree sky-plane geometric-optics maps can reveal the presence of intrinsic three-dimensional voids. We carry out a cosmological N-body simulation and place it further than a gigaparsec from the observer, at redshift 0.5. We infer three-dimensional void structures using the watershed algorithm. Independently, we calculate a surface overdensity map and maps of weak gravitational lensing and geometric-optics scalars. We propose and implement a heuristic algorithm for detecting (projected) radial void profiles from these maps. We find in our simulation that given the sky-plane centres of the three-dimensional watershed-detected voids, there is significant evidence of finding corresponding void centres in the surface overdensity Σ, the averaged weak-lensing tangential shear $\overline{{\gamma }_{\perp }}$, the Sachs expansion θ, and the Sachs shear modulus |σ|. Recovering the centres of the three-dimensional voids from the sky-plane information alone is significant given the Sachs expansion θ, or the Sachs shear |σ|, mildly significant given the weak-lensing shear $\overline{{\gamma }_{\perp }}$, and not significant for the surface overdensity Σ. Void radii are uncorrelated between three-dimensional and two-dimensional voids; our algorithm is not designed to distinguish voids that are nearly concentric in projection. This investigation shows preliminary evidence encouraging observational studies of gravitational lensing through individual voids, either blind or with spectroscopic/photometric redshifts. The former case – blind searches – should generate falsifiable predictions of intrinsic three-dimensional void centres.

Curved-spacetime geometric-optics maps derived from a deep photometric survey should contain information about the three-dimensional matter distribution and thus about cosmic voids in the survey, despite projection effects. We explore to what degree sky-plane geometric-optics maps can reveal the presence of intrinsic three-dimensional voids. We carry out a cosmological N-body simulation and place it further than a gigaparsec from the observer, at redshift 0.5. We infer three-dimensional void structures using the watershed algorithm. Independently, we calculate a surface overdensity map and maps of weak gravitational lensing and geometric-optics scalars. We propose and implement a heuristic algorithm for detecting (projected) radial void profiles from these maps. We find in our simulation that given the sky-plane centres of the three-dimensional watershed-detected voids, there is significant evidence of correlated void centres in the surface overdensity $\Sigma$, the averaged weak-lensing tangential shear $\overline{\gamma_\perp}$, the Sachs expansion $\theta$, and the Sachs shear modulus $\lvert\sigma\rvert$. Recovering the centres of the three-dimensional voids from the sky-plane information alone is significant given the weak-lensing shear $\overline{\gamma_\perp}$, the Sachs expansion $\theta$, or the Sachs shear $\lvert\sigma\rvert$, but not significant for the surface overdensity $\Sigma$. Void radii are uncorrelated between three-dimensional and two-dimensional voids; our algorithm is not designed to distinguish voids that are nearly concentric in projection. This investigation shows preliminary evidence encouraging observational studies of gravitational lensing through individual voids, either blind or with spectroscopic/photometric redshifts. The former case - blind searches - should generate falsifiable predictions of intrinsic three-dimensional void centres.

Several software packages for relativistic cosmological simulations that do not fully implement the Einstein equation have recently been developed. Two of the free-licensed ones are inhomog and gevolution. A key question is whether globally emergent volume evolution that is faster than that of a Friedmannian reference model results from the averaged effects of structure formation. Checking that emergent volume evolution is correctly modelled by the packages is thus needed. We numerically replace the software's default random realisation of initial seed fluctuations by a fluctuation of spatially constant amplitude in a simulation's initial conditions. The average volume evolution of the perturbed model should follow that of a Friedmannian expansion history that corresponds to the original Friedmannian reference solution modified by the insertion of the spatially constant perturbation. We derive the equations that convert from the perturbed reference solution to the effective solution. We find that inhomog allows emergent volume evolution correctly at first order through to the current epoch. For initial conditions with a resolution of N = 128^3 particles and an initial non-zero extrinsic curvature invariant I_i = 0.001, inhomog matches an exact Friedmannian solution to -0.0058% (Einstein-de Sitter, EdS) or -0.0033% (LCDM). We find that gevolution models the decaying mode to fair accuracy, and excludes the growing mode by construction. For N = 128^3 and an initial scalar potential Phi = 0.001, gevolution is accurate for the decaying mode to 0.012% (EdS) or 0.013% (LCDM). We conclude that this special case of an exact non-linear solution for a perturbed Friedmannian model provides a robust calibration for relativistic cosmological simulations.

Several software packages for relativistic cosmological simulations that do not fully implement the Einstein equation have recently been developed. Two of the free-licensed ones are inhomog and gevolution. A key question is whether globally emergent volume evolution that is faster than that of a Friedmannian reference model results from the averaged effects of structure formation. Checking that emergent volume evolution is correctly modelled by the packages is thus needed. We numerically replace the software's default random realisation of initial seed fluctuations by a fluctuation of spatially constant amplitude in a simulation's initial conditions. The average volume evolution of the perturbed model should follow that of a Friedmannian expansion history that corresponds to the original Friedmannian reference solution modified by the insertion of the spatially constant perturbation. We find that inhomog allows emergent volume evolution correctly at first order through to the current epoch. For initial conditions with a resolution of $N = 128^3$ particles and an initial non-zero extrinsic curvature invariant $I_i = 0.001$, inhomog matches an exact Friedmannian solution to -0.00576% (Einstein-de Sitter, EdS) or -0.00326% (LCDM). We find that gevolution models the decaying mode to fair accuracy, and excludes the growing mode by construction. For $N = 128^3$ and an initial scalar potential $\Phi$ = 0.001, gevolution is accurate for the decaying mode to 0.0125% (EdS) or 0.0125% (LCDM). We conclude that this special case of an exact non-linear solution for a perturbed Friedmannian model provides a robust calibration for relativistic cosmological simulations.

The noise in daily infection counts of an epidemic should be super-Poissonian due to intrinsic epidemiological and administrative clustering. Here, we use this clustering to classify the official national SARS-CoV-2 daily infection counts and check for infection counts that are unusually anti-clustered. We adopt a one-parameter model of $\phi _i^{\prime}$ϕi′ infections per cluster, dividing any daily count ni into $n_i/ _i^{\prime}$ni/ϕi′ ‘clusters’, for ‘country’ i. We assume that ${n_i}/\phi _i^{\prime}$ni/ϕi′ on a given day j is drawn from a Poisson distribution whose mean is robustly estimated from the four neighbouring days, and calculate the inferred Poisson probability $P_{ij}^{\prime}$Pij′ of the observation. The $P_{ij}^{\prime}$Pij′ values should be uniformly distributed. We find the value $\phi_i$ϕi that minimises the Kolmogorov–Smirnov distance from a uniform distribution. We investigate the (ϕi, Ni) distribution, for total infection count Ni. We consider consecutive count sequences above a threshold of 50 daily infections. We find that most of the daily infection count sequences are inconsistent with a Poissonian model. Most are found to be consistent with the ϕi model. The 28-, 14- and 7-day least noisy sequences for several countries are best modelled as sub-Poissonian, suggesting a distinct epidemiological family. The 28-day least noisy sequence of Algeria has a preferred model that is strongly sub-Poissonian, with $\phi _i^{28} < 0.1$ϕi28

Voids may affect galaxy formation via weakening mass infall or increasing disk sizes, which could potentially play a role in the formation of giant low surface brightness galaxies (LSBGs). If a dark matter halo forms at the potential hill corresponding to a void of the cosmic web, which we denote the “elaphrocentre” in contrast to a barycentre, then the elaphrocentre should weaken the infall rate to the halo when compared to infall rates towards barycentres. We investigate this hypothesis numerically. We present a complete software pipeline to simulate galaxy formation, starting from a power spectrum of initial perturbations and an N-body simulation through to merger-history-tree based mass infall histories. The pipeline is built from well-established, free-licensed cosmological software packages, and aims at highly portable long-term reproducibility. We find that the elaphrocentric accelerations tending to oppose mass infall are modest. We do not find evidence of location in a void or elaphrocentric position weakening mass infall towards a galaxy. However, we find indirect evidence of voids influencing galaxy formation: while void galaxies are of lower mass compared to galaxies in high density environments, their spin parameters are typically higher. For a fixed mass, the implied disk scale length would be greater. Tangential accelerations in voids are found to be high and might significantly contribute to the higher spin parameters. We find significantly later formation epochs for void galaxies; this should give lower matter densities and may imply lower surface densities of disk galaxies. Thus, void galaxies have higher spin parameters and later formation epochs; both are factors that may increase the probability of forming LSBGs in voids.