Search

Accreting supermassive black holes (SMBHs) located at the centers of galaxies are typically surrounded by large quantities of gas and dust. The structure and evolution of this circumnuclear material can be studied at different wavelengths, from the submillimeter to the X-ray. Recent X-ray studies have shown that the covering factor of the obscuring material tends to decrease with increasing Eddington ratio, likely due to radiative feedback on dusty gas. Here we study a sample of 549 nearby ( z ≲ 0.1) hard X-ray (14–195 keV) selected nonblazar active galactic nuclei (AGN) and use the ratio between the AGN infrared and bolometric luminosity as a proxy of the covering factor. We find that, in agreement with what has been found by X-ray studies of the same sample, the covering factor decreases with increasing Eddington ratio. We also confirm previous findings that showed that obscured AGN typically have larger covering factors than unobscured sources. Finally, we find that the median covering factors of AGN located in different regions of the column density–Eddington ratio diagram are in good agreement with what would be expected from a radiation-regulated growth of SMBHs.

We present ALMA Band 6 (ν = 233 GHz, λ = 1.3 mm) continuum observations toward 68 "normal" star-forming galaxies within two Coma-like progenitor structures at z = 2.10 and 2.47, from which ISM masses are derived, providing the largest census of molecular gas mass in overdense environments at these redshifts. Our sample comprises galaxies with a stellar mass range of 1 × 109M⊙–4 × 1011M⊙ with a mean M⋆ ≈ 6 × 1010M⊙. Combining these measurements with multiwavelength observations and spectral energy distribution modeling, we characterize the gas mass fraction and the star formation efficiency, and infer the impact of the environment on galaxies' evolution. Most of our detected galaxies (≳70%) have star formation efficiencies and gas fractions similar to those found for coeval field galaxies and in agreement with the field scaling relations. However, we do find that the protoclusters contain an increased fraction of massive, gas-poor galaxies, with low gas fractions (fgas ≲ 6%–10%) and red rest-frame ultraviolet/optical colors typical of post-starburst and passive galaxies. The relatively high abundance of passive galaxies suggests an accelerated evolution of massive galaxies in protocluster environments. The large fraction of quenched galaxies in these overdense structures also implies that environmental quenching takes place during the early phases of cluster assembly, even before virialization. From our data, we derive a quenching efficiency of ϵq ≈ 0.45 and an upper limit on the quenching timescale of τq < 1 Gyr. [ABSTRACT FROM AUTHOR]

We analyze the evolution of the dual active galactic nucleus (AGN) in ESO 509-IG066, a galaxy pair located at z = 0.034 whose nuclei are separated by 11 kpc. Previous observations with XMM-Newton on this dual AGN found evidence for two moderately obscured ( ∼ 1022 cm−2) X-ray luminous ( ∼ 1043 erg s−1) nuclear sources. We present an analysis of subsequent Chandra, NuSTAR, and Swift/XRT observations that show one source has dropped in flux by a factor of 10 between 2004 and 2011, which could be explained by either an increase in the absorbing column or an intrinsic fading of the central engine, possibly due to a decrease in mass accretion. Both of these scenarios are predicted by galaxy merger simulations. The source that has dropped in flux is not detected by NuSTAR, which argues against absorption, unless it is extreme. However, new Keck/LRIS optical spectroscopy reveals a previously unreported broad Hα line that is highly unlikely to be visible under the extreme absorption scenario. We therefore conclude that the black hole in this nucleus has undergone a dramatic drop in its accretion rate. From AO-assisted near-infrared integral-field spectroscopy of the other nucleus, we find evidence that the galaxy merger is having a direct effect on the kinematics of the gas close to the nucleus of the galaxy, providing a direct observational link between the galaxy merger and the mass accretion rate onto the black hole. [ABSTRACT FROM AUTHOR]

We present a study of the average X-ray spectral properties of the sources detected by the NuSTAR extragalactic survey, comprising observations of the Extended Chandra Deep Field South (E-CDFS), Extended Groth Strip (EGS), and the Cosmic Evolution Survey (COSMOS). The sample includes 182 NuSTAR sources (64 detected at 8–24 keV), with 3–24 keV fluxes ranging between and 6 × 10−13 erg cm−2 s−1 ( erg cm−2 s−1) and redshifts in the range of . We produce composite spectra from the Chandra + NuSTAR data (, rest frame) for all the sources with redshift identifications (95%) and investigate the intrinsic, average spectra of the sources, divided into broad-line (BL) and narrow-line (NL) active galactic nuclei (AGNs), and also in different bins of X-ray column density and luminosity. The average power-law photon index for the whole sample is , flatter than the typically found for AGNs. While the spectral slope of BL and X-ray unabsorbed AGNs is consistent with the typical values (), a significant flattening is seen in NL AGNs and heavily absorbed sources ( and , respectively), likely due to the effect of absorption and to the contribution from the Compton reflection component to the high-energy flux ( keV). We find that the typical reflection fraction in our spectra is (for ), with a tentative indication of an increase of the reflection strength with X-ray column density. While there is no significant evidence for a dependence of the photon index on X-ray luminosity in our sample, we find that decreases with luminosity, with relatively high levels of reflection () for erg s−1 and for erg s−1 AGNs, assuming a fixed spectral slope of . [ABSTRACT FROM AUTHOR]

Using Chandra observations in the 2.15 deg2 COSMOS-legacy field, we present one of the most accurate measurements of the Cosmic X-ray Background (CXB) spectrum to date in the [0.3–7] keV energy band. The CXB has three distinct components: contributions from two Galactic collisional thermal plasmas at kT ∼ 0.27 and 0.07 keV and an extragalactic power law with a photon spectral index Γ = 1.45 ± 0.02. The 1 keV normalization of the extragalactic component is 10.91 ± 0.16 keV cm−2 s−1 sr−1 keV−1. Removing all X-ray-detected sources, the remaining unresolved CXB is best fit by a power law with normalization 4.18 ± 0.26 keV cm−2 s−1 sr−1 keV−1 and photon spectral index Γ = 1.57 ± 0.10. Removing faint galaxies down to leaves a hard spectrum with and a 1 keV normalization of ∼1.37 keV cm−2 s−1 sr−1 keV−1. This means that ∼91% of the observed CXB is resolved into detected X-ray sources and undetected galaxies. Unresolved sources that contribute ∼8%–9% of the total CXB show marginal evidence of being harder and possibly more obscured than resolved sources. Another ∼1% of the CXB can be attributed to still undetected star-forming galaxies and absorbed active galactic nuclei. According to these limits, we investigate a scenario where early black holes totally account for non-source CXB fraction and constrain some of their properties. In order to not exceed the remaining CXB and the accreted mass density, such a population of black holes must grow in Compton-thick envelopes with 1.6 × 1025 cm−2 and form in extremely low-metallicity environments . [ABSTRACT FROM AUTHOR]

We present the first full catalog and science results for the Nuclear Spectroscopic Telescope Array (NuSTAR) serendipitous survey. The catalog incorporates data taken during the first 40 months of NuSTAR operation, which provide ≈20 Ms of effective exposure time over 331 fields, with an areal coverage of 13 deg2, and 497 sources detected in total over the 3–24 keV energy range. There are 276 sources with spectroscopic redshifts and classifications, largely resulting from our extensive campaign of ground-based spectroscopic follow-up. We characterize the overall sample in terms of the X-ray, optical, and infrared source properties. The sample is primarily composed of active galactic nuclei (AGNs), detected over a large range in redshift from z = 0.002 to 3.4 (median of ), but also includes 16 spectroscopically confirmed Galactic sources. There is a large range in X-ray flux, from to −11, and in rest-frame 10–40 keV luminosity, from to 46, with a median of 44.1. Approximately 79% of the NuSTAR sources have lower-energy (<10 keV) X-ray counterparts from XMM-Newton, Chandra, and Swift XRT. The mid-infrared (MIR) analysis, using WISE all-sky survey data, shows that MIR AGN color selections miss a large fraction of the NuSTAR-selected AGN population, from ≈15% at the highest luminosities ( erg s−1) to ≈80% at the lowest luminosities ( erg s−1). Our optical spectroscopic analysis finds that the observed fraction of optically obscured AGNs (i.e., the type 2 fraction) is , for a well-defined subset of the 8–24 keV selected sample. This is higher, albeit at a low significance level, than the type 2 fraction measured for redshift- and luminosity-matched AGNs selected by <10 keV X-ray missions. [ABSTRACT FROM AUTHOR]

We present new Atacama Large Millimeter/submillimeter Array Band 7 (∼340 GHz) observations of the dense gas tracers HCN, HCO+, and CS in the local, single-nucleus, ultraluminous infrared galaxy IRAS 13120–5453. We find centrally enhanced HCN (4–3) emission, relative to HCO+ (4–3), but do not find evidence for radiative pumping of HCN. Considering the size of the starburst (0.5 kpc) and the estimated supernovae rate of ∼1.2 yr−1, the high HCN/HCO+ ratio can be explained by an enhanced HCN abundance as a result of mechanical heating by the supernovae, though the active galactic nucleus and winds may also contribute additional mechanical heating. The starburst size implies a high ΣIR of 4.7 × 1012L⊙ kpc−2, slightly below predictions of radiation-pressure limited starbursts. The HCN line profile has low-level wings, which we tentatively interpret as evidence for outflowing dense molecular gas. However, the dense molecular outflow seen in the HCN line wings is unlikely to escape the Galaxy and is destined to return to the nucleus and fuel future star formation. We also present modeling of Herschel observations of the H2O lines and find a nuclear dust temperature of ∼40 K. IRAS 13120–5453 has a lower dust temperature and ΣIR than is inferred for the systems termed “compact obscured nuclei (CONs)” (such as Arp 220 and Mrk 231). If IRAS 13120–5453 has undergone a CON phase, we are likely witnessing it at a time when the feedback has already inflated the nuclear ISM and diluted star formation in the starburst/active galactic nucleus core. [ABSTRACT FROM AUTHOR]

A correction to the article "Excitation Mechanisms for HCN (1-0) and HCO+ in Galaxies From the Great Observatories All-Sky Lirg Survey," in the 2015 issue is presented.

We present the X-ray spectral analysis of the 1855 extragalactic sources in the Chandra COSMOS-Legacy survey catalog having more than 30 net counts in the 0.5–7 keV band. A total of 38% of the sources are optically classified type 1 active galactic nuclei (AGNs), 60% are type 2 AGNs, and 2% are passive, low-redshift galaxies. We study the distribution of AGN photon index Γ and of the intrinsic absorption based on the sources’ optical classification: type 1 AGNs have a slightly steeper mean photon index Γ than type 2 AGNs, which, on the other hand, have average times higher than type 1 AGNs. We find that ∼15% of type 1 AGNs have cm−2, i.e., are obscured according to the X-ray spectral fitting; the vast majority of these sources have 1044 erg s−1. The existence of these objects suggests that optical and X-ray obscuration can be caused by different phenomena, the X-ray obscuration being, for example, caused by dust-free material surrounding the inner part of the nuclei. Approximately 18% of type 2 AGNs have cm−2, and most of these sources have low X-ray luminosities (L 1043 erg s−1). We expect a part of these sources to be low-accretion, unobscured AGNs lacking broad emission lines. Finally, we also find a direct proportional trend between and host-galaxy mass and star formation rate, although part of this trend is due to a redshift selection effect. [ABSTRACT FROM AUTHOR]

We present the largest high-redshift (3 < z < 6.85) sample of X-ray-selected active galactic nuclei (AGNs) on a contiguous field, using sources detected in the Chandra COSMOS-Legacy survey. The sample contains 174 sources, 87 with spectroscopic redshift and the other 87 with photometric redshift (zphot). In this work, we treat zphot as a probability-weighted sum of contributions, adding to our sample the contribution of sources with zphot < 3 but zphot probability distribution >0 at z > 3. We compute the number counts in the observed 0.5–2 keV band, finding a decline in the number of sources at z > 3 and constraining phenomenological models of the X-ray background. We compute the AGN space density at z > 3 in two different luminosity bins. At higher luminosities (logL(2–10 keV) > 44.1 erg s−1), the space density declines exponentially, dropping by a factor of ∼20 from z ∼ 3 to z ∼ 6. The observed decline is ∼80% steeper at lower luminosities (43.55 erg s−1 < logL(2–10 keV) < 44.1 erg s−1) from z ∼ 3 to z ∼ 4.5. We study the space density evolution dividing our sample into optically classified Type 1 and Type 2 AGNs. At logL(2–10 keV) > 44.1 erg s−1, unobscured and obscured objects may have different evolution with redshift, with the obscured component being three times higher at z ∼ 5. Finally, we compare our space density with predictions of quasar activation merger models, whose calibration is based on optically luminous AGNs. These models significantly overpredict the number of expected AGNs at logL (2–10 keV) > 44.1 erg s−1 with respect to our data. [ABSTRACT FROM AUTHOR]

We present near-infrared integral-field spectroscopic observations targeting Hα in eight submillimeter galaxies (SMGs) at z = 1.3–2.5 using the Very Large Telescope/Spectrograph for Integral Field Observations in the Near Infrared, obtaining significant detections for six of them. The star formation rates derived from the Hα emission are ∼100 M⊙ yr−1, which account for only ∼20%–30% of the infrared-derived values, thus suggesting that these systems are very dusty. Two of these systems present [N ii]/Hα ratios indicative of the presence of an active galactic nucleus. We mapped the spatial distribution and kinematics of the star-forming regions in these galaxies on kiloparsec scales. In general, the Hα morphologies tend to be highly irregular and/or clumpy, showing spatial extents of ∼3–11 kpc. We find evidence for significant spatial offsets, of ∼0.″1–0.″4 or 1.2–3.4 kpc, between the Hα and the continuum emission in three of the sources. Performing a kinemetry analysis, we conclude that the majority of the sample is not consistent with disk-like rotation-dominated kinematics. Instead, they tend to show irregular and/or clumpy and turbulent velocity and velocity dispersion fields. This can be interpreted as evidence for a scenario in which these extreme star formation episodes are triggered by galaxy–galaxy interactions and major mergers. In contrast to recent results for SMGs, these sources appear to follow the same relations between gas and star-forming rate densities as less luminous and/or normal star-forming galaxies. [ABSTRACT FROM AUTHOR]

We present results of five Nuclear Spectroscopic Telescope Array (NuSTAR) observations of the type 2 active galactic nucleus (AGN) in IC 751, three of which were performed simultaneously with XMM-Newton or Swift/X-Ray Telescope. We find that the nuclear X-ray source underwent a clear transition from a Compton-thick () to a Compton-thin () state on timescales of months, which makes IC 751 the first changing look AGN discovered by NuSTAR. Changes of the line of sight column density at the ∼2σ level are also found on a timescale of ∼48 hr (). From the lack of spectral variability on timescales of ∼100 ks, we infer that the varying absorber is located beyond the emission-weighted average radius of the broad-line region (BLR), and could therefore be related either to the external part of the BLR or a clumpy molecular torus. By adopting a physical torus X-ray spectral model, we are able to disentangle the column density of the non-varying absorber () from that of the varying clouds , and to constrain that of the material responsible for the reprocessed X-ray radiation (). We find evidence of significant intrinsic X-ray variability, with the flux varying by a factor of five on timescales of a few months in the 2–10 and 10–50 keV band. [ABSTRACT FROM AUTHOR]

We report the detection of a heavily obscured active galactic nucleus (AGN) in the luminous infrared galaxy (LIRG) NGC 6286 identified in a 17.5 ks Nuclear Spectroscopic Telescope Array observation. The source is in an early merging stage and was targeted as part of our ongoing NuSTAR campaign observing local luminous and ultra-luminous infrared galaxies in different merger stages. NGC 6286 is clearly detected above 10 keV and by including the quasi-simultaneous Swift/XRT and archival XMM-Newton and Chandra data, we find that the source is heavily obscured (NH ≃(0.95−1.32) × 1024 cm−2) with a column density consistent with being Compton-thick (CT, ). The AGN in NGC 6286 has a low absorption-corrected luminosity (L2−10 keV ∼ 3−20 × 1041 erg s−1) and contributes ≲1% to the energetics of the system. Because of its low luminosity, previous observations carried out in the soft X-ray band (<10 keV) and in the infrared did not notice the presence of a buried AGN. NGC 6286 has multiwavelength characteristics typical of objects with the same infrared luminosity and in the same merger stage, which might imply that there is a significant population of obscured low-luminosity AGNs in LIRGs that can only be detected by sensitive hard X-ray observations. [ABSTRACT FROM AUTHOR]

We present the catalog of optical and infrared counterparts of the ChandraCOSMOS-Legacy Survey, a 4.6 Ms Chandra program on the 2.2 deg2 of the COSMOS field, combination of 56 new overlapping observations obtained in Cycle 14 with the previous C-COSMOS survey. In this Paper we report the i, K, and 3.6 μm identifications of the 2273 X-ray point sources detected in the new Cycle 14 observations. We use the likelihood ratio technique to derive the association of optical/infrared (IR) counterparts for 97% of the X-ray sources. We also update the information for the 1743 sources detected in C-COSMOS, using new K and 3.6 μm information not available when the C-COSMOS analysis was performed. The final catalog contains 4016 X-ray sources, 97% of which have an optical/IR counterpart and a photometric redshift, while ≃54% of the sources have a spectroscopic redshift. The full catalog, including spectroscopic and photometric redshifts and optical and X-ray properties described here in detail, is available online. We study several X-ray to optical (X/O) properties: with our large statistics we put better constraints on the X/O flux ratio locus, finding a shift toward faint optical magnitudes in both soft and hard X-ray band. We confirm the existence of a correlation between X/O and the the 2–10 keV luminosity for Type 2 sources. We extend to low luminosities the analysis of the correlation between the fraction of obscured AGNs and the hard band luminosity, finding a different behavior between the optically and X-ray classified obscured fraction. [ABSTRACT FROM AUTHOR]

We present the first direct measurements of the rest-frame 10–40 keV X-ray luminosity function (XLF) of active galactic nuclei (AGNs) based on a sample of 94 sources at 0.1 < z < 3, selected at 8–24 keV energies from sources in the Nuclear Spectroscopic Telescope Array (NuSTAR) extragalactic survey program. Our results are consistent with the strong evolution of the AGN population seen in prior, lower-energy studies of the XLF. However, different models of the intrinsic distribution of absorption, which are used to correct for selection biases, give significantly different predictions for the total number of sources in our sample, leading to small, systematic differences in our binned estimates of the XLF. Adopting a model with a lower intrinsic fraction of Compton-thick sources and a larger population of sources with column densities cm−2 or a model with stronger Compton reflection component (with a relative normalization of R ∼ 2 at all luminosities) can bring extrapolations of the XLF from 2–10 keV into agreement with our NuSTAR sample. Ultimately, X-ray spectral analysis of the NuSTAR sources is required to break this degeneracy between the distribution of absorbing column densities and the strength of the Compton reflection component and thus refine our measurements of the XLF. Furthermore, the models that successfully describe the high-redshift population seen by NuSTAR tend to over-predict previous, high-energy measurements of the local XLF, indicating that there is evolution of the AGN population that is not fully captured by the current models. [ABSTRACT FROM AUTHOR]

We present new Institut de Radioastronomie Millimétrique (IRAM) 30 m spectroscopic observations of the ∼88 GHz band, including emission from the multiplet, HCN (, , and , for a sample of 58 local luminous and ultraluminous infrared galaxies from the Great Observatories All-sky LIRG Survey (GOALS). By combining our new IRAM data with literature data and Spitzer/IRS spectroscopy, we study the correspondence between these putative tracers of dense gas and the relative contribution of active galactic nuclei (AGNs) and star formation to the mid-infrared luminosity of each system. We find the HCN (1–0) emission to be enhanced in AGN-dominated systems ( / ), compared to composite and starburst-dominated systems ( / and 0.88, respectively). However, some composite and starburst systems have / ratios comparable to those of AGNs, indicating that enhanced HCN emission is not uniquely associated with energetically dominant AGNs. After removing AGN-dominated systems from the sample, we find a linear relationship (within the uncertainties) between log10() and log10(LIR), consistent with most previous findings. /LIR, typically interpreted as the dense-gas depletion time, appears to have no systematic trend with LIR for our sample of luminous and ultraluminous infrared galaxies, and has significant scatter. The galaxy-integrated and emission do not appear to have a simple interpretation in terms of the AGN dominance or the star formation rate, and are likely determined by multiple processes, including density and radiative effects. [ABSTRACT FROM AUTHOR]

We present new Nuclear Spectroscopic Telescope Array (NuSTAR ) and Chandra observations of NGC 3393, a galaxy reported to host the smallest separation dual active galactic nuclei (AGN) resolved in the X-rays. While past results suggested a 150 pc separation dual AGN, three times deeper Chandra imaging, combined with adaptive optics and radio imaging suggest a single, heavily obscured, radio-bright AGN. Using Very Large Array and Very Long Baseline Array data, we find an AGN with a two-sided jet rather than a dual AGN and that the hard X-ray, UV, optical, near-infrared, and radio emission are all from a single point source with a radius <0 .″ 2. We find that the previously reported dual AGN is most likely a spurious detection resulting from the low number of X-ray counts (<160) at 6–7 keV and Gaussian smoothing of the data on scales much smaller than the point-spread function (PSF) (0 .″ 25 versus 0 .″ 80 FWHM). We show that statistical noise in a single Chandra PSF generates spurious dual peaks of the same separation (0 .″ 55±0 .″ 07 versus 0 .″ 6) and flux ratio (39% ± 9% versus 32% counts) as the purported dual AGN. With NuSTAR, we measure a Compton-thick source ( ) with a large torus half-opening angle, which we postulate results from feedback from strong radio jets. This AGN shows a 2–10 keV intrinsic-to-observed flux ratio of ( versus ). Using simulations, we find that even the deepest Chandra observations would severely underestimate the intrinsic luminosity of NGC 3393 above , but would detect an unobscured AGN of this luminosity out to high redshift (). [ABSTRACT FROM AUTHOR]

Performing joint analysis of gene expression datasets from different experiments can present challenges brought on by multiple factors-differences in equipment, protocols, climate etc. "Cross-study normalization" is a general term for transformations aimed at eliminating such effects, thus making datasets more comparable. However, joint analysis of datasets from different species is rarely done, and there are no dedicated normalization methods for such inter-species analysis. In order to test the usefulness of cross-studies normalization methods for inter-species analysis, we first applied three cross-study normalization methods, EB, DWD and XPN, to RNA sequencing datasets from different species. We then developed a new approach to evaluate the performance of cross-study normalization in eliminating experimental effects, while also maintaining the biologically significant differences between species and conditions. Our results indicate that all normalization methods performed relatively well in the cross-species setting. We found XPN to be better at reducing experimental differences, and found EB to be better at preserving biological differences. Still, according to our in-silico experiments, in all methods it is not possible to enforce the preservation of the biological differences in the normalization process. In addition to the study above, in this work we propose a new dedicated cross-studies and cross-species normalization method. Our aim is to address the shortcoming mentioned above: in the normalization process, we wish to reduce the experimental differences while preserving the biological differences. We term our method as CSN, and base it on the performance evaluation criteria mentioned above. Repeating the same experiments, the CSN method obtained a better and more balanced conservation of biological differences within the datasets compared to existing methods. To summarize, we demonstrate the usefulness of cross-study normalization methods in the inter-species settings, and suggest a dedicated cross-study cross-species normalization method that will hopefully open the way to the development of improved normalization methods for the inter-species settings., Competing Interests: The authors have declared that no competing interests exist., (Copyright: © 2024 Feldman et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Graph convolutional networks (GCNs) are widely used in a variety of applications and can be seen as an unstructured version of standard convolutional neural networks (CNNs). As in CNNs, the computational cost of GCNs for large input graphs (such as large point clouds or meshes) can be high and inhibit the use of these networks, especially in environments with low computational resources. To ease these costs, quantization can be applied to GCNs. However, aggressive quantization of the feature maps can lead to a significant degradation in performance. On a different note, the Haar wavelet transforms are known to be one of the most effective and efficient approaches to compress signals. Therefore, instead of applying aggressive quantization to feature maps, we propose to use Haar wavelet compression and light quantization to reduce the computations involved with the network. We demonstrate that this approach surpasses aggressive feature quantization by a significant margin, for a variety of problems ranging from node classification to point cloud classification and both part and semantic segmentation.

Quantized neural networks (QNNs) are among the main approaches for deploying deep neural networks on low-resource edge devices. Training QNNs using different levels of precision throughout the network (mixed-precision quantization) typically achieves superior trade-offs between performance and computational load. However, optimizing the different precision levels of QNNs can be complicated, as the values of the bit allocations are discrete and difficult to differentiate for. Moreover, adequately accounting for the dependencies between the bit allocation of different layers is not straightforward. To meet these challenges, in this work, we propose GradFreeBits: a novel joint optimization scheme for training mixed-precision QNNs, which alternates between gradient-based optimization for the weights and gradient-free optimization for the bit allocation. Our method achieves a better or on par performance with the current state-of-the-art low-precision classification networks on CIFAR10/100 and ImageNet, semantic segmentation networks on Cityscapes, and several graph neural networks benchmarks. Furthermore, our approach can be extended to a variety of other applications involving neural networks used in conjunction with parameters that are difficult to optimize for.

Recent advancements in machine learning techniques for protein structure prediction motivate better results in its inverse problem-protein design. In this work we introduce a new graph mimetic neural network, MimNet, and show that it is possible to build a reversible architecture that solves the structure and design problems in tandem, allowing to improve protein backbone design when the structure is better estimated. We use the ProteinNet data set and show that the state of the art results in protein design can be met and even improved, given recent architectures for protein folding., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2022 Eliasof, Boesen , Haber , Keasar  and Treister .)

Major galaxy mergers are thought to play an important part in fuelling the growth of supermassive black holes 1 . However, observational support for this hypothesis is mixed, with some studies showing a correlation between merging galaxies and luminous quasars 2,3 and others showing no such association 4,5 . Recent observations have shown that a black hole is likely to become heavily obscured behind merger-driven gas and dust, even in the early stages of the merger, when the galaxies are well separated 6-8 (5 to 40 kiloparsecs). Merger simulations further suggest that such obscuration and black-hole accretion peaks in the final merger stage, when the two galactic nuclei are closely separated 9 (less than 3 kiloparsecs). Resolving this final stage requires a combination of high-spatial-resolution infrared imaging and high-sensitivity hard-X-ray observations to detect highly obscured sources. However, large numbers of obscured luminous accreting supermassive black holes have been recently detected nearby (distances below 250 megaparsecs) in X-ray observations 10 . Here we report high-resolution infrared observations of hard-X-ray-selected black holes and the discovery of obscured nuclear mergers, the parent populations of supermassive-black-hole mergers. We find that obscured luminous black holes (bolometric luminosity higher than 2 × 10 44 ergs per second) show a significant (P < 0.001) excess of late-stage nuclear mergers (17.6 per cent) compared to a sample of inactive galaxies with matching stellar masses and star formation rates (1.1 per cent), in agreement with theoretical predictions. Using hydrodynamic simulations, we confirm that the excess of nuclear mergers is indeed strongest for gas-rich major-merger hosts of obscured luminous black holes in this final stage.

Theoretical models and numerical simulations have established a framework of galaxy evolution in which galaxies merge and create dual supermassive black holes (with separations of one to ten kiloparsecs), which eventually sink into the centre of the merger remnant, emit gravitational waves and coalesce. The merger also triggers star formation and supermassive black hole growth, and gas outflows regulate the stellar content 1-3 . Although this theoretical picture is supported by recent observations of starburst-driven and supermassive black hole-driven outflows 4-6 , it remains unclear how these outflows interact with the interstellar medium. Furthermore, the relative contributions of star formation and black hole activity to galactic feedback remain unknown 7-9 . Here we report observations of dual outflows in the central region of the prototypical merger NGC 6240. We find a black-hole-driven outflow of [O III] to the northeast and a starburst-driven outflow of Hα to the northwest. The orientations and positions of the outflows allow us to isolate them spatially and study their properties independently. We estimate mass outflow rates of 10 and 75 solar masses per year for the Hα bubble and the [O III] cone, respectively. Their combined mass outflow is comparable to the star formation rate 10 , suggesting that negative feedback on star formation is occurring.