Search

This paper reports observations of a 22 GHz water maser `superburst' in the G25.65+1.05 massive star forming region, conducted in response to an alert from the Maser Monitoring Organisation (M2O). Very long baseline interferometry (VLBI) observations using the European VLBI Network (EVN) recorded a maser flux density of $1.2 \times 10^{4}$ Jy. The superburst was investigated in the spectral, structural and temporal domains and its cause was determined to be an increase in maser path length generated by the superposition of multiple maser emitting regions aligning in the line of sight to the observer. This conclusion was based on the location of the bursting maser in the context of the star forming region, its complex structure, and its rapid onset and decay.

Attention mechanisms in deep learning architectures have often been used as a means of transparency and, as such, to shed light on the inner workings of the architectures. Recently, there has been a growing interest in whether or not this assumption is correct. In this paper we investigate the interpretability of multi-head attention in abstractive summarization, a sequence-to-sequence task for which attention does not have an intuitive alignment role, such as in machine translation. We first introduce three metrics to gain insight in the focus of attention heads and observe that these heads specialize towards relative positions, specific part-of-speech tags, and named entities. However, we also find that ablating and pruning these heads does not lead to a significant drop in performance, indicating redundancy. By replacing the softmax activation functions with sparsemax activation functions, we find that attention heads behave seemingly more transparent: we can ablate fewer heads and heads score higher on our interpretability metrics. However, if we apply pruning to the sparsemax model we find that we can prune even more heads, raising the question whether enforced sparsity actually improves transparency. Finally, we find that relative positions heads seem integral to summarization performance and persistently remain after pruning.

We report the detection of new 12.178, 12.229, 20.347, and 23.121 GHz methanol masers in the massive star-forming region G358.93-0.03, which are flaring on similarly short timescales (days) as the 6.668 GHz methanol masers also associated with this source. The brightest 12.178 GHz channel increased by a factor of over 700 in just 50 d. The masers found in the 12.229 and 20.347 GHz methanol transitions are the first ever reported and this is only the fourth object to exhibit associated 23.121 GHz methanol masers. The 12.178 GHz methanol maser emission appears to have a higher flux density than that of the 6.668 GHz emission, which is unusual. No associated near-infrared flare counterpart was found, suggesting that the energy source of the flare is deeply embedded., Comment: 9 pages, 3 figures, 2 tables, Published in MNRAS

A total of 188 high-mass outflows have been identified from a sample of 694 clumps from the Millimetre Astronomy Legacy Team 90 GHz survey, representing a detection rate of approximately 27%. The detection rate of outflows increases from the protostellar stage to the H II stage, but decreases again at the photodissociation (PDR) stage suggesting that outflows are being switched off during the PDR stage. An intimate relationship is found between outflow action and the presence of masers, and water masers appear together with 6.7 GHz methanol masers. Comparing the infall detection rate of clumps with and without outflows, we find that outflow candidates have a lower infall detection rate. Finally, we find that outflow action has some influence on the local environment and the clump itself, and this influence decreases with increasing evolutionary time as the outflow action ceases., Comment: 10 pages, 11 figures, MNRAS, Accepted

Learning algorithms become more powerful, often at the cost of increased complexity. In response, the demand for algorithms to be transparent is growing. In NLP tasks, attention distributions learned by attention-based deep learning models are used to gain insights in the models' behavior. To which extent is this perspective valid for all NLP tasks? We investigate whether distributions calculated by different attention heads in a transformer architecture can be used to improve transparency in the task of abstractive summarization. To this end, we present both a qualitative and quantitative analysis to investigate the behavior of the attention heads. We show that some attention heads indeed specialize towards syntactically and semantically distinct input. We propose an approach to evaluate to which extent the Transformer model relies on specifically learned attention distributions. We also discuss what this implies for using attention distributions as a means of transparency., Comment: To appear at FACTS-IR 2019, SIGIR

We present a detailed comparison of two types of sequence to sequence models trained to conduct a compositional task. The models are architecturally identical at inference time, but differ in the way that they are trained: our baseline model is trained with a task-success signal only, while the other model receives additional supervision on its attention mechanism (Attentive Guidance), which has shown to be an effective method for encouraging more compositional solutions (Hupkes et al.,2019). We first confirm that the models with attentive guidance indeed infer more compositional solutions than the baseline, by training them on the lookup table task presented by Li\v{s}ka et al. (2019). We then do an in-depth analysis of the structural differences between the two model types, focusing in particular on the organisation of the parameter space and the hidden layer activations and find noticeable differences in both these aspects. Guided networks focus more on the components of the input rather than the sequence as a whole and develop small functional groups of neurons with specific purposes that use their gates more selectively. Results from parameter heat maps, component swapping and graph analysis also indicate that guided networks exhibit a more modular structure with a small number of specialized, strongly connected neurons., Comment: To appear at BlackboxNLP 2019, ACL

A multi-wavelength analysis of the large Galactic infrared bubble N 24 is has been presented in this paper in order to investigate the molecular and star formation environment around expanding HII regions. Using archival data from Herschel and ATLASGAL, the distribution and physical properties of the dust over the entire bubble are studied. Twenty three dense clumps are identified using the Clumpfind2d algorithm with sizes and masses in the range 0.65-1.73 pc and 600-16300 Msun, respectively. To analyse the molecular environment in N 24, observations of NH3 (1,1) and (2,2) were carried out using the Nanshan 26m radio telescope. Analysis of the kinetic temperature and gravitational stability of these clumps suggests gravitational collapse in several of them. The mass-size distributions of the clumps and the presence of massive young protostars indicate that the shell of N 24 is a region of ongoing massive star formation. The compatibility of the dynamical and fragmentation time-scales and the overabundance of YSOs and clumps on the rim suggest that the collect and collapse mechanism is in play at the boundary of the bubble, but the existence of the IRDC at the edge of bubble indicates that radiation-driven implosion mechanism may also have played a role there., Comment: Accepted for publication in Monthly Notices of the Royal Astronomical Society

The formaldehyde $\rm H_2CO(1_{10} - 1_{11})$ absorption line and H$110\alpha$ radio recombination line (RRL) have been observed toward the Aquila Molecular Cloud using the Nanshan 25 m telescope operated by the Xinjiang Astronomical Observatory CAS. These first observations of the $\rm H_2CO$ $(1_{10} - 1_{11})$ absorption line determine the extent of the molecular regions that are affected by the ongoing star formation in the Aquila molecular complex and show some of the dynamic properties. The distribution of the excitation temperature $T_{ex}$ for $\rm H_2CO$ identifies the two known star formation regions W40 and Serpens South as well as a smaller new region Serpens 3. The intensity and velocity distributions of $\rm H_2CO$ and $\rm ^{13}CO(1-0)$ do not agree well with each other, which confirms that the $\rm H_2CO$ absorption structure is mostly determined by the excitation of the molecules resulting from the star formation rather than by the availability of molecular material as represented by the distribution. Some velocity-coherent linear $\rm ^{13}CO(1-0)$ structures have been identified in velocity channel maps of $\rm H_2CO$ and it is found that the three star formation regions lie on the intersect points of filaments. The $\rm H110\alpha$ emission is found only at the location of the W40 H II region and spectral profile indicates a redshifted spherical outflow structure in the outskirts of the H II region. Sensitive mapping of $\rm H_2CO$ absorption of the Aquila Complex has correctly identified the locations of star-formation activity in complex molecular clouds and the spectral profiles reveal the dominant velocity components and may identify the presence of outflows., Comment: Accepted by ApJ

In a rare and spectacular display, two well-known massive star forming regions, W49N and G25.65+1.05, recently underwent maser 'super burst' - their fluxes suddenly increasing above 30,000 and 18,000 Jy, respectively, reaching several orders of magnitude above their usual values. In quick-response, ToO observations with the EVN, VLBA and KaVA were obtained constituting a 4 week campaign - producing a high-cadence multi-epoch VLBI investigation of the maser emission. The combination of high-resolution, polarisation and flux monitoring during the burst provides one of the best accounts, to date, of the maser super burst phenomenon, aiding their use as astrophysical tools. These proceedings contain the preliminary results of our campaign.

Active galactic nuclei (AGN) accreting at rates close to the Eddington limit can host radiatively driven mildly relativistic outflows. Some of these X-ray absorbing but powerful outflows may produce strong shocks resulting in a significant non-thermal emission. This outflow-driven radio emission may be detectable in the radio-quiet quasar PDS 456 since it has a bolometric luminosity reaching the Eddington limit and a relativistic wide-aperture X-ray outflow with a kinetic power high enough to quench the star formation in its host galaxy. To investigate this possibility, we performed very-long-baseline interferometric (VLBI) observations of the quasar with the European VLBI Network (EVN) at 5 GHz. The EVN image with the full resolution reveals two faint and diffuse radio components with a projected separation of about 20 pc and an average brightness temperature of around two million Kelvin. In relation to the optical sub-mas-accuracy position measured by the Gaia mission, the two components are very likely on opposite sides of an undetected radio core. The VLBI structure at the deca-pc scale can thus be either a young jet or a bidirectional radio-emitting outflow, launched in the vicinity of a strongly accreting central engine. Two diffuse components at the hecto-pc scale, likely the relic radio emission from the past AGN activity, are tentatively detected on each side in the low-resolution EVN image., Comment: 6 pages, 2 figures, 1 table. Accepted for publication in MNRAS

An unbiased search of molecular outflows within the region of the COHRS survey has identified 157 high-mass outflows from a sample of 770 ATLASGAL clumps with a detection rate of 20\%. The detection rate of outflows increases for clumps with higher M$_{clump}$, L$_{bol}$, L$_{bol}$/M$_{clump}$, N$_{H_{2}}$, and T$_{dust}$ compared to the clumps with no outflow. The detection rates of the outflow increases from protostellar (8\%) to YSO clump (17\%) and to MSF clump (29\%). The detection rate 26\% for quiescent clump is preliminary, because the sample of quiescent clumps is small. A statistical relation between the outflow and clump masses for our sample is $\log(M_{out}/M_{\bigodot}) = (-1.1\pm0.21) + (0.9\pm0.07)\log(M_{clump}/M_{\bigodot})$. The detection rate of outflows and the outflow mass-loss rate show an increase with increasing M$_{clump}$, L$_{bol}$, N$_{H_{2}}$, and T$_{dust}$, which indicates that clumps with outflow with higher parameter values are at a more advanced evolutionary stage. The outflow mechanical force increases with increasing bolometric luminosities. No clear evidence has yet been found that higher mass outflows have different launching conditions than low-mass outflows., Comment: 16 pages, 20 figures, accepted for publication in ApJ

NGC 4945 is one of the nearest (~3.8 Mpc; 1" ~ 19 pc) starburst galaxies. ALMA band 3 (3--4\,mm) observations of HCN, HCO+, CS, C3H2, SiO, HCO, and CH3C2H were carried out with ~2" resolution. The lines reveal a rotating nuclear disk of projected size 10" x 2" with position angle ~45 deg, inclination ~75 deg and an unresolved bright central core of size <2.5". The continuum source (mostly free-free radiation) is more compact than the nuclear disk by a linear factor of two but shows the same position angle and is centered 0.39" +_ 0.14" northeast of the nuclear accretion disk defined by H2O maser emission. Outside the nuclear disk, both HCN and CS delineate molecular arms on opposite sides of the dynamical center. These are connected by a (deprojected) 0.6 kpc sized molecular bridge, likely a dense gaseous bar seen almost ends-on, shifting gas from the front and back side into the nuclear disk. Modeling this nuclear disk located farther inside <100 pc) with tilted rings indicates a coplanar outflow reaching a characteristic deprojectd velocity of ~50 km/s. All our molecular lines, with the notable exception of CH3C2H, show significant absorption near the systemic velocity (~571 km/s), within a range of ~500-660 km/s. Apparently, only molecular transitions with low critical H2-density do not show absorption. The velocity field of the nuclear disk, derived from CH3C2H, provides evidence for rigid rotation in the inner few arcseconds and a dynamical mass of M = (2.1+_0.2) x 10^8 Mo inside a galactocentric radius of 2.45", with a significantly flattened rotation curve farther out. Velocity integrated line intensity maps with most pronounced absorption show molecular peak positions up to 1.5" southwest of the continuum peak, presumably due to absorption, which appears to be most severe slightly northeast of the nuclear maser disk., Comment: 31 pages, 29 figures, 6 tables, accepted for publication in A&A; for the draft showing figures with full resolution, see https://gongyan2444.github.io/pdf/Henkel-NGC4945.pdf

We present the first VLBI observations of a Galactic water maser (in Chepeus A) made with a very long baseline interferometric array involving the RadioAstron Earth-orbiting satellite station as one of its elements. We detected two distinct components at -16.9 and 0.6 km/s with a fringe spacing of 66 microarcseconds. In total power, the 0.6 km/s component appears to be a single Gaussian component of strength 580 Jy and width of 0.7 km/s. Single-telescope monitoring showed that its lifetime was only 8~months. The absence of a Zeeman pattern implies the longitudinal magnetic field component is weaker than 120 mG. The space-Earth cross power spectrum shows two unresolved components smaller than 15 microarcseconds, corresponding to a linear scale of 1.6 x 10^11 cm, about the diameter of the Sun, for a distance of 700 pc, separated by 0.54 km/s in velocity and by 160 +/-35 microarcseconds in angle. This is the smallest angular structure ever observed in a Galactic maser. The brightness temperatures are greater than 2 x 10^14K, and the line widths are 0.5 km/s. Most of the flux (about 87%) is contained in a halo of angular size of 400 +/- 150 microarcseconds. This structure is associated with the compact HII region HW3diii. We have probably picked up the most prominent peaks in the angular size range of our interferometer. We discuss three dynamical models: (1) Keplerian motion around a central object, (2) two chance overlapping clouds, and (3) vortices caused by flow around an obstacle (i.e., von Karman vortex street) with Strouhal number of about~0.3., Comment: 15 pages, 9 figures. Accepted for publication in ApJ, February 16, 2018

Observations of the masers in the course of RadioAstron mission yielded detections of fringes for a number of sources in both water and hydroxyl maser transitions. Several sources display numerous ultra-compact details. This proves that implementation of the space VLBI technique for maser studies is possible technically and is not always prevented by the interstellar scattering, maser beaming and other effects related to formation, transfer, and detection of the cosmic maser emission. For the first time, cosmic water maser emission was detected with projected baselines exceeding Earth Diameter. It was detected in a number of star-forming regions in the Galaxy and megamaser galaxies NGC 4258 and NGC 3079. RadioAstron observations provided the absolute record of the angular resolution in astronomy. Fringes from the NGC 4258 megamaser were detected on baseline exceeding 25 Earth Diameters. This means that the angular resolution sufficient to measure the parallax of the water maser source in the nearby galaxy LMC was directly achieved in the cosmic maser observations. Very compact features with angular sizes about 20 microarcsec have been detected in star-forming regions of our Galaxy. Corresponding linear sizes are about 5-10 million kilometers. So, the major step from milli- to micro-arcsecond resolution in maser studies is done in the RadioAstron mission. The existence of the features with extremely small angular sizes is established. Further implementations of the space-VLBI maser instrument for studies of the nature of cosmic objects, studies of the interaction of extremely high radiation field with molecular material and studies of the matter on the line of sight are planned., Comment: To be published in Astrophysical Masers: Unlocking the Mysteries of the Universe, IAU Symposium 336, 2017

The RadioAstron space-VLBI mission has successfully detected extragalactic H2O MegaMaser emission regions at very long Earth to space baselines ranging between 1.4 and 26.7 Earth Diameters (ED). The preliminary results for two galaxies, NGC3079 and NGC4258, at baselines longer than one ED indicate masering environments and excitation conditions in these galaxies that are distinctly different. Further observations of NGC4258 at longer baselines will reveal more of the physics of individual emission regions., Comment: To be published in Astrophysical Masers: Unlocking the Mysteries of the Universe, IAU Symposium 336, 2017

Deep galaxy surveys have revealed that the global star formation rate (SFR) density in the Universe peaks at 1 < z < 2 and sharply declines towards z = 0. But a clear picture of the underlying processes, in particular the evolution of cold atomic (~100 K) and molecular gas phases, that drive such a strong evolution is yet to emerge. MALS is designed to use MeerKAT's L- and UHF-band receivers to carry out the most sensitive (N(HI)>10$^{19}$ cm$^{-2}$) dust-unbiased search of intervening HI 21-cm and OH 18-cm absorption lines at 0 < z < 2. This will provide reliable measurements of the evolution of cold atomic and molecular gas cross-sections of galaxies, and unravel the processes driving the steep evolution in the SFR density. The large sample of HI and OH absorbers obtained from the survey will (i) lead to tightest constraints on the fundamental constants of physics, and (ii) be ideally suited to probe the evolution of magnetic fields in disks of galaxies via Zeeman Splitting or Rotation Measure synthesis. The survey will also provide an unbiased census of HI and OH absorbers, i.e. cold gas associated with powerful AGNs (>10$^{24}$ W Hz$^{-1}$) at 0 < z < 2, and will simultaneously deliver a blind HI and OH emission line survey, and radio continuum survey. Here, we describe the MALS survey design, observing plan and the science issues to be addressed under various science themes., Comment: 16 pages, 3 figures Accepted for publication, Proceedings of Science, Workshop on "MeerKAT Science: On the Pathway to the SKA", held in Stellenbosch 25-27 May, 2016

We have observed 104 gravitationally-lensed quasars at $z\sim1-4$ with Herschel/SPIRE, the largest such sample ever studied. By targeting gravitational lenses, we probe intrinsic far-infrared (FIR) luminosities and star formation rates (SFRs) more typical of the population than the extremely luminous sources that are otherwise accessible. We detect 72 objects with Herschel/SPIRE and find 66 percent (69 sources) of the sample have spectral energy distributions (SEDs) characteristic of dust emission. For 53 objects with sufficiently constrained SEDs, we find a median effective dust temperature of $38^{+12}_{-5}$ K. By applying the radio-infrared correlation, we find no evidence for an FIR excess which is consistent with star-formation-heated dust. We derive a median magnification-corrected FIR luminosity of $3.6^{+4.8}_{-2.4}~\times 10^{11}~{\rm L_{\odot}}$ and median SFR of $120^{+160}_{-80}~{\rm M_{\odot}~yr^{-1}}$ for 94 quasars with redshifts. We find $\sim10$ percent of our sample have FIR properties similar to typical dusty star-forming galaxies at $z\sim2-3$ and a range of SFRs $<20-10000~{\rm M_{\odot}~yr^{-1}}$ for our sample as a whole. These results are in line with current models of quasar evolution and suggests a coexistence of dust-obscured star formation and AGN activity is typical of most quasars. We do not find a statistically-significant difference in the FIR luminosities of quasars in our sample with a radio excess relative to the radio-infrared correlation. Synchrotron emission is found to dominate at FIR wavelengths for $<15$ percent of those sources classified as powerful radio galaxies., Comment: 47 pages, 89 figures, accepted for publication in MNRAS

The formaldehyde MegaMaser emission has been mapped for the three host galaxies IC\,860. IRAS\,15107$+$0724, and Arp\,220. Elongated emission components are found at the nuclear centres of all galaxies with an extent ranging between 30 to 100 pc. These components are superposed on the peaks of the nuclear continuum. Additional isolated emission components are found superposed in the outskirts of the radio continuum structure. The brightness temperatures of the detected features ranges from 0.6 to 13.4 $\times 10^{4}$ K, which confirms their masering nature. The masering scenario is interpreted as amplification of the radio continuum by foreground molecular gas that is pumped by far-infrared radiation fields in these starburst environments of the host galaxies., Comment: Accepted MNRAS