Search

The recently discovered C-19 stellar stream is a collection of kinematically associated metal-poor stars in the halo of the Milky Way lacking an obvious progenitor. The stream spans an arc of ~15 degrees in the sky, and orbit-fitting suggests an apocentric distance of ~20 kpc and a pericentre of ~10 kpc. The narrow metallicity dispersion of stars with available spectra, together with light element abundance variations, suggests a globular cluster (GC) origin. The observed metallicity ([Fe/H] ~ -3.4), however, is much lower than that of any known GC. In addition, the width and velocity dispersion of the stream are similar to those expected from disrupting dwarf galaxies, and substantially larger than the tidal debris of GCs able to disrupt on C-19's orbit. We propose here an unconventional model where the C-19 progenitor is a dark matter-dominated stellar system with GC-like abundance patterns. We use N-body simulations to show that the tidal disruption of a ~100 pc King-model stellar component embedded in a ~20 km/s cuspy cold dark matter halo yields debris consistent with C-19's observed width and velocity dispersion. The stellar component of the progenitor is fully disrupted, and is spread over two distinct streams; one corresponding to C-19 and another possibly hiding behind the Galactic plane. If such companion stream were found, it would suggest that dark matter-dominated dwarfs may also develop GC-like enrichment patterns, a finding that would inform our theoretical understanding of the formation of multiple populations in GCs and dwarf galaxies alike., Comment: 9 pages, 9 figures. minor edits to match accepted version

We study the impact of the dark matter velocity distribution modelling on signals from velocity-dependent dark matter annihilation in Milky Way dwarf spheroidal galaxies. Using the high resolution APOSTLE simulations, we identify analogues corresponding to Milky Way dwarf spheroidal galaxies, and from these directly determine the dark matter pair-wise relative velocity distribution, and compare to best-fitting Maxwell-Boltzmann distribution models. For three velocity-dependent annihilation models, p-wave, d-wave, and the Sommerfeld model, we quantify the errors introduced when using the Maxwell-Boltzmann parameterization. We extract a simple power-law relation between the maximum circular velocity of the dwarf spheroidal analogue and the peak speed of the Maxwell-Boltzmann distribution. We show that this relation can be used to accurately calculate the dark matter relative velocity distribution, and find that it allows us to estimate the dark matter annihilation signal without the need to directly calculate the relative velocity distribution for each galaxy. The scatter in the J-factors calculated from the analogues dominates the uncertainty obtained when compared to the J-factor as determined from the observational data for each dwarf spheroidal, with the largest scatter from d-wave models and the smallest from Sommerfeld models., 17 pages, 7 figures, 2 tables

The C-19 stream is the most metal poor stellar system ever discovered, with a mean metallicity $[Fe/H] = -3.38\pm0.06$. Its low metallicity dispersion ($\sigma_{\rm [Fe/H]}$ $, Comment: 7 pages, 6 figures, accepted to MNRAS. The C-19 member list is published via https://github.com/zyuan-astro/C-19

The arrangement of dwarf galaxies in a thin plane surrounding the Milky Way has been thought to contradict the prevailing cosmological model of cold dark matter in the Universe. New work suggests that this arrangement may just be a temporary alignment, bringing our galaxy back into agreement with theoretical expectations once the radial distribution of satellites is taken into account., Comment: Invited 'News and Views' contribution for Nature Astronomy on paper by Sawala et al. 2022 "The Milky Way's plane of satellites is consistent with {\Lambda}CDM"

The formation of "stellar halos" in dwarf galaxies have been discussed in terms of early mergers or Galactic tides, although fluctuations in the gravitational potential due to stellar feedback is also a possible candidate mechanism. A Bayesian algorithm is used to find new candidate members in the extreme outskirts of the Sculptor dwarf galaxy. Precise metallicities and radial velocities for two distant stars are measured from their spectra taken with the Gemini South GMOS spectrograph. The radial velocity, proper motion and metallicity of these targets are consistent with Sculptor membership. As a result, the known boundary of the Sculptor dwarf extends now out to an elliptical distance of $\sim10$ half-light radii, which corresponds to a projected physical distance of $\sim3$ kpc. As reported in earlier work, the overall distribution of radial velocities and metallicities indicate the presence of a more spatially and kinematically dispersed metal-poor population that surrounds the more concentrated and colder metal-rich stars. Sculptor's density profile shows a "kink" in its logarithmic slope at a projected distance of $\sim25$ arcmin (620 pc), which we interpret as evidence that Galactic tides have helped to populate the distant outskirts of the dwarf. We discuss further ways to test and validate this tidal interpretation for the origin of these distant stars., Comment: Accepted for publication on MNRAS. Figure 1 now displays all the candidate members. arXiv admin note: text overlap with arXiv:2301.13214

The cuspy central density profiles of cold dark matter (CDM) haloes make them highly resilient to disruption by tides. Self-interactions between dark matter particles, or the cycling of baryons, may result in the formation of a constant-density core which would make haloes more susceptible to tidal disruption. We use N-body simulations to study the evolution of NFW-like "cored" subhaloes in the tidal field of a massive host, and identify the criteria and timescales for full tidal disruption. Our results imply that the survival of Milky Way satellites places constraints on the sizes of dark matter cores. Indeed, we find that no subhaloes with cores larger than 1 per cent of their initial NFW scale radius can survive for a Hubble time on orbits with pericentres 10 Gyr ago. These results suggest that many satellites have vanishingly small core sizes, consistent with CDM cusps. The discovery of further Milky Way satellites on orbits with small pericentric radii would strengthen these conclusions and allow for stricter upper limits on the core sizes., 13 pages, 13 figures, minor edits to match accepted version

The investigation of the metal-poor tail in the Galactic bulge provides unique information on the early Milky Way assembly and evolution. A chemo-dynamical analysis of 17 very metal-poor stars (VMP, [Fe/H] $, 22 pages + appendix, 14 Figures + 3 in the appendix, Accepted for publication in MNRAS

We present a new spectroscopic study of the faint Milky Way satellite Sagittarius II. Using multiobject spectroscopy from the Fibre Large Array Multi-Element Spectrograph, we supplement the data set of Longeard et al. with 47 newly observed stars, 19 of which are identified as members of the satellite. These additional member stars are used to put tighter constraints on the dynamics and the metallicity properties of the system. We find a low velocity dispersion of $\sigma _\mathrm{v}^\mathrm{SgrII} = 1.7 \pm 0.5$ km s−1, in agreement with the dispersion of Milky Way globular clusters of similar luminosity. We confirm the very metal-poor nature of the satellite ([Fe/H]$_\mathrm{spectro}^\mathrm{SgrII} = -2.23 \pm 0.07$) and find that the metallicity dispersion of Sgr II is not resolved, reaching only 0.20 at the 95 per cent confidence limit. No star with a metallicity below −2.5 is confidently detected. Therefore, despite the unusually large size of the system (r$_h = 35.5 ^{+1.4}_{-1.2}$ pc), we conclude that Sgr II is an old and metal-poor globular cluster of the Milky Way.

We use $\Lambda$CDM cosmological hydrodynamical simulations to explore the kinematics of gaseous discs in late-type dwarf galaxies. We create high-resolution 21-cm 'observations' of simulated dwarfs produced in two variations of the EAGLE galaxy formation model: one where supernova-driven gas flows redistribute dark matter and form constant-density central 'cores', and another where the central 'cusps' survive intact. We 'observe' each galaxy along multiple sight lines and derive a rotation curve for each observation using a conventional tilted-ring approach to model the gas kinematics. We find that the modelling process introduces systematic discrepancies between the recovered rotation curve and the actual circular velocity curve driven primarily by (i) non-circular gas orbits within the discs; (ii) the finite thickness of gaseous discs, which leads to overlap of different radii in projection; and (iii) departures from dynamical equilibrium. Dwarfs with dark matter cusps often appear to have a core, whilst the inverse error is less common. These effects naturally reproduce an observed trend which other models struggle to explain: late-type dwarfs with more steeply-rising rotation curves appear to be dark matter-dominated in the inner regions, whereas the opposite seems to hold in galaxies with core-like rotation curves. We conclude that if similar effects affect the rotation curves of observed dwarfs, a late-type dwarf population in which all galaxies have sizeable dark matter cores is most likely incompatible with current measurements., Comment: 21 pages, 13 figures, 1 table. Accepted for publication in MNRAS. Appendices included as ancillary file

The dark matter (DM) content of the Fornax dwarf spheroidal galaxy inferred from its kinematics is substantially lower than expected from LCDM cosmological simulations. We use N-body simulations to examine whether this may be the result of Galactic tides. We find that, despite improved proper motions from the Gaia mission, the pericentric distance of Fornax remains poorly constrained, mainly because its largest velocity component is roughly anti-parallel to the solar motion. Translating Fornax's proper motion into a Galactocentric velocity is thus sensitively dependent on Fornax's assumed distance: the observed distance uncertainty, $\pm 8\%$, implies pericentric distances that vary between $r_{\rm peri}\sim 50$ and $r_{\rm peri}\sim 150$ kpc. Our simulations show that for $r_{\rm peri}$ in the lower range of that estimate, a LCDM subhalo with maximum circular velocity $V_{\rm max}=40$ km s$^{-1}$ (or virial mass $M_{200}\approx 10^{10} M_\odot$, as expected from LCDM) would be tidally stripped to $V_{\rm max} \sim 23$ km s$^{-1}$ over $10$ Gyr. This would reduce the DM mass within the Fornax stellar half-mass radius to about half its initial value, bringing it into agreement with observations. Tidal stripping affects mainly Fornax's DM halo; its stellar component is affected little, losing less than $5\%$ of its initial mass in the process. We also explore the effect of Galactic tides on the dynamical friction decay times of Fornax's population of globular clusters (GC) and find little evidence for substantial changes, compared with models run in isolation. A population of GCs with initial orbital radii between $1$ and $2$ kpc is consistent with the present-day spatial distribution of Fornax GCs, despite assuming a cuspy halo. Neither the DM content nor the spatial distribution of GCs seem inconsistent with a simple model where Fornax inhabits a tidally-stripped cuspy cold DM halo., Comment: 10 pages, 8 figures, 4 tables. Accepted for publication in MNRAS

Stellar ejecta gradually enrich the gas out of which subsequent stars form, making the least chemically enriched stellar systems direct fossils of structures formed in the early universe. Although a few hundred stars with metal content below one thousandth of the solar iron content are known in the Galaxy, none of them inhabit globular clusters, some of the oldest known stellar structures. These show metal content of at least ~0.2 percent of the solar metallicity ([Fe/H] > -2.7). This metallicity floor appears universal and it has been proposed that proto-galaxies that merge into the galaxies we observe today were simply not massive enough to form clusters that survived to the present day. Here, we report the discovery of a stellar stream, C-19, whose metallicity is less than 0.05 per cent the solar metallicity ([Fe/H]=-3.38 +/- 0.06 (stat.) +/- 0.20 (syst.)). The low metallicity dispersion and the chemical abundances of the C-19 stars show that this stream is the tidal remnant of the most metal-poor globular cluster ever discovered, and significantly below the purported metallicity floor: clusters with significantly lower metallicities than observed today existed in the past and contributed their stars to the Milky Way halo., 14 pages, 6 figures, 5 tables. Nature, accepted

We use the APOSTLE Local Group (LG) cosmological hydro-simulations to examine the properties of "backsplash" galaxies, i.e, dwarfs which were within the virial boundaries of the Milky Way (MW) or M31 in the past, but are today outside their virial radius ($r_{200}$). More than half of all dwarfs between $1-2\,r_{200}$ of each primary are backsplash. More distant backsplash systems, i.e., those reaching distances well beyond $2\,r_{200}$, are typically close to apocentre of nearly radial orbits, and, therefore, essentially at rest relative to their primary. We use this result to investigate which LG dwarfs beyond $\sim500$ kpc of either primary could be a distant backsplash satellite of MW or M31. Tucana dSph, one of the few known quiescent LG field dwarfs, at $d_{\rm M31}\approx1350$ kpc and $d_{\rm MW}\approx880$ kpc, is a promising candidate. Tucana's radial velocity is consistent with being at rest relative to M31. Further, Tucana is located close to M33's orbital plane around M31, and simple orbit integrations indicate that Tucana may have been ejected during an early pericentric passage of M33 $\sim11$ Gyr ago, a timing which approximately coincides with Tucana's last episode of star formation. We suggest that Tucana may have been an early-infalling satellite of M31 or M33, providing a compelling explanation for its puzzling lack of gas and ongoing star formation despite its isolated nature. In this scenario, M33 should have completed some orbits around M31, a result that may help to explain the relative dearth of M33 satellite-candidates identified so far., Comment: 9 pages, 5 figures, Accepted for publication in MNRAS

The unusually low velocity dispersion and large size of Crater II pose a challenge to our understanding of dwarf galaxies in the Lambda Cold Dark Matter (LCDM) cosmogony. The low velocity dispersion suggests either a dark halo mass much lower than the minimum expected from hydrogen cooling limit arguments, or one that is in the late stages of extreme tidal stripping. The tidal interpretation has been favoured in recent work and is supported by the small pericentric distances consistent with available kinematic estimates. We use N-body simulations to examine this interpretation in detail, assuming a Navarro-Frenk-White (NFW) profile for Crater II's progenitor halo. Our main finding is that, although the low velocity dispersion can indeed result from the effect of tides, the large size of Crater II is inconsistent with this hypothesis. This is because galaxies stripped to match the observed velocity dispersion are also reduced to sizes much smaller than the observed half-light radius of Crater II. Unless its size has been substantially overestimated, reconciling this system with LCDM requires that either (i) it is not bound and near equilibrium (unlikely, given its crossing time is shorter than the time elapsed since pericentre), or that (ii) its progenitor halo deviates from the assumed NFW profile. The latter alternative may signal that baryons can affect the inner halo cusp even in extremely faint dwarfs or, more intriguingly, may signal effects associated with the intimate nature of the dark matter, such as finite self-interactions, or other such deviations from the canonical LCDM paradigm., 12 pages, 10 figures, 3 tables. Accepted for publication in MNRAS

We evaluate the rotational velocity of stars observed by the Pristine survey towards the Galactic anticentre, spanning a wide range of metallicities from the extremely metal-poor regime ($\mathrm{[Fe/H]} -1.5$. Furthermore, we detect signatures of a "heated disc", the so-called Splash, at metallicities higher than $\sim-1.5$. Finally, at $\mathrm{[Fe/H]} < -1.5$ our anticentre sample is dominated by a kinematical halo with a net prograde motion., Comment: 18 pages, 14 figures. Accepted for publication in MNRAS

We examine the spatial clustering of blue horizontal branch (BHB) stars from the u-band of the Canada–France Imaging Survey (CFIS, a component of the Ultraviolet Near-Infrared Optical Northern Survey, or UNIONS). All major groupings of stars are associated with previously known satellites, and among these is NGC 5466, a distant (16 kpc) globular cluster. NGC 5466 reportedly possesses a long stellar stream, although no individual members of the stream have previously been identified. Using both BHBs and more numerous red giant branch stars cross-matched to Gaia Data Release 2, we identify extended tidal tails from NGC 5466 that are both spatially and kinematically coherent. Interestingly, we find that this stream does not follow the same path as the previous detection at large distances from the cluster. We trace the stream across 31° of sky and show that it exhibits a very strong distance gradient in the range 10

We use N-body simulations to revisit the globular cluster (GC) ``timing problem'' in the Fornax dwarf spheroidal (dSph). In agreement with earlier work, we find that, due to dynamical friction, GCs sink to the center of dark matter halos with a cuspy inner density profile but ``stall'' at roughly 1/3 of the core radius ($r_{\rm core}$) in halos with constant-density cores. The timescales to sink or stall depend strongly on the mass of the GC and on the initial orbital radius, but are essentially the same for either cuspy (NFW) or cored halos normalized to have the same total mass within $r_{\rm core}$. Arguing against a cusp on the basis that GCs have not sunk to the center is thus no different from arguing against a core, unless all clusters are today at $\sim (1/3)\, r_{\rm core}$. This would imply a core radius exceeding $\sim 3$ kpc, much larger than seems plausible in any core-formation scenario. (The average projected distance of Fornax GCs is $\langle R_{\rm GC,Fnx}\rangle\sim 1$ kpc and its effective radius is $\sim 700$ pc.) A simpler explanation is that Fornax GCs have only been modestly affected by dynamical friction, as expected if clusters started orbiting at initial radii of order $\sim 1$-$2$ kpc, just outside Fornax's present-day half-light radius but well within the tidal radius imprinted by Galactic tides. This is not entirely unexpected. Fornax GCs are significantly older and more metal-poor than most Fornax stars, and such populations in dSphs tend to be more spatially extended than their younger and more metal-rich counterparts. Contrary to some earlier claims, our simulations further suggest that GCs do not truly ``stall'' at $\sim 0.3\, r_{\rm core}$, but rather continue decaying toward the center, albeit at reduced rates. We conclude that dismissing the presence of a cusp in Fornax based on the spatial distribution of its GC population is unwarranted., 7 pages, 4 figures. Submitted to MNRAS

Measurements of the rotation curves of dwarf galaxies are often interpreted as requiring a constant density core at the centre, at odds with the "cuspy" inner profiles predicted by $N$-body simulations of cold dark matter (CDM) haloes. It has been suggested that this conflict could be resolved by fluctuations in the inner gravitational potential caused by the periodic removal of gas following bursts of star formation. Earlier work has suggested that core formation requires a bursty and extended star formation history (SFH). Here we investigate the structure of CDM haloes of dwarf galaxies ($M_{{\rm DM}} \sim 10^9-5\times10^{10}\,{\rm M}_\odot$) formed in the APOSTLE ('A Project of Simulating the Local Environment') and AURIGA cosmological hydrodynamic simulations. Our simulations have comparable or better resolution than others that make cores ($M_{{\rm gas}} \sim 10^4\,{\rm M}_\odot$, gravitational softening $\sim 150$ pc). Yet, we do not find evidence of core formation at {\it any} mass or any correlation between the inner slope of the DM density profile and temporal variations in the SFH. APOSTLE and AURIGA dwarfs display a similar diversity in their cumulative SFHs to available data for Local Group dwarfs. Dwarfs in both simulations are DM-dominated on all resolved scales at all times, likely limiting the ability of gas outflows to alter significantly the central density profiles of their haloes. We conclude that recurrent bursts of star formation are not sufficient to cause the formation of cores, and that other conditions must also be met for baryons to be able to modify the central DM cusp., Comment: 15 pages, 9 figures, 2 tables. Updated version accepted for publication in MNRAS

We use astrometry, broad-band photometry and variability information from the Data Release 2 of ESA's Gaia mission (GDR2) to identify members of the Orphan Stream (OS) across the whole sky. The stream is traced above and below the celestial equator and in both Galactic hemispheres, thus increasing its visible length to ~ 210 degrees equivalent to ~150 kpc in physical extent. Taking advantage of the large number of RR Lyrae stars in the OS, we extract accurate distances and proper motions across the entire stretch of the tidal debris studied. As delineated by the GDR2 RR Lyrae, the stream exhibits two prominent twists in its shape on the sky which are accompanied by changes in the tangential motion. We complement the RR Lyrae maps with those created using GDR2 Red Giants and the DECam Legacy Survey Main Sequence Turn-Off stars. The behavior of the OS track on the sky is consistent across all three tracers employed. We detect a strong non-zero motion in the across-stream direction for a substantial portion of the stream. Such a misalignment between the debris track and the streaming velocity cannot be reproduced in a static gravitational potential and signals an interaction with a massive perturber., 18 pages; accepted to MNRAS

The early Universe presented a star formation environment that was almost devoid of heavy elements. The lowest metallicity stars thus provide a unique window into the earliest Galactic stages, but are exceedingly rare and difficult to find. Here we present the discovery of an ultra-metal-poor star, Pristine_221.8781+9.7844, using narrow-band Ca H&K photometry from the Pristine survey. Follow-up medium and high-resolution spectroscopy confirms the ultra-metal-poor nature of Pristine_221.8781+9.7844 ([Fe/H] = -4.66 +/- 0.13 in 1D LTE) with an enhancement of 0.3-0.4 dex in alpha-elements relative to Fe, and an unusually low carbon abundance. We derive an upper limit of A(C) = 5.6, well below typical A(C) values for such ultra metal-poor stars. This makes Pristine_221.8781+9.7844 one of the most metal-poor stars; in fact, it is very similar to the most metal-poor star known (SDSS J102915+172927). The existence of a class of ultra metal-poor stars with low(er) carbon abundances suggest that there must have been several formation channels in the early Universe through which long-lived, low-mass stars were formed., Re-submitted to MNRAS after minor revisions

We compare the star-forming properties of satellites around Milky Way (MW) analogs from the Stage~II release of the Satellites Around Galactic Analogs Survey (SAGA-II) to those from the APOSTLE and Auriga cosmological zoom-in simulation suites. We use archival GALEX UV imaging as a star-formation indicator for the SAGA-II sample and derive star-formation rates (SFRs) to compare with those from APOSTLE and Auriga. We compare our detection rates from the NUV and FUV bands to the SAGA-II H$\alpha$ detections and find that they are broadly consistent with over $85\%$ of observed satellites detected in all three tracers. We apply the same spatial selection criteria used around SAGA-II hosts to select satellites around the MW-like hosts in APOSTLE and Auriga. We find very good overall agreement in the derived SFRs for the star-forming satellites as well as the number of star-forming satellites per host in observed and simulated samples. However, the number and fraction of quenched satellites in the SAGA-II sample are significantly lower than those in APOSTLE and Auriga below a stellar mass of $M_*\sim10^{8}\,M_{\odot}$, even when the SAGA-II incompleteness and interloper corrections are included. This discrepancy is robust with respect to the resolution of the simulations and persists when alternative star-formation tracers are employed. We posit that this disagreement is not readily explained by vagaries in the observed or simulated samples considered here, suggesting a genuine discrepancy that may inform the physics of satellite populations around MW analogs., Comment: Accepted for publication in ApJ Letters after minor changes to text

The most metal-deficient stars hold important clues about the early build-up and chemical evolution of the Milky Way, and carbon-enhanced metal-poor (CEMP) stars are of special interest. However, little is known about CEMP stars in the Galactic bulge. In this paper, we use the large spectroscopic sample of metal-poor stars from the Pristine Inner Galaxy Survey (PIGS) to identify CEMP stars ([C/Fe] > +0.7) in the bulge region and to derive a CEMP fraction. We identify 96 new CEMP stars in the inner Galaxy, of which 62 are very metal-poor ([Fe/H] < -2.0); this is more than a ten-fold increase compared to the seven previously known bulge CEMP stars. The cumulative fraction of CEMP stars in PIGS is $42^{\,+14\,}_{\,-13} \%$ for stars with [Fe/H] < -3.0, and decreases to $16^{\,+3\,}_{\,-3} \%$ for [Fe/H] < -2.5 and $5.7^{\,+0.6\,}_{\,-0.5} \%$ for [Fe/H] < -2.0. The PIGS inner Galaxy CEMP fraction for [Fe/H] < -3.0 is consistent with the halo fraction found in the literature, but at higher metallicities the PIGS fraction is substantially lower. While this can partly be attributed to a photometric selection bias, such bias is unlikely to fully explain the low CEMP fraction at higher metallicities. Considering the typical carbon excesses and metallicity ranges for halo CEMP-s and CEMP-no stars, our results point to a possible deficiency of both CEMP-s and CEMP-no stars (especially the more metal-rich) in the inner Galaxy. The former is potentially related to a difference in the binary fraction, whereas the latter may be the result of a fast chemical enrichment in the early building blocks of the inner Galaxy., 15 pages, 9 figures, accepted for publication in MNRAS

We determine the dark matter pair-wise relative velocity distribution in a set of Milky Way-like halos in the Auriga and APOSTLE simulations. Focusing on the smooth halo component, the relative velocity distribution is well-described by a Maxwell-Boltzmann distribution over nearly all radii in the halo. We explore the implications for velocity-dependent dark matter annihilation, focusing on four models which scale as different powers of the relative velocity: Sommerfeld, s-wave, p-wave, and d-wave models. We show that the J-factors scale as the moments of the relative velocity distribution, and that the halo-to-halo scatter is largest for d-wave, and smallest for Sommerfeld models. The J-factor is strongly correlated with the dark matter density in the halo, and is very weakly correlated with the velocity dispersion. This implies that if the dark matter density in the Milky Way can be robustly determined, one can accurately predict the dark matter annihilation signal, without the need to identify the dark matter velocity distribution in the Galaxy., 25 pages, 11 figures, 2 tables; added figure 11; JCAP accepted version

The kinematics of the most metal-poor stars provide a window into the early formation and accretion history of the Milky Way. Here, we use 5~high-resolution cosmological zoom-in simulations ($\sim~5\times10^6$ star particles) of Milky Way-like galaxies taken from the NIHAO-UHD project, to investigate the origin of low-metallicity stars ([Fe/H]$\leq-2.5$). The simulations show a prominent population of low-metallicity stars confined to the disk plane, as recently discovered in the Milky Way. The ubiquity of this finding suggests that the Milky Way is not unique in this respect. Independently of the accretion history, we find that $\gtrsim~90$ per cent of the retrograde stars in this population are brought in during the initial build-up of the galaxies during the first few Gyrs after the Big Bang. Our results therefore highlight the great potential of the retrograde population as a tracer of the early build-up of the Milky Way. The prograde planar population, on the other hand, is accreted during the later assembly phase and samples the full galactic accretion history. In case of a quiet accretion history, this prograde population is mainly brought in during the first half of cosmic evolution ($t\lesssim7$~Gyr), while, in the case of an on-going active accretion history, later mergers on prograde orbits are also able to contribute to this population. Finally, we note that the Milky Way shows a rather large population of eccentric, very metal-poor planar stars. This is a feature not seen in most of our simulations, with the exception of one simulation with an exceptionally active early building phase., Comment: Submitted to MNRAS, 8 figures, 13 pages

Elliptical (E) and spiral (S) galaxies follow tight, but different, scaling laws that link their stellar masses, radii, and characteristic velocities. Mass and velocity, for example, scale tightly in spirals with little dependence on galaxy radius (the "Tully-Fisher relation"; TFR). On the other hand, ellipticals appear to trace a 2D surface in size-mass-velocity space (the "Fundamental Plane"; FP). Over the years, a number of studies have attempted to understand these empirical relations, usually in terms of variations of the virial theorem for E galaxies and in terms of the scaling relations of dark matter halos for spirals. We use Lambda cold dark matter (LCDM) cosmological hydrodynamical simulations to show that the scaling relations of both ellipticals and spirals arise as the result of (i) a tight galaxy mass-dark halo mass relation and (ii) the self-similar mass profile of cold dark matter halos. In this interpretation, E and S galaxies of a given stellar mass inhabit halos of similar masses, and their different scaling laws result from the varying amounts of dark matter enclosed within their luminous radii. This scenario suggests a new galaxy distance indicator applicable to galaxies of all morphologies and provides simple and intuitive explanations for long-standing puzzles, such as why the TFR is independent of surface brightness, or what causes the "tilt" in the FP. Our results provide strong support for the predictions of LCDM in the strongly non-linear regime, as well as guidance for further improvements to cosmological simulations of galaxy formation., Comment: 17 pages, 13 figures. Accepted for publication in A&A

Elemental abundances of the most metal-poor stars reflect the conditions in the early Galaxy and the properties of the first stars. We present a spectroscopic follow-up of two ultra metal-poor stars ([Fe/H], Comment: 17 pages, 11 figures, accepted for publication in MNRAS

We use N-body simulations to study the evolution of cuspy cold dark matter (CDM) halos in the gravitational potential of a massive host. Tidal mass losses reshape CDM halos, leaving behind bound remnants whose characteristic densities are set by the mean density of the host at the pericentre of their respective orbit. The evolution to the final bound remnant state is essentially complete after ~5 orbits for nearly circular orbits, while reaching the same remnant requires ~25 and ~40 orbits for eccentric orbits with 1:5 and 1:20 pericentre-to-apocentre ratios, respectively. The density profile of tidal remnants is fully specified by the fraction of mass lost, and approaches an exponentially-truncated Navarro-Frenk-White profile in the case of heavy mass loss. Resolving tidal remnants requires excellent numerical resolution; poorly resolved subhalos have systematically lower characteristic densities and are more easily disrupted. Even simulations with excellent spatial and time resolution fail when the final remnant is resolved with fewer than 3000 particles. We derive a simple empirical model that describes the evolution of the mass and the density profile of the tidal remnant applicable to a wide range of orbital eccentricities and pericentric distances. Applied to the Milky Way, our results suggest that $10^8$ - $10^{10}$ solar mass halos accreted ~10 Gyrs ago on 1:10 orbits with pericentric distance ~10 kpc should have been stripped to 0.1 - 1 per cent of their original mass. This implies that estimates of the survival and structure of such halos (the possible hosts of ultra-faint Milky Way satellites) based on direct cosmological simulations may be subject to substantial revision., 15 pages, 15 figures. Edited to match accepted version

Local Group satellite galaxies show a wide diversity of star formation histories (SFHs) whose origin is yet to be fully understood. Using hydrodynamical simulations from the Constrained Local UniversE project, we study the SFHs of satellites of Milky Way-like galaxies in a cosmological context: while in the majority of the cases the accretion onto their host galaxy causes the satellites to lose their gas, with a subsequent suppression in SF, in about 25$\%$ of our sample we observe a clear enhancement of SF after infall. Peaks in SF clearly correlate with the satellite pericentric passage around its host and, in one case, with a satellite-satellite interaction. We identify two key ingredients that result in enhanced SF after infall: galaxies must enter the host's virial radius with a reservoir of cold gas $M_{\rm gas,inf}/M_{\rm vir,inf}\gtrsim 10^{-2}$ and with a minimum pericentric distance $\gtrsim$10 kpc (mean distance $\sim$50 kpc for the full sample), in order to form new stars due to compression of cold gas at pericentric passage. On the other hand, satellites that infall with little gas or whose pericentric distance is too small, have their gas ram-pressure stripped and subsequent SF quenched. The pericentric passage of satellites likewise correlates with SF peaks in their hosts, suggesting that this mechanism induces bursts of SF in satellites and central galaxies alike, in agreement with recent studies of our Galaxy's SFH. Our findings can explain the recently reported multiple stellar populations observed in dwarf galaxies such as Carina and Fornax, and should be taken into account in semi-analytic models of galaxy formation and satellite quenching., Comment: 15 pages, 12 figures, matches accepted MNRAS version. Expanded sec 3.2. New figures 2 and 3 show comparison between simulated and observed SFHs of satellites and isolated dwarfs

The abundance of the faintest galaxies provides insight into the nature of dark matter and the process of dwarf galaxy formation. In the LCDM scenario, low mass halos are so numerous that the efficiency of dwarf formation must decline sharply with decreasing halo mass in order to accommodate the relative scarcity of observed dwarfs and satellites in the Local Group. The nature of this decline contains important clues to the mechanisms regulating the onset of galaxy formation in the faintest systems. We explore here two possible models for the stellar mass ($M_*$)-halo mass ($M_{200}$) relation at the faint end, motivated by some of the latest LCDM cosmological hydrodynamical simulations. One model includes a sharp mass threshold below which no luminous galaxies form, as expected if galaxy formation proceeds only in systems above the Hydrogen-cooling limit. In the second model, $M_*$ scales as a steep power-law of $M_{200}$ with no explicit cutoff, as suggested by recent semianalytic work. Although both models predict satellite numbers around Milky Way-like galaxies consistent with current observations, they predict vastly different numbers of ultra-faint dwarfs and of satellites around isolated dwarf galaxies. Our results illustrate how the satellite mass function around dwarfs may be used to probe the $M_*$-$M_{200}$ relation at the faint end and to elucidate the mechanisms that determine which low-mass halos "light up" or remain dark in the LCDM scenario., Comment: 14 pages, 11 figures. Accepted version for publication in MNRAS

[Context] Data-driven methods play an increasingly important role in the field of astrophysics In the context of large spectroscopic surveys of stars, data-driven methods are key in deducing physical parameters for millions of spectra in a short time. Convolutional neural networks (CNNs) enable us to connect observables (e.g. spectra, stellar magnitudes) to physical properties (atmospheric parameters, chemical abundances, or labels in general)., [Aims] We test whether it is possible to transfer the labels derived from a high-resolution stellar survey to intermediate-resolution spectra of another survey by using a CNN., [Methods] We trained a CNN, adopting stellar atmospheric parameters and chemical abundances from APOGEE DR16 (resolution R = 22 500) data as training set labels. As input, we used parts of the intermediate-resolution RAVE DR6 spectra (R ∼ 7500) overlapping with the APOGEE DR16 data as well as broad-band ALL_WISE and 2MASS photometry, together with Gaia DR2 photometry and parallaxes., [Results] We derived precise atmospheric parameters Teff, log(g), and [M/H], along with the chemical abundances of [Fe/H], [α/M], [Mg/Fe], [Si/Fe], [Al/Fe], and [Ni/Fe] for 420 165 RAVE spectra. The precision typically amounts to 60 K in Teff, 0.06 in log(g) and 0.02−0.04 dex for individual chemical abundances. Incorporating photometry and astrometry as additional constraints substantially improves the results in terms of the accuracy and precision of the derived labels, as long as we operate in those parts of the parameter space that are well-covered by the training sample. Scientific validation confirms the robustness of the CNN results. We provide a catalogue of CNN-trained atmospheric parameters and abundances along with their uncertainties for 420 165 stars in the RAVE survey., [Conclusions] CNN-based methods provide a powerful way to combine spectroscopic, photometric, and astrometric data without the need to apply any priors in the form of stellar evolutionary models. The developed procedure can extend the scientific output of RAVE spectra beyond DR6 to ongoing and planned surveys such as Gaia RVS, 4MOST, and WEAVE. We call on the community to place a particular collective emphasis and on efforts to create unbiased training samples for such future spectroscopic surveys., Funding for RAVE has been provided by: the Leibniz-Institut für Astrophysik Potsdam (AIP); the Australian Astronomical Observatory; the Australian National University; the Australian Research Council; the French National Research Agency (Programme National Cosmology et Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES); the German Research Foundation (SPP 1177 and SFB 881); the European Research Council (ERC-StG 240271 Galactica); the Istituto Nazionale di Astrofisica at Padova; The Johns Hopkins University; the National Science Foundation of the USA (AST-0908326); the W. M. Keck foundation; the Macquarie University; the Netherlands Research School for Astronomy; the Natural Sciences and Engineering Research Council of Canada; the Slovenian Research Agency (research core funding no. P1-0188); the Swiss National Science Foundation; the Science & Technology Facilities Council of the UK; Opticon; Strasbourg Observatory; and the Universities of Basel, Groningen, Heidelberg, and Sydney. TZ acknowledges financial support of the Slovenian Research Agency (research core funding No. P1-0188) and of the ESA project PHOTO2CHEM (C4000127986). FA is grateful for funding from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 800502. This work has made use of data from the European Space Agency (ESA) mission Gaia (http://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, http://www.cosmos.esa. int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement. This publication makes use of data products from the Wide-field Infrared Survey Explorer, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, funded by the National Aeronautics and Space Administration. This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

The pattern speed with which galactic bars rotate is intimately linked to the amount of dark matter in the inner regions of their host galaxies. In particular, dark matter haloes act to slow down bars via torques exerted through dynamical friction. Observational studies of barred galaxies tend to find that bars rotate fast, while hydrodynamical cosmological simulations of galaxy formation and evolution in the $\Lambda$CDM framework have previously found that bars slow down excessively. This has led to a growing tension between fast bars and the $\Lambda$CDM cosmological paradigm. In this study we revisit this issue, using the Auriga suite of high resolution, magneto-hydrodynamical cosmological zoom-in simulations of galaxy formation and evolution in the $\Lambda$CDM framework, finding that bars remain fast down to $z=0$. In Auriga, bars form in galaxies that have higher stellar-to-dark matter ratios and are more baryon-dominated than in previous cosmological simulations; this suggests that in order for bars to remain fast, massive spiral galaxies must lie above the commonly used abundance matching relation. While this reduces the aforementioned tension between the rotation speed of bars and $\Lambda$CDM, it accentuates the recently reported discrepancy between the dynamically inferred stellar-to-dark matter ratios of massive spirals and those inferred from abundance matching. Our results highlight the potential of using bar dynamics to constrain models of galaxy formation and evolution., Comment: 5 pages + 4 pages of appendices; 7 figures; Accepted for publication in A&A Letters

We use the APOSTLE $\Lambda$CDM cosmological hydrodynamical simulations of the Local Group to study the recent accretion of massive satellites into the halo of Milky Way (MW)-sized galaxies. These systems are selected to be close analogues to the Large Magellanic Cloud (LMC), the most massive satellite of the MW. The simulations allow us to address, in a cosmological context, the impact of the Clouds on the MW, including the contribution of Magellanic satellites to the MW satellite population, and the constraints placed on the Galactic potential by the motion of the LMC. We show that LMC-like satellites are twice more common around Local Group-like primaries than around isolated halos of similar mass; these satellites come from large turnaround radii and are on highly eccentric orbits whose velocities at first pericentre are comparable with the primary's escape velocity. This implies $V_{\rm esc}^{\rm MW} (50 $ kpc$)\sim 365$ km/s, a strong constraint on Galactic potential models. LMC analogues contribute about 2 satellites with $M_*>10^5\, M_\odot$, having thus only a mild impact on the luminous satellite population of their hosts. At first pericentre, LMC-associated satellites are close to the LMC in position and velocity, and are distributed along the LMC's orbital plane. Their orbital angular momenta roughly align with the LMC's, but, interestingly, they may appear to "counter-rotate" the MW in some cases. These criteria refine earlier estimates of the LMC association of MW satellites: only the SMC, Hydrus1, Car3, Hor1, Tuc4, Ret2 and Phoenix2 are compatible with all criteria. Carina, Grus2, Hor2 and Fornax are less probable associates given their large LMC relative velocity., Comment: 17 pages, 14 figures, 1 table. Accepted for publication in MNRAS

We test a non-parametric higher-order Jeans analysis method, GravSphere, on 32 simulated dwarf galaxies comparable to classical Local Group dwarfs like Fornax. The galaxies are selected from the APOSTLE suite of cosmological hydrodynamics simulations with Cold Dark Matter (CDM) and Self-Interacting Dark Matter (SIDM) models, allowing us to investigate cusps and cores in density distributions. We find that, for CDM dwarfs, the recovered enclosed mass profiles have a bias of no more than 10 per cent, with a 50 per cent scatter in the inner regions and a 20 per cent scatter near the half-light radius, consistent with standard mass estimators. The density profiles are also recovered with a bias of no more than 10 per cent and a scatter of 30 per cent in the inner regions. For SIDM dwarfs, the mass and density profiles are recovered within our 95 per cent confidence intervals, but are biased towards cuspy dark matter distributions. This is mainly due to a lack of sufficient constraints from the data. We explore the sources of scatter in the accuracy of the recovered profiles and suggest a $\chi^2$ statistic to separate successful models from biased ones. Finally, we show that the uncertainties on the mass profiles obtained with GravSphere are smaller than those for comparable Jeans methods, and that they can be further improved if stronger priors, motivated by cosmological simulations, are placed on the velocity anisotropy. We conclude that GravSphere is a promising Jeans-based approach for modelling dark matter distributions in dwarf galaxies., Comment: 22 pages, 13 figures. Accepted in MNRAS

The Radial Velocity Experiment (Rave) is a magnitude-limited (9 < I < 12) spectroscopic survey of Galactic stars randomly selected in Earth's southern hemisphere. The Rave medium-resolution spectra (R ∼ 7500) cover the Ca-triplet region (8410-8795 Å). The sixth and final data release (DR6) is based on 518,387 observations of 451,783 unique stars. Rave observations were taken between 2003 April 12 and 2013 April 4. Here we present the genesis, setup, and data reduction of Rave as well as wavelength-calibrated and flux-normalized spectra and error spectra for all observations in Rave DR6. Furthermore, we present derived spectral classification and radial velocities for the Rave targets, complemented by cross-matches with Gaia DR2 and other relevant catalogs. A comparison between internal error estimates, variances derived from stars with more than one observing epoch, and a comparison with radial velocities of Gaia DR2 reveals consistently that 68% of the objects have a velocity accuracy better than 1.4 km s-1, while 95% of the objects have radial velocities better than 4.0 km s-1. Stellar atmospheric parameters, abundances and distances are presented in a subsequent publication. The data can be accessed via the Rave website (http://rave-survey.org) or the Vizier database., Funding for Rave has been provided by: the Leibniz-Institut f¨ur Astrophysik Potsdam (AIP); the Australian Astronomical Observatory; the Australian National University; the Australian Research Council; the French National Research Agency (Programme National Cosmology et Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES); the German Research Foundation (SPP 1177 and SFB 881); the European Research Council (ERC-StG 240271 Galactica); the Istituto Nazionale di Astrofisica at Padova; The Johns Hopkins University; the National Science Foundation of the USA (AST-0908326); the W. M. Keck foundation; the Macquarie University; the Netherlands Research School for Astronomy; the Natural Sciences and Engineering Research Council of Canada; the Slovenian Research Agency (research core funding no. P1-0188); the Swiss National Science Foundation; the Science & Technology Facilities Council of the UK; Opticon; Strasbourg Observatory; and the Universities of Basel, Groningen, Heidelberg, and Sydney. PJM is supported by grant 2017-03721 from the Swedish Research Council. LC is the recipient of the ARC Future Fellowship FT160100402. RAG acknowledges the support from the PLATO CNES grant. SM would like to acknowledge support from the Spanish Ministry with the Ramon y Cajal fellowship number RYC-2015-17697. MS thanks the Research School of Astronomy & Astrophysics in Canberra for support through a Distinguished Visitor Fellowship. RFGW thanks the Kavli Institute for Theoretical Physics and the Simons Foundation for support as a Simons Distinguished Visiting Scholar. This research was supported in part by the National Science Foundation under Grant No. NSF PHY-1748958 to KITP.

We use cosmological simulations of isolated Milky Way-mass galaxies, as well as Local Group analogues, to define the "edge" -- a caustic manifested in a drop in density or radial velocity -- of Galactic-sized haloes, both in dark matter and in stars. In the dark matter, we typically identify two caustics: the outermost caustic located at ~1.4r_200m corresponding to the "splashback" radius, and a second caustic located at ~0.6r_200m which likely corresponds to the edge of the virialized material which has completed at least two pericentric passages. The splashback radius is ill defined in Local Group type environments where the halos of the two galaxies overlap. However, the second caustic is less affected by the presence of a companion, and is a more useful definition for the boundary of the Milky Way halo. Curiously, the stellar distribution also has a clearly defined caustic, which, in most cases, coincides with the second caustic of the dark matter. This can be identified in both radial density and radial velocity profiles, and should be measurable in future observational programmes. Finally, we show that the second caustic can also be identified in the phase-space distribution of dwarf galaxies in the Local Group. Using the current dwarf galaxy population, we predict the edge of the Milky Way halo to be 292 +/- 61 kpc., Comment: 14 pages, 11 figures. MNRAS in press

We present part 2 of the sixth and final Data Release (DR6) of the Radial Velocity Experiment (Rave), a magnitude-limited spectroscopic survey of Galactic stars randomly selected in Earth's southern hemisphere. The Rave medium-resolution spectra (R ∼ 7500) cover the Ca triplet region (8410-8795 Å) and span the complete time frame from the start of Rave observations on 2003 April 12 to their completion on 2013 April 4. In the second of two publications, we present the data products derived from 518,387 observations of 451,783 unique stars using a suite of advanced reduction pipelines focusing on stellar atmospheric parameters, in particular purely spectroscopically derived stellar atmospheric parameters, and the overall metallicity), enhanced stellar atmospheric parameters inferred via a Bayesian pipeline using Gaia DR2 astrometric priors, and asteroseismically calibrated stellar atmospheric parameters for giant stars based on asteroseismic observations for 699 K2 stars. In addition, we provide abundances of the elements Fe, Al, and Ni, as well as an overall [α/Fe] ratio obtained using a new pipeline based on the GAUGUIN optimization method that is able to deal with variable signal-to-noise ratios. The Rave DR6 catalogs are cross-matched with relevant astrometric and photometric catalogs, and are complemented by orbital parameters and effective temperatures based on the infrared flux method. The data can be accessed via the Rave website (http://rave-survey.org) or the Vizier database., Funding for Rave has been provided by: the Leibniz-Institut f¨ur Astrophysik Potsdam (AIP); the Australian Astronomical Observatory; the Australian National University; the Australian Research Council; the French National Research Agency (Programme National Cosmology et Galaxies (PNCG) of CNRS/INSU with INP and IN2P3, co-funded by CEA and CNES); the German Research Foundation (SPP 1177 and SFB 881); the European Research Council (ERC-StG 240271 Galactica); the Istituto Nazionale di Astrofisica at Padova; The Johns Hopkins University; the National Science Foundation of the USA (AST-0908326); the W. M. Keck foundation; the Macquarie University; the Netherlands Research School for Astronomy; the Natural Sciences and Engineering Research Council of Canada; the Slovenian Research Agency (research core funding no. P1-0188); the Swiss National Science Foundation; the Science & Technology Facilities Council of the UK; Opticon; Strasbourg Observatory; and the Universities of Basel, Groningen, Heidelberg, and Sydney. PJM is supported by grant 2017-03721 from the Swedish Research Council. LC is the recipient of the ARC Future Fellowship FT160100402. RAG acknowledges the support from the PLATO CNES grant. SM would like to acknowledge support from the Spanish Ministry with the Ramon y Cajal fellowship number RYC-2015-17697. MS thanks the Research School of Astronomy & Astrophysics in Canberra for support through a Distinguished Visitor Fellowship. RFGW thanks the Kavli Institute for Theoretical Physics and the Simons Foundation for support as a Simons Distinguished Visiting Scholar. This research was supported in part by the National Science Foundation under Grant No. NSF PHY-1748958 to KITP.

This compilation is the fourth data release from the $R$-Process Alliance (RPA) search for $r$-process-enhanced stars, and the second release based on "snapshot" high-resolution ($R \sim 30,000$) spectra collected with the du Pont 2.5m Telescope. In this data release, we propose a new delineation between the $r$-I and $r$-II stellar classes at $\mathrm{[Eu/Fe]} = +0.7$, instead of the empirically chosen $\mathrm{[Eu/Fe]} = +1.0$ level previously in use, based on statistical tests of the complete set of RPA data released to date. We also statistically justify the minimum level of [Eu/Fe] for definition of the $r$-I stars, [Eu/Fe] $> +0.3$. Redefining the separation between $r$-I and $r$-II stars will aid in analysis of the possible progenitors of these two classes of stars and whether these signatures arise from separate astrophysical sources at all. Applying this redefinition to previous RPA data, the number of identified $r$-II and $r$-I stars changes to 51 and 121, respectively, from the initial set of data releases published thus far. In this data release, we identify 21 new $r$-II, 111 new $r$-I (plus three re-identified), and 7 new (plus one re-identified) limited-$r$ stars out of a total of 232 target stars, resulting in a total sample of 72 new $r$-II stars, 232 new $r$-I stars, and 42 new limited-$r$ stars identified by the RPA to date., 17 pages, 7 figures, 6 tables; accepted to ApJS

Metal-poor stars are important tools for tracing the early history of the Milky Way, and for learning about the first generations of stars. Simulations suggest that the oldest metal-poor stars are to be found in the inner Galaxy. Typical bulge surveys, however, lack low metallicity ([Fe/H] < -1.0) stars because the inner Galaxy is predominantly metal-rich. The aim of the Pristine Inner Galaxy Survey (PIGS) is to study the metal-poor and very metal-poor (VMP, [Fe/H] < -2.0) stars in this region. In PIGS, metal-poor targets for spectroscopic follow-up are selected from metallicity-sensitive CaHK photometry from the CFHT. This work presents the ~250 deg^2 photometric survey as well as intermediate-resolution spectroscopic follow-up observations for ~8000 stars using AAOmega on the AAT. The spectra are analysed using two independent tools: ULySS with an empirical spectral library, and FERRE with a library of synthetic spectra. The comparison between the two methods enables a robust determination of the stellar parameters and their uncertainties. We present a sample of 1300 VMP stars -- the largest sample of VMP stars in the inner Galaxy to date. Additionally, our spectroscopic dataset includes ~1700 horizontal branch stars, which are useful metal-poor standard candles. We furthermore show that PIGS photometry selects VMP stars with unprecedented efficiency: 86%/80% (lower/higher extinction) of the best candidates satisfy [Fe/H] < -2.0, as do 80%/63% of a larger, less strictly selected sample. We discuss future applications of this unique dataset that will further our understanding of the chemical and dynamical evolution of the innermost regions of our Galaxy., accepted for publication in MNRAS, 17 pages, 9 figures