arXiv subjects

Gerry Gilmore

Publications and source records attributed to Gerry Gilmore.

At least 19 recordsLinked to original sources

arXiv2025

Misspecification-robust amortised simulation-based inference using variational methods

Recent advances in neural density estimation have enabled powerful simulation-based inference (SBI) methods that can flexibly approximate Bayesian inference for intractable stochastic models. Although these methods have demonstrated reliable posterior estimation when the simulator accurately represents the underlying data generative process (DGP), recent work has shown that they perform poorly in the presence of model misspecification. This poses a significant issue for their use in real-world problems, due to simulators always misrepresenting the true DGP to a certain degree. In this paper, we introduce robust variational neural posterior estimation (RVNP), a method which addresses the problem of misspecification in amortised SBI by bridging the simulation-to-reality gap using variational inference and error modelling. We test RVNP on multiple benchmark tasks, including using real data from astronomy, and show that it can recover robust posterior inference in a data-driven manner without adopting hyperparameters or priors governing the misspecification influence.

stat.ML↗

arXiv2025

Data-driven dust inference at mid-to-high Galactic latitudes using probabilistic machine learning

We present a method for accurately and precisely inferring photometric dust extinction towards stars at mid-to-high Galactic latitudes using probabilistic machine learning to model the colour-magnitude distribution of zero-extinction stars in these regions. Photometric dust maps rely on a robust method for inferring stellar reddening. At high Galactic latitudes, where extinction is low, such inferences are particularly susceptible to contamination from modelling errors and prior assumptions, potentially introducing artificial structure into dust maps. In this work, we demonstrate the use of normalising flows to learn the conditional probability distribution of the photometric colour-magnitude relations of zero-extinction stars, conditioned on Galactic cylindrical coordinates for stars at mid-to-high Galactic latitudes. By using the normalising flow to model the colour-magnitude diagram, we infer the posterior distribution of dust extinction towards stars along different lines of sight by marginalising over the colour-magnitude flow. We validate our method using data from Gaia, Pan-STARRS, and 2MASS, showing that we can recover unbiased posteriors and successfully detect dust along the line-of-sight in two calibration regions at mid-Galactic latitude that have been extensively studied in the context of polarisation surveys.

astro-ph.GA↗

arXiv2023

Quantifying Interstellar Extinction at High Galactic Latitudes

A detailed map of the distribution of dust at high Galactic latitudes is essential for future cosmic microwave background (CMB) polarization experiments because the dust, while diffuse, remains a significant foreground in these regions. We develop a Bayesian model to identify a region of the Hertzsprung-Russell (HR) diagram suited to constrain the single-star extinction accurately at high Galactic latitudes. Using photometry from Gaia, 2MASS and ALLWISE, and parallax from Gaia, we employ nested sampling to fit the model to the data and analyse the posterior over stellar parameters for both synthetic and real data. Charting low variations in extinction is complex due to systematic errors and degeneracies between extinction and other stellar parameters. The systematic errors can be minimised by restricting our data to a region of the HR diagram where the stellar models are most accurate. Moreover, the degeneracies can be reduced by including astrophysical priors and spectroscopic constraints. We show that accounting for the measurement error of the data and the assumed inaccuracies of the stellar models are critical in accurately recovering small variations in extinction. We compare the posterior distribution for individual stars with spectroscopic stellar parameter estimates from LAMOST and Gaia ESO and illustrate the importance of estimating extinction and effective temperature as a joint solution.

astro-ph.GA↗

arXiv2023

Assembling a high-precision abundance catalogue of solar twins in GALAH for phylogenetic studies

Stellar chemical abundances have proved themselves a key source of information for understanding the evolution of the Milky Way, and the scale of major stellar surveys such as GALAH have massively increased the amount of chemical data available. However, progress is hampered by the level of precision in chemical abundance data as well as the visualization methods for comparing the multidimensional outputs of chemical evolution models to stellar abundance data. Machine learning methods have greatly improved the former; while the application of tree-building or phylogenetic methods borrowed from biology are beginning to show promise with the latter. Here we analyse a sample of GALAH solar twins to address these issues. We apply The Cannon algorithm to generate a catalogue of about 40,000 solar twins with 14 high precision abundances which we use to perform a phylogenetic analysis on a selection of stars that have two different ranges of eccentricities. From our analyses we are able to find a group with mostly stars on circular orbits and some old stars with eccentric orbits whose age-[Y/Mg] relation agrees remarkably well with the chemical clocks published by previous high precision abundance studies. Our results show the power of combining survey data with machine learning and phylogenetics to reconstruct the history of the Milky Way.

astro-ph.GA↗

arXiv2020

The Gaia-ESO Survey: Oxygen abundance in the Galactic thin and thick disks

We analyze the oxygen abundances of a stellar sample representative of the two major Galactic populations: the thin and thick disks. The aim is to investigate the differences between members of the Galactic disks and to contribute to the understanding on the origin of oxygen chemical enrichment in the Galaxy. The analysis is based on the [O\,{\sc i}]=6300.30\,Å~ oxygen line in HR spectra ($R\sim$52,500) obtained from the GES Survey. By comparing the observed spectra with a theoretical dataset, computed in LTE with the SPECTRUM synthesis and ATLAS12 codes, we derive the oxygen abundances of 516 FGK dwarfs for which we have previously measured carbon abundances. Based on kinematic, chemical and dynamical considerations we identify 20 thin and 365 thick disk members. We study potential trends of both subsamples in terms of their chemistry ([O/H], [O/Fe], [O/Mg], and [C/O] versus [Fe/H] and [Mg/H]), age, and position in the Galaxy. Main results are: (a) [O/H] and [O/Fe] ratios versus [Fe/H] show systematic differences between thin and thick disk stars with enhanced O abundance of thick disk stars with respect to thin disk members and a monotonic decrement of [O/Fe] with increasing metallicity, even at metal-rich regime; (b) a smooth correlation of [O/Mg] with age in both populations, suggesting that this abundance ratio can be a good proxy of stellar ages within the Milky Way; (c) thin disk members with [Fe/H]$\simeq0$ display a [C/O] ratio smaller than the solar value, suggesting a possibly outward migration of the Sun from lower Galactocentric radii.

astro-ph.GA↗

arXiv2019

The Gaia-ESO Survey: Carbon abundance in the Galactic thin and thick disks

This paper focuses on carbon that is one of the most abundant elements in the Universe and is of high importance in the field of nucleosynthesis and galactic and stellar evolution. Even nowadays, the origin of carbon and the relative importance of massive and low- to intermediate-mass stars in producing it is still a matter of debate. In this paper we aim at better understanding the origin of carbon by studying the trends of [C/H], [C/Fe],and [C/Mg] versus [Fe/H], and [Mg/H] for 2133 FGK dwarf stars from the fifth Gaia-ESO Survey internal data release (GES iDR5). The availability of accurate parallaxes and proper motions from Gaia DR2 and radial velocities from GES iDR5 allows us to compute Galactic velocities, orbits and absolute magnitudes and, for 1751 stars, ages via a Bayesian approach. Three different selection methodologies have been adopted to discriminate between thin and thick disk stars. In all the cases, the two stellar groups show different abundance ratios, [C/H], [C/Fe], and [C/Mg], and span different age intervals, with the thick disk stars being, on average, older than those in the thin disk. The behaviours of [C/H], [C/Fe], and [C/Mg] versus [Fe/H], [Mg/H], and age all suggest that C is primarily produced in massive stars like Mg. The increase of [C/Mg] for young thin disk stars indicates a contribution from low-mass stars or the increased C production from massive stars at high metallicities due to the enhanced mass loss. The analysis of the orbital parameters Rmed and |Zmax| support an "inside-out" and "upside-down" formation scenario for the disks of Milky Way.

astro-ph.SR↗

arXiv2018

Gaia-ESO Survey: INTRIGOSS - A new library of High Resolution Synthetic Spectra

We present a high resolution synthetic spectral library, INTRIGOSS, designed for studying FGK stars. The library is based on atmosphere models computed with specified individual element abundances via ATLAS12 code. Normalized SPectra (NSP) and surface Flux SPectra (FSP), in the 4830-5400 A, wavelength range, were computed with the SPECTRUM code. INTRIGOSS uses the solar composition by Grevesse et al. 2007 and four [alpha/Fe] abundance ratios and consists of 15,232 spectra. The synthetic spectra are computed with astrophysical gf-values derived by comparing synthetic predictions with a very high SNR solar spectrum and the UVES-U580 spectra of five cool giants. The validity of the NSPs is assessed by using the UVES-U580 spectra of 2212 stars observed in the framework of the Gaia-ESO Survey and characterized by homogeneous and accurate atmospheric parameter values and by detailed chemical compositions. The greater accuracy of NSPs with respect to spectra from the AMBRE, GES_Grid, PHOENIX, C14, and B17 synthetic spectral libraries is demonstrated by evaluating the consistency of the predictions of the different libraries for the UVES-U580 sample stars. The validity of the FSPs is checked by comparing their prediction with both observed spectral energy distribution and spectral indices. The comparison of FSPs with SEDs derived from ELODIE, INDO--U.S., and MILES libraries indicates that the former reproduce the observed flux distributions within a few percent and without any systematic trend. The good agreement between observational and synthetic Lick/SDSS indices shows that the predicted blanketing of FSPs well reproduces the observed one, thus confirming the reliability of INTRIGOSS FSPs.

astro-ph.SR↗

arXiv2017

The selection function of the RAVE survey

We characterize the selection function of RAVE using 2MASS as our underlying population, which we assume represents all stars which could have potentially been observed. We evaluate the completeness fraction as a function of position, magnitude, and color in two ways: first, on a field-by-field basis, and second, in equal-size areas on the sky. Then, we consider the effect of the RAVE stellar parameter pipeline on the final resulting catalogue, which in principle limits the parameter space over which our selection function is valid. Our final selection function is the product of the completeness fraction and the selection function of the pipeline. We then test if the application of the selection function introduces biases in the derived parameters. To do this, we compare a parent mock catalogue generated using Galaxia with a mock-RAVE catalogue where the selection function of RAVE has been applied. We conclude that for stars brighter than I = 12, between $4000 \rm K < T_{\rm eff} < 8000 \rm K$ and $0.5 < \rm{log}\,g < 5.0$, RAVE is kinematically and chemically unbiased with respect to expectations from Galaxia.

astro-ph.GA↗

arXiv2016

The RAVE-on catalog of stellar atmospheric parameters and chemical abundances for chemo-dynamic studies in the Gaia era

The orbits, atmospheric parameters, chemical abundances, and ages of individual stars in the Milky Way provide the most comprehensive illustration of galaxy formation available. The Tycho-Gaia Astrometric Solution (TGAS) will deliver astrometric parameters for the largest ever sample of Milky Way stars, though its full potential cannot be realized without the addition of complementary spectroscopy. Among existing spectroscopic surveys, the RAdial Velocity Experiment (RAVE) has the largest overlap with TGAS ($\gtrsim$200,000 stars). We present a data-driven re-analysis of 520,781 RAVE spectra using The Cannon. For red giants, we build our model using high-fidelity APOGEE stellar parameters and abundances for stars that overlap with RAVE. For main-sequence and sub-giant stars, our model uses stellar parameters from the K2/EPIC. We derive and validate effective temperature $T_{\rm eff}$, surface gravity $\log{g}$, and chemical abundances of up to seven elements (O, Mg, Al, Si, Ca, Fe, Ni). We report a total of 1,685,851 elemental abundances with a typical precision of 0.07 dex, a substantial improvement over previous RAVE data releases. The synthesis of RAVE-on and TGAS is the most powerful data set for chemo-dynamic analyses of the Milky Way ever produced.

astro-ph.SR↗

arXiv2016

An accurate and self-consistent chemical abundance catalogue for the APOGEE/Kepler sample

Context. The APOGEE survey has obtained high-resolution infrared spectra of more than 100,000 stars. Deriving chemical abundances patterns of these stars is paramount to piecing together the structure of the Milky Way. While the derived chemical abundances have been shown to be precise for most stars, some calibration problems have been reported, in particular for more metal- poor stars. Aims. In this paper, we aim to (1) re-determine the chemical abundances of the APOGEE+Kepler stellar sample (APOKASC) with an independent procedure, line list and line selection, and high quality surface gravity information from astroseismology, and (2) extend the abundance catalogue by including abundances that are not currently reported in the most recent APOGEE release (DR12). Methods. We fixed the Teff and log g to those determined using spectrophotometric and asteroseismic techniques, respectively. We made use of the Brussels Automatic Stellar Parameter (BACCHUS) code to derive the metallicity and broadening parameters for the APOKASC sample. In addition, we derived differential abundances with respect to Arcturus. Results. We have validated the BACCHUS code on APOGEE data using several well-known stars, and stars from open and globular clusters. We also provide the abundances of C, N, O, Mg, Ca, Si, Ti, S, Al, Na, Ni, Mn, Fe, K, P, Cr, Co, Cu, Rb, Yb and V for every star, line, and show the impact of line selection on the final abundances. These include abundances of five new elements and improved abundances for Si, Ti, S, and V. Conclusions. In this paper, we present an independent analysis of the APOKASC sample and provide abundances of up to 21 elements. This catalogue can be used not only to study chemical abundance patterns of the Galaxy but also to train data driven spectral approaches which can improve the abundance precision in a restricted dataset, but also full APOGEE sample.

astro-ph.GA↗

arXiv2016

The Gaia-ESO survey: Hydrogen lines in red giants directly trace stellar mass

Red giant stars are perhaps the most important type of stars for Galactic and extra-galactic archaeology: they are luminous, occur in all stellar populations, and their surface temperatures allow precise abundance determinations for many different chemical elements. Yet, the full star formation and enrichment history of a galaxy can be traced directly only if two key observables can be determined for large stellar samples - age and chemical composition. While spectroscopy is a powerful method to analyse the detailed abundances of stars, stellar ages are the "missing link in the chain", since they are not a direct observable. However, spectroscopy should be able to estimate stellar masses, which for red giants directly infer ages provided their chemical composition is known. Here we establish a new empirical relation between the shape of the hydrogen line in the observed spectra of red giants and stellar mass determined from asteroseismology. The relation allows to determine stellar masses and ages with the accuracy of 10-15%. The method can be used with confidence for stars in the following range of stellar parameters: 4000 < Teff < 5000 K, 0.5 < log g < 3.5, -2.0 < [Fe/H] < 0.3, and luminosities log L/LSun < 2.5. Our analysis provides observational evidence that the Halpha spectral characteristics of red giant stars are tightly correlated with their mass and therefore their age. We also show that the method samples well all stellar populations with ages above 1 Gyr. Targeting bright giants, the method allows to obtain simultaneous age and chemical abundance information far deeper than would be possible with asteroseismology, extending the possible survey volume to remote regions of the Milky Way and even to neighbouring galaxies like Andromeda or the Magellanic Clouds already with present instrumentation, like VLT and Keck facilities.

astro-ph.SR↗

arXiv2016

Gaia FGK Benchmark Stars: New Candidates At Low-Metallicities

We have entered an era of large spectroscopic surveys in which we can measure, through automated pipelines, the atmospheric parameters and chemical abundances for large numbers of stars. Calibrating these survey pipelines using a set of "benchmark stars" in order to evaluate the accuracy and precision of the provided parameters and abundances is of utmost importance. The recent proposed set of Gaia FGK benchmark stars of Heiter et al. (2015) has no recommended stars within the critical metallicity range of $-2.0 <$ [Fe/H] $< -1.0$ dex. In this paper, we aim to add candidate Gaia benchmark stars inside of this metal-poor gap. We began with a sample of 21 metal-poor stars which was reduced to 10 stars by requiring accurate photometry and parallaxes, and high-resolution archival spectra. The procedure used to determine the stellar parameters was similar to Heiter et al. (2015) and Jofre et al. (2014) for consistency. The effective temperature (T$_{\mathrm{eff}}$) of all candidate stars was determined using the Infrared Flux Method utilizing multi-band photometry. The surface gravity (log g) was determined through fitting stellar evolutionary tracks. The [Fe/H] was determined using four different spectroscopic methods fixing the T$_{\mathrm{eff}}$ and log g from the values determined independent of spectroscopy. We discuss, star-by-star, the quality of each parameter including how it compares to literature, how it compares to a spectroscopic run where all parameters are free, and whether Fe I ionisation-excitation balance is achieved. From the 10 stars, we recommend a sample of five new metal-poor benchmark candidate stars which have consistent T$_{\mathrm{eff}}$ , log g, and [Fe/H] determined through several means. These stars can be used for calibration and validation purpose of stellar parameter and abundance pipelines and should be of highest priority for future interferometric studies.

astro-ph.SR↗

arXiv2016

The Cannon 2: A data-driven model of stellar spectra for detailed chemical abundance analyses

We have shown that data-driven models are effective for inferring physical attributes of stars (labels; Teff, logg, [M/H]) from spectra, even when the signal-to-noise ratio is low. Here we explore whether this is possible when the dimensionality of the label space is large (Teff, logg, and 15 abundances: C, N, O, Na, Mg, Al, Si, S, K, Ca, Ti, V, Mn, Fe, Ni) and the model is non-linear in its response to abundance and parameter changes. We adopt ideas from compressed sensing to limit overall model complexity while retaining model freedom. The model is trained with a set of 12,681 red-giant stars with high signal-to-noise spectroscopic observations and stellar parameters and abundances taken from the APOGEE Survey. We find that we can successfully train and use a model with 17 stellar labels. Validation shows that the model does a good job of inferring all 17 labels (typical abundance precision is 0.04 dex), even when we degrade the signal-to-noise by discarding ~50% of the observing time. The model dependencies make sense: the spectral derivatives with respect to abundances correlate with known atomic lines, and we identify elements belonging to atomic lines that were previously unknown. We recover (anti-)correlations in abundance labels for globular cluster stars, consistent with the literature. However we find the intrinsic spread in globular cluster abundances is 3--4 times smaller than previously reported. We deliver 17 labels with associated errors for 87,563 red giant stars, as well as open-source code to extend this work to other spectroscopic surveys.

astro-ph.SR↗

arXiv2015

Using chemical tagging to redefine the interface of the Galactic disk and halo

We present a chemical abundance distribution study in 14 $α$, odd-Z, even-Z, light, and Fe-peak elements of approximately 3200 intermediate metallicity giant stars from the APOGEE survey. The main aim of our analysis is to explore the Galactic disk-halo transition region between -1.20 $<$ [Fe/H] $<$ -0.55 as a means to study chemical difference (and similarities) between these components. In this paper, we show that there is an $α$-poor and $α$-rich sequence within both the metal-poor and intermediate metallicity regions. Using the Galactic rest-frame radial velocity and spatial positions, we further separate our sample into the canonical Galactic components. We then studied the abundances ratios, of Mg, Ti, Si, Ca, O, S, Al, C+N, Na, Ni, Mn, V, and K for each of the components and found the following: (1) the $α$-poor halo subgroup is chemically distinct in the $α$-elements (particularly O, Mg, and S), Al, C+N, and Ni from the $α$-rich halo, consistent with the literature confirming the existence of an $α$-poor accreted halo population; (2) the canonical thick disk and halo are not chemically distinct in all elements indicating a smooth transition between the thick disk and halo; (3) a subsample of the $α$-poor stars at metallicities as low as [Fe/H] $\sim$ -0.85 dex are chemically and dynamically consistent with the thin disk indicating that the thin disk may extend to lower metallicities than previously thought, and (4) that the location of the most metal-poor thin disk stars are consistent with a negative radial metallicity gradient. Finally, we used our analysis to suggest a new set of chemical abundance planes ([$α$/Fe], [C+N/Fe], [Al/Fe], and [Mg/Mn]) that may be able to chemically label the Galactic components in a clean and efficient way independent of kinematics.

astro-ph.GA↗

arXiv2014

On the Relative Ages of the $α$-Rich and $α$-Poor Stellar Populations in the Galactic Halo

We study the ages of $α$-rich and $α$-poor stars in the halo using a sample of F and G dwarfs from the Sloan Digital Sky Survey (SDSS). To separate stars based on [$α$/Fe], we have developed a new semi-empirical spectral-index based method and applied it to the low-resolution, moderate signal-to-noise SDSS spectra. The method can be used to estimate the [$α$/Fe] directly providing a new and widely applicable way to estimate [$α$/Fe] from low-resolution spectra. We measured the main-sequence turnoff temperature and combined it with the metallicities and a set of isochrones to estimate the age of the $α$-rich and $α$-poor populations in our sample. We found all stars appear to be older than 8 Gyr confirming the idea that the Galactic halo was formed very early on. A bifurcation appears in the age-metallicity relation such that in the low metallicity regime the $α$-rich and $α$-poor populations are coeval while in the high metallicity regime the $α$-rich population is older than the $α$-poor population. Our results indicate the $α$-rich halo population, which has shallow age-metallicity relation, was formed in a rapid event with high star formation, while the $α$-poor stars were formed in an environment with a slower chemical evolution timescale.

astro-ph.GA↗

arXiv2014

Pseudo Three-dimensional Maps of the Diffuse Interstellar Band at 862 nm

The diffuse interstellar bands (DIBs) are absorption lines observed in visual and near infrared spectra of stars. Understanding their origin in the interstellar medium (ISM) is one of the oldest problems in astronomical spectroscopy, as DIBs ahave been known since 1922. In a completely new approach to understanding DIBs, we combined information from nearly 500,000 stellar spectra obtained by the massive spectroscopic survey RAVE (Radial Velocity Experiment) to produce the first pseudo three-dimensional map of the strength of the DIB at 8620~Ångstroms covering the nearest 3 kiloparsecs from the Sun, and show that it follows our independently constructed spatial distribution of extinction by interstellar dust along the Galactic plane. Despite having a similar distribution in the Galactic plane, the DIB 8620 carrier has a significantly larger vertical scale height than the dust. Even if one DIB may not represent the general DIB population, our observations outline the future direction of DIB research.

astro-ph.GA↗

arXiv2014

New distances to RAVE stars

Probability density functions are determined from new stellar parameters for the distance moduli of stars for which the RAdial Velocity Experiment (RAVE) has obtained spectra with S/N>=10. Single-Gaussian fits to the pdf in distance modulus suffice for roughly half the stars, with most of the other half having satisfactory two-Gaussian representations. As expected, early-type stars rarely require more than one Gaussian. The expectation value of distance is larger than the distance implied by the expectation of distance modulus; the latter is itself larger than the distance implied by the expectation value of the parallax. Our parallaxes of Hipparcos stars agree well with the values measured by Hipparcos, so the expectation of parallax is the most reliable distance indicator. The latter are improved by taking extinction into account. The effective temperature absolute-magnitude diagram of our stars is significantly improved when these pdfs are used to make the diagram. We use the method of kinematic corrections devised by Schoenrich, Binney & Asplund to check for systematic errors for general stars and confirm that the most reliable distance indicator is the expectation of parallax. For cool dwarfs and low-gravity giants tends to be larger than the true distance by up to 30 percent. The most satisfactory distances are for dwarfs hotter than 5500 K. We compare our distances to stars in 13 open clusters with cluster distances from the literature and find excellent agreement for the dwarfs and indications that we are over-estimating distances to giants, especially in young clusters.

astro-ph.GA↗

arXiv2014

Galactic kinematics and dynamics from RAVE stars

We analyse the kinematics of ~400000 RAVE stars. We split the sample into hot and cold dwarfs, red-clump and non-clump giants. The kinematics of the clump giants are consistent with being identical with those of non-clump giants. We fit Gaussian velocity ellipsoids to the meridional-plane components of velocity of each star class and give formulae from which the shape and orientation of the velocity ellipsoid can be determined at any location. The data are consistent with the giants and the cool dwarfs sharing the same velocity ellipsoids; sigma_z rises from 21 kms in the plane to sim 55 kms at |z|=2 kpc, while sigma_r rises from 37 kms to 82 kms. At (R,z) the longest axis of one of these velocity ellipsoids is inclined to the Galactic plane by an angle ~0.8 arctan(z/R). We use a novel formula to obtain precise fits to the highly non-Gaussian distributions of v_phi components. We compare the observed velocity distributions with the predictions of a dynamical model fitted to the velocities of stars that lie within ~150 pc of the Sun and star counts towards the Galactic pole. The model accurately reproduces the non-Gaussian nature of the v_r and v_z distributions and provides excellent fits to the data for v_z at all locations. The model v_phi distributions for the cool dwarfs fit the data extremely well, while those for the hot dwarfs have displacements to low v_phi that grow with |z| from very small values near the plane. At |z|>0.5 kpc, the theoretical v_phi distributions for giants show a deficit of stars with large v_phi and the model v_r distributions are too narrow. Systematically over-estimating distances by 20 per cent introduces asymmetry into the model v_r and v_z distributions near the plane and but significantly improves the fits to the data at |z|>0.5 kpc. The quality of the fits lends credence to the assumed, disc-dominated, gravitational potential.

astro-ph.GA↗