Parameterizing Stellar Spectra Using Deep Neural Networks

Large-scale sky surveys are observing massive amounts of stellar spectra.The large number of stellar spectra makes it necessary to automatically parameterize spectral data,which in turn helps in statistically exploring properties related to the atmospheric parameters.This work focuses on designing an automatic scheme to estimate effective temperature(T_(eff)),surface gravity(log g) and metallicity[Fe/H] from stellar spectra.A scheme based on three deep neural networks(DNNs) is proposed.This scheme consists of the following three procedures:first,the configuration of a DNN is initialized using a series of autoencoder neural networks;second,the DNN is fine-tuned using a gradient descent scheme;third,three atmospheric parameters T_(eff),log g and [Fe/H] are estimated using the computed DNNs.The constructed DNN is a neural network with six layers(one input layer,one output layer and four hidden layers),for which the number of nodes in the six layers are 3821,1000,500,100,30 and 1,respectively.This proposed scheme was tested on both real spectra and theoretical spectra from Kurucz's new opacity distribution function models.Test errors are measured with mean absolute errors(MAEs).The errors on real spectra from the Sloan Digital Sky Survey(SDSS) are 0.1477,0.0048 and 0.1129 dex for log g,log T_(eff) and [Fe/H](64.85 K for T_(eff)),respectively.Regarding theoretical spectra from Kurucz's new opacity distribution function models,the MAE of the test errors are 0.0182,0.0011 and 0.0112 dex for log g,log T_(eff) and [Fe/H](14.90 K for T_(eff)),respectively.     

Validation of LAMOST stellar parameters with the PASTEL catalog

The Large Sky Area Multi-Object Fiber Spectroscopic Telescope(LAMOST) published its first data release(DRl)in 2013,which is currently the largest dataset of stellar spectra in the world.We combine the PASTEL catalog and SIMBAD radial velocities as a testing standard to validate stellar parameters(effective temperature T_(eff),surface gravity log g,metallicity[Fe/H]and radial velocity V_r)derived from DR1.Through cross-identification of the DR1 catalogs and the PASTEL catalog,we obtain a preliminary sample of 422 stars.After removal of stellar parameter measurements from problematic spectra and applying effective temperature constraints to the sample,we compare the stellar parameters from DR1 with those from PASTEL and SIMBAD to demonstrate that the DR1 results are reliable in restricted ranges of T_(eff).We derive standard deviations of 110 K,0.19 dex and 0.11 dex for T_(eff),log g and[Fe/H]respectively when T_(eff)8000 K,and 4.91 km S~(-1) for V_r when T_(eff)10 000 K.Systematic errors are negligible except for those of V_r.In addition,metallicities in DR1 are systematically higher than those in PASTEL,in the range of PASTEL[Fe/H]-1.5.     

Estimating stellar atmospheric parameters based on Lasso features

With the rapid development of large scale sky surveys like the Sloan Digital Sky Survey （SDSS）, GAIA and LAMOST （Guoshoujing telescope）, stellar spectra can be obtained on an ever-increasing scale. Therefore, it is necessary to estimate stel- lar atmospheric parameters such as Teff, log g and [Fe/H] automatically to achieve the scientific goals and make full use of the potential value of these observations. Feature selection plays a key role in the automatic measurement of atmospheric parameters. We propose to use the least absolute shrinkage selection operator （Lasso） algorithm to select features from stellar spectra. Feature selection can reduce redundancy in spectra, alleviate the influence of noise, improve calculation speed and enhance the robustness of the estimation system. Based on the extracted features, stellar atmospheric param- eters are estimated by the support vector regression model. Three typical schemes are evaluated on spectral data from both the ELODIE library and SDSS. Experimental results show the potential performance to a certain degree. In addition, results show that our method is stable when applied to different spectra.     

A Method of Estimating the [α/Fe] Ratios from Low-resolution Stellar Spectratwo

A method for the determination of [α/Fe] from low-resolution stellar spectra is presented. The proposed scheme includes the following three steps: firstly, the spectrum is decomposed by the multi-scale Haar wavelet, and the high-frequency components are removed to suppress the high-frequency noise; then, based on the correlation of the spectral data component with [α/Fe], the spectral features are selected by the LASSO (Least Absolute Shrinkage and Selection Operator) algorithm; finally, [α/Fe] is measured by the multiple linear regression method based on the MARCS stellar spectrum library. The effectiveness of the method is verified with the low-resolution stellar spectra of ELODIE, SDSS (Sloan Digital Sky Survey), LAMOST (Large Sky Area Multi-Object Fibre Spectroscopic Telescope), and four star clusters. The systematic deviations and accuracies are as follows: (0.04 dex, 0.064 dex) for the 317 ELODIE spectra; (0.16 dex, 0.065 dex) for the 412 SDSS spectra; (0.05 dex, 0.062 dex) for the 1276 LAMOST spectra (with the signal-noise ratio in the g band (SNRG) greater than 20). The averages of [α/Fe] obtained for the likely members of the globular star clusters (M13, M15) and open star clusters (NGC2420, M67) are in agreement with the literature values.     

Automatic determination of stellar atmospheric parameters and construction of stellar spectral templates of the Guoshoujing Telescope (LAMOST)

A number of spectroscopic surveys have been carried out or are planned to study the origin of the Milky Way. Their exploitation requires reliable automated methods and softwares to measure the fundamental parameters of the stars. Adopting the ULySS package,we have tested the effect of different resolutions and signal-tonoise ratios(SNR) on the measurement of the stellar atmospheric parameters(effective temperature Teff,surface gravity log g,and metallicity [Fe/H]) . We show that ULySS is reliable for determ...     

Analysis and calibration of Ca ii triplet spectroscopy of red giant branch stars from VLT/FLAMES observations

We demonstrate that low-resolution Ca  ii triplet (CaT) spectroscopic estimates of the overall metallicity ([Fe/H]) of individual red giant branch (RGB) stars in two nearby dwarf spheroidal galaxies (dSphs) agree to ±0.1–0.2 dex with detailed high-resolution spectroscopic determinations for the same stars over the range  −2.5 < [Fe/H] < −0.5  . For this study, we used a sample of 129 stars observed in low- and high-resolution modes with VLT/FLAMES in the Sculptor and Fornax dSphs. We also present the data-reduction steps we used in our low-resolution analysis and show that the typical accuracy of our velocity and CaT [Fe/H] measurement is ∼2 km s⁻¹ and 0.1 dex, respectively. We conclude that CaT–[Fe/H] relations calibrated on globular clusters can be applied with confidence to RGB stars in composite stellar populations over the range  −2.5 < [Fe/H] < −0.5  .     

High Resolution Spectroscopy of Halo Stars in the Near UV and Blue Region I. Spectra in the Wavelength Region 3550-5000 A

An atlas of high resolution (R = 60 000) CCD-spectra in the wavelength range 3500-5000A is presented for four objects in metallicity range -3.0 < [Fe/H] < -0.6, temperature range 4750 < Teff < 5900 K, and surface gravity range 1.6 < lgg < 5.0. We describe the calibration of the stellar atmospheric parameters using Alonso's formula based on the method of infrared flux and outline the determination of the abundances of a total number of 25 chemical elements. An analysis of the abundance determination errors for different chemical elements is carried out, and a method is provided for the observations and reduction of spectral material. Properties of the method of producing an atlas of spectra and line identifications are described.     

Empirical calibration of the near-infrared Ca ii triplet – II. The stellar atmospheric parameters

We present a homogeneous set of stellar atmospheric parameters ( T _eff, log  g , [Fe/H]) for a sample of about 700 field and cluster stars which constitute a new stellar library in the near-IR developed for stellar population synthesis in this spectral region ( λ 8350–9020) . Having compiled the available atmospheric data in the literature for field stars, we have found systematic deviations between the atmospheric parameters from different bibliographic references. The Soubiran, Katz & Cayrel sample of stars with very well determined fundamental parameters has been taken as our standard reference system, and other papers have been calibrated and bootstrapped against it. The obtained transformations are provided in this paper. Once most of the data sets were on the same system, final parameters were derived by performing error weighted means. Atmospheric parameters for cluster stars have also been revised and updated according to recent metallicity scales and colour–temperature relations.     

The chemical composition of the Small Magellanic Cloud

The Small Magellanic Cloud is a close, irregular galaxy that has experienced a complex star formation history due to the strong interactions occurred both with the Large Magellanic Cloud and the Galaxy. Despite its importance, the chemical composition of its stellar populations older than ∼ 1–2 Gyr is still poorly investigated. I present the first results of a spectroscopic survey of ∼ 200 Small Magellanic Cloud giant stars performed with FLAMES@VLT. The derived metallicity distribution peaks at [Fe/H] ∼ –0.9/–1.0 dex, with a secondary peak at [Fe/H] ∼ –0.6 dex. All these stars show [α /Fe] abundance ratios that are solar or mildly enhanced (∼+0.1 dex). Also, three metal‐poor stars (with [Fe/H] ∼ –2.5 dex and enhanced [α /Fe] ratios compatible with those of the Galactic Halo) have been detected in the outskirts of the SMC: These giants are the most metal‐poor stars discovered so far in the Magellanic Clouds. (© 2014 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Kinematics and stellar populations of the dwarf elliptical galaxy IC 3653

We present the first 3D observations of a diffuse elliptical galaxy (dE). The good quality data (S/N up to 40) reveal the kinematical signature of an embedded stellar disc, reminiscent of what is commonly observed in elliptical galaxies, though similarity of their origins is questionable. Colour map built from Hubble Space Telescope Advanced Camera for Surveys (ACS) images confirms the presence of this disc. Its characteristic scale (about 3 arcsec =250 pc) is about a half of galaxy's effective radius, and its metallicity is 0.1–0.2 dex larger than the underlying population. Fitting the spectra with synthetic single stellar populations (SSP), we found an SSP-equivalent age of 5 Gyr and nearly solar metallicity [Fe/H]  =−0.06  dex. We checked that these determinations are consistent with those based on Lick indices, but have smaller error bars. The kinematical discovery of a stellar disc in dE gives additional support to an evolutionary link from dwarf irregular galaxies due to stripping of the gas against the intracluster medium.     

A Model of the Circumstellar Envelope of Luminous Blue Variables

The continuum energy distributions of the luminous blue variables R127 and R110 in the outburst phase are fitted with a circumstellar envelope model.Both stars show two peaks in their continuum, one near 1250A and the other in the optical band. We suggest that their UV and optical fluxes may have different origins: the UV flux comes from the central star while the optical flux comes from an expanding circumstellar envelope. We construct a model for LBVs consisting of two LTE atmosphere models with different temperatures, and find it to be in agreement with the observed spectral energy distributions of R127 and R110.According to our numerical experiments, R127‘s continuum is composed of fluxes from a circumstellar envelope of Teff = 8000K, R = 485R⊙, and log g = 1,and from a central star of Tef = 17000K, R = 135R⊙, and log g = 2.5 with a permeating factor f = 0.5; while R110‘s continuum can be fitted by a circumstellar envelope of Teff =7000K, R = 350R⊙, and logg = 0.5, and a central star of Teff =25 000 K, R = 27R⊙, and log g 3.0 with a permeating factor f = 0.65.Both models show that the non-spherically symmetric, optically thick regions are formed surrounding the central star in the outburst phase. The light of the central star is shielded by the circumstellar envelope so that the visual brightness increases with the decrease/increase of the temperature/radius of the optically thick regions.     

Analysis of UBV photometry in selected area 133

We interpret the de‐reddened UBV data for the field SA 133 to deduce the stellar density and metallicity distribution functions. The logarithmic local space density for giants, D*(0) = 6.40, and the agreement of the luminosity function for dwarfs and sub‐giants with the one of Hipparcos confirms the empirical method used for their separation. The metallicity distribution for dwarfs gives a narrow peak at [Fe/H] = +0.13 dex, due to apparently bright limiting magnitude, V_o = 16.5, whereas late‐type giants extending up to z ∼ 4.5 kpc from the galactic plane have a multimodal distribution. The metallicity distribution for giants gives a steep gradient d[Fe/H]/dz = –0.75 dex kpc^–1 for thin disk and thick disk whereas a smaller value for the halo, i.e. d[Fe/H]/dz = –0.45 dex kpc^–1. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Precise determination of fundamental parameters of six exoplanet host stars and their planets

The aim of this paper is to determinate the fundamental parameters of six exoplanet host(EH) stars and their planets. Because techniques for detecting exoplanets yield properties of the planet only as a function of the properties of the host star, we must accurately determine the parameters of the EH stars first. For this reason, we constructed a grid of stellar models including diffusion and rotation-induced extra-mixing with given ranges of input parameters(i.e. mass, metallicity and initial rotation rate). In addition to the commonly used observational constraints such as the effective temperature Teff, luminosity L and metallicity [Fe/H], we added two observational constraints, the lithium abundance log N(Li) and the rotational period Prot.These two additional observed parameters can set further constraints on the model due to their correlations with mass, age and other stellar properties. Hence, our estimations of the fundamental parameters for these EH stars and their planets have a higher precision than previous works. Therefore, the combination of rotational period and lithium helps us to obtain more accurate parameters for stars, leading to an improvement in knowledge about the physical state of EH stars and their planets.     

Gas properties of H ii and starburst galaxies: relation with the stellar population

We study the gas emission of galaxies with active star formation, consisting mostly of H  ii and starburst galaxies, as well as some Seyfert 2 galaxies, and determine chemical and physical parameters. The data consist of 19 high signal-to-noise ratio optical templates, a result of grouping 185 emission-line galaxy spectra. Underlying stellar population models (from Raimann et al.) were subtracted from the templates in order to isolate the pure emission component.
We analyse the distribution of these improved signal-to-noise ratio emission spectra in diagnostic diagrams and find that the H  ii templates show a smaller spread in log([O  iii ]/H β ) values than the individual galaxies, apparently as a result of the population subtraction and a better signal-to-noise ratio. We thus suggest the template sequence as a fiducial observational locus for H  ii galaxies which can be used as reference for models. The sequence of line ratios presented by the H  ii galaxies in the diagram log([O  iii ] λ 5007/H β ) versus log([N  ii ] λ 6584/H α ) is primarily owing to the gas metallicity, of which the log([N  ii ]/H α ) ratio is a direct estimator. We also study the properties of the starburst galaxies and those intermediate between H  ii and starburst galaxies, which are more metal rich and sit on more massive galaxies.
We discuss the present results in the frame of a recently proposed equivalent-width diagnostic diagram for emission-line galaxies (by Rola et al.) and conclude that the observed ranges in W ([O  ii ])/ W (H β ) and W (H β ) are mostly owing to the non-ionizing stellar population contribution. We propose that W (H β ) be used as an estimator of this contribution to the continuum, and briefly discuss implications to the cosmological use of H  ii galaxies.     

High Resolution Optical Spectroscopy of Hot Post-AGB Star Candidates LS IV-04 1 and LB3116

We present LTE analysis of high resolution optical spectra for B-type hot PAGB stars LS IV-04 1 and LB3116 (LSE 237). The spectra of these high Galactic latitude stars were obtained with the 3.9-m Anglo-Australian Telescope (AAT) and the UCLES spectrograph. The standard 1D LTE analysis with line-blanketed LTE model atmospheres and spectral synthesis provided fundamental atmospheric parameters of T_eff= 15 000±1000 K, log g= 2.5±0.2, ξ = 5.0±1.0 km s^?1, [M/H] = ?1.81 dex, and v sin i= 5 km s^?1 for LSIV-04 1 and T_eff= 16 000±1000 K, log g= 2.5±0.1, v sin i= 25 km s^?1, and [Fe/H] = ?0.93 dex for LB 3116. Chemical abundances of ten different elements were obtained. For LS IV-04 1, its derived model temperature contradicts with previous analysis results. The upper limits for its nitrogen and oxygen abundances were reported for the first time. The magnesium, silicon and calcium were overabundant (i.e. [Mg/Fe] = 0.8 dex, [Si/Fe] = 0.5 dex, [Ca/Fe] = 0.9 dex). With its metal-poor photosphere and V_LSR ≈ 96 km s^?1, LSIV-04 1 is likely a population II star and most probably a PAGB star. LTE abundances of LB 3116 were reported for the first time. The spectrum of this helium rich star shows 0.9 dex enhancement in the nitrogen. The photosphere of the star is slightly deficient in Mg, Si, and S. (i.e. [Mg/Fe] = ?0.2 dex, [Si/Fe] = ?0.4 dex, [S/Fe] = ?0.2 dex). The Al is slightly enhanced. The phosphorus is overabundant, i.e. [P/Fe] ≈ 1.7 ± 0.47 dex, hence LB3116 may be the first example of a PAGB star which is rich in phosphorus. With its high radial velocity (i.e.V_LSR = 73 km s^?1), and the deficiencies observed in C, Mg, Si, and S indicate that LB 3116 is likely a hot PAGB star at high galactic latitude.     

Metallicity calibration for solar type stars based on red spectra

Based on a high resolution and high signal-to-noise ratio(S/N) spectral analysis of 90 solar-type stars,we have established several new metallicity calibrations in the Teff range [5600,6500]K based on red spectra with the wavelength range of 560-880nm.The new metallicity calibrations are applied to determine the metallicity of solar analogs selected from Sloan Digital Sky Survey(SDSS) spectra.There is a good consistent result with the adopted value presented in SDSS-DR7 and a small scatter of 0.26dex for st...     

中等贫金属星高分辨率光谱分析Ⅱ.化学元素丰度

在第1篇论文的基础上，确定了样本星的恒星大气参数，得到这些星中9种元素的丰度。讨论了各种元素丰度随[Fe/H]的变化。平均的[Na/Fe]～－0．01dex，接近于太阳丰度。α元素Si和Ca具有几乎相同的丰度模式，而[Ti/Fe]弥散较大，但三者均有随[Fe/H]的减小而增加的趋势。铁峰元素V、Cr、Ni在不同丰度处有较大的弥散，[Cr/Fe]在所有样本星中均表现超丰；而[Mn/Fe]却明显过贫，且随金属丰度的增加而增加。     

Optical spectrum of the bipolar nebula AFGL 2688

Based on echelle spectra obtained at the prime focus of the 6-m telescope, we have determined for the first time the detailed chemical composition of one of the components of the bipolar nebula identified with the intense infrared source AFGL 2688 by the model-atmosphere method. The iron abundance [Fe/H]=?0.59 dex derived for AFGL 2688 suggests that the object probably belongs to an intermediate population of the Galaxy. The stellar atmosphere exhibits high carbon and nitrogen abundances, [C/Fe]=+0.73 and [N/Fe]=+2.00, and C/O>1, confirming that the object is at the post-AGB stage. However, the detected overabundance of s-process elements (yttrium and barium) relative to iron is modest: [X/Fe]=+0.55. The lanthanides are even less enhanced: for La, Ce, Pr, and Nd, the mean abundance relative to iron is [la/Fe]=+0.26. This behavior of the heavy metals is consistent with the low intensity of the 21-μm band in the infrared spectrum of AFGL 2688; the intensity of this emission band is great in the spectra of all the studied PPN with large overabundances of s-process elements. An analysis of the radial velocities measured from spectral features originating in the atmosphere and in the circumstellar shell has revealed a high-velocity (～60 km s^?1) component of the stellar wind from AFGL 2688.     

Velocity and abundance precisions for future high‐resolution spectroscopic surveys: A study for 4MOST

In preparation for future, large‐scale, multi‐object, high‐resolution spectroscopic surveys of the Galaxy, we present a series of tests of the precision in radial velocity and chemical abundances that any such project can achieve at a 4 m class telescope. We briefly discuss a number of science cases that aim at studying the chemo‐dynamical history of the major Galactic components (bulge, thin and thick disks, and halo) – either as a follow‐up to the Gaia mission or on their own merits. Based on a large grid of synthetic spectra that cover the full range in stellar parameters of typical survey targets, we devise an optimal wavelength range and argue for a moderately high‐resolution spectrograph. As a result, the kinematic precision is not limited by any of these factors, but will practically only suffer from systematic effects, easily reaching uncertainties <1km s^–1. Under realistic survey conditions (namely, considering stars brighter than r = 16 mag with reasonable exposure times) we prefer an ideal resolving power of R ∼20 000 on average, for an overall wavelength range (with a common two‐arm spectrograph design) of [395;456.5] nm and [587;673] nm. We show for the first time on a general basis that it is possible to measure chemical abundance ratios to better than 0.1 dex for many species (Fe, Mg, Si, Ca, Ti, Na, Al, V, Cr, Mn, Co, Ni, Y, Ba, Nd, Eu) and to an accuracy of about 0.2 dex for other species such as Zr, La, and Sr. While our feasibility study was explicitly carried out for the 4MOST facility, the results can be readily applied to and used for any other conceptual design study for high‐resolution spectrographs. (© 2013 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Radio emission at 1.4 GHz from luminous compact galaxies

We analyse the properties of the 1.4 GHz continuum emission in 52 star‐forming compact luminous galaxies (LCGs). The fluxes of the 1.4 GHz thermal (free‐free) component are derived from the extinction‐ and aperture‐corrected fluxes of the Hα emission line. The fraction of the thermal 1.4 GHz emission is shown tobe in the range of 2–52 % with a median value of about 15–17 %, and its distribution is close to a log‐normal one. Both the thermal and non‐thermal radio continuum luminosities are shown to be proportional to the masses of the young stellar population of the LCGs. We discuss the approximation of the ratio of the 1.4 GHz luminosity‐to‐the mass of the young stellar population as a function of starburst age. The derived approximation is shown to be similar but less clearly defined compared to ones obtained previously for LCGs from their Hα and UV continuum luminosities. (© 2015 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)