Source-linked AI summary
ULySS: A Full Spectrum Fitting Package
Mina Koleva, Philippe Prugniel, Antoine Bouchard, Yue Wu
TL;DR
ULySS addresses the need for an accessible full-spectrum fitting package for stellar atmospheric parameters and stellar-population histories. It fits observations with flexible model components, velocity distributions, and continuum terms while mapping degeneracies and estimating errors. The package supports robust analysis for simple models, whereas complex models require Monte-Carlo validation.
Problem
Automated full-spectrum analysis is needed to extract stellar atmospheric parameters and stellar-population histories from spectroscopic data while using the distributed signal.
Method
ULySS fits spectra with linear combinations of non-linear components, optional LOSVD convolution, and a multiplicative polynomial, using parameter-space maps and Monte-Carlo simulations.
Results
ULySS provides full-spectrum analysis in which simple models often have a unique or recognizable minimum, while complex models require Monte-Carlo simulations to assess solution validity.
Takeaways & Limitations
The public package is simple and flexible, uses the full signal, and supports joint analysis of stellar kinematics and population mixtures.
Takeaways & Limitations
Complex fits can contain statistically indistinguishable local minima, so a less probable solution cannot always be objectively rejected.
Abstract
from arXiv · showhide
Aims. We provide an easy-to-use full-spectrum fitting package and explore its applications to (i) the determination of the stellar atmospheric parameters and (ii) the study of the history of stellar populations. Methods. We developed ULySS, a package to fit spectroscopic observations against a linear combination of non-linear model components convolved with a parametric line-of-sight velocity distribution. The minimization can be either local or global, and determines all the parameters in a single fit. We use chi2 maps, convergence maps and Monte-Carlo simulations to study the degeneracies, local minima and to estimate the errors. Results. We show the importance of determining the shape of the continuum simultaneously to the other parameters by including a multiplicative polynomial in the model (without prior pseudo-continuum determination, or rectification of the spectrum). We also stress the benefice of using an accurate line-spread function, depending on the wavelength, so that the line-shape of the models properly match the observation. For simple models, i. e., to measure the atmospheric parameters or the age/metallicity of a single-age stellar population, there is often a unique minimum, or when local minima exist they can unambiguously be recognized. For more complex models, Monte-Carlo simulations are required to assess the validity of the solution. Conclusions. The ULySS package is public, simple to use and flexible. The full spectrum fitting makes optimal usage of the signal.
1. Introduction
ULySS is introduced as a public, user-friendly full-spectrum fitting package for stellar atmospheric parameters and the histories of stellar populations. It fits entire spectra with flexible models while addressing continuum shape, velocity effects, and parameter degeneracies.
- Motivation: ULySS responds to the data avalanche by providing an automated, objective alternative to interactive spectral-signature identification and analysis.Direct pixel-by-pixel comparison with models has applications in both stellar atmospheres and stellar-population histories.
- Motivation: Full-spectrum methods use the entire measured signal rather than selected spectral features, although this makes direct feature-to-physics relationships less simple.The paper attributes this complexity to redundant information distributed across broad wavelength ranges.
- Scope: ULySS applies full spectral fitting to stellar atmospheric parameters and galaxy star-formation and metal-enrichment histories.The package handles both contexts within a single framework because of similarities between them.
- Approach: ULySS compares observed spectra with linear combinations of non-linear model components, optionally incorporating velocity distributions and a multiplicative polynomial.The polynomial absorbs flux-calibration, extinction, and other spectral-shape errors without prior pseudo-continuum rectification.
- Dissemination: The package is publicly available and is presented with descriptions of its method, stellar-atmosphere applications, stellar-population examples, and conclusions.The paper explicitly organizes these topics across Sections 2–5.
- Scope: The package was broadened from stellar-population analysis to atmospheric parameters and other applications, with mathematical changes intended to improve precision, robustness, and performance.The authors also added documentation and tutorials for public distribution.
2. Description of the method and package
ULySS fits spectra in pixel space with composite non-linear models, flexible minimization, and tools for examining degeneracies and solution reliability. The package emphasizes simultaneous parameter estimation, accurate instrumental modeling, and cautious treatment of continuum and additive terms.
- Core model and minimization: ULySS minimizes pixel-space χ2 between observations and models generated at matching resolution and sampling, using all spectral bins and their errors.The model combines non-linear components and can include LOSVD convolution and polynomial terms.
- Core model and minimization: A single fit determines all free parameters together, helping handle degeneracies such as temperature–metallicity coupling.This avoids the additional complexity of sequential estimation and iteration described for stepwise procedures.
- Core model and minimization: The model uses a LOSVD parameterized by systemic velocity and dispersion, optionally including h3 and h4, while components encode context-specific physical parameters.Examples include Teff, surface gravity, and [Fe/H] for stellar atmospheres, or age and abundance parameters for stellar populations.
- Continuum and instrumental treatment: The multiplicative polynomial absorbs flux-calibration and extinction effects, replacing prior pseudo-continuum rectification; additive terms require caution because they can bias component parameters.The paper recommends simulations to evaluate additive-term effects rather than relying only on reduced χ2.
- Continuum and instrumental treatment: Matching the wavelength-dependent LSF smooths residuals and primarily improves LOSVD measurements, while neglecting LSF shifts can bias velocity dispersion for distant galaxies.For SDSS, the cited bias for σ = 50 km s−1 is about 0.1 km s−1 at z=0.03 and 2 km s−1 at z=0.4.
- Core model and minimization: ULySS supports local or global minimization and provides Monte-Carlo simulations, convergence maps, and χ2 maps to explore parameter-space structure.These tools address sensitivity to starting guesses, degeneracies, local minima, and uncertainty assessment.
3. Determination of stellar atmospheric parameters
ULySS determines stellar atmospheric parameters by fitting spectra with interpolated stellar models and evaluates solution stability, convergence, and agreement with independent measurements.
- Method: ULySS derives Teff, log(g), and [Fe/H] by comparing observed spectra with a parametric interpolator of stellar reference spectra.The TGM component performs the minimization over the three atmospheric parameters, with the current ELODIE library covering 3 600 K < Teff < 30 000 K.
- Comparison with literature: For the comparison sample, Δ(Teff)/Teff = −0.013 ± 0.010, Δ(log(g)) = 0.14 ± 0.22, and Δ([Fe/H]) = 0.01 ± 0.11.Excluding the discrepant M star, the metallicity difference becomes Δ([Fe/H]) = −0.01 ± 0.07.
- Comparison with literature: ULySS temperatures were systematically cooler by 60 K than da Silva et al. (2006), while the reported program errors were about 20 times smaller than the external errors.The 60 K offset was consistent with a temperature offset reported by da Silva et al. in their own literature comparison.
- Multiplicative polynomial continuum: Including a multiplicative polynomial determines continuum normalization during the fit instead of requiring prior pseudo-continuum rectification.The polynomial absorbs flux-calibration and extinction effects while allowing the physical parameters and continuum shape to be fitted together.
- Multiplicative polynomial continuum: Atmospheric parameters generally reached stable plateaus at polynomial degrees n = 10 to 15 for F, G, K, and O stars, and n = 35 for the M star.The A1Ib star HD 195324 showed persistent metallicity dependence on n, attributed to limitations of the sparsely populated ELODIE interpolator and continuum treatment.
- Multiplicative polynomial continuum: Parameter variations with polynomial degree exceeded the formal error bars, with typical dispersions about twice the reported errors.For Teff, log(g), and [Fe/H], the typical errors were 0.1%, 0.006, and 0.005, versus dispersions of 0.2%, 0.01, and 0.01 for n > 20.
- Reliability assessment: Convergence maps, χ2 maps, and Monte-Carlo simulations assess degeneracies, local minima, and error reliability before applying the method to larger datasets.For HD 76151, guesses with Teff < 10000 K converged to the correct solution, whereas hotter guesses could reach an unphysical local minimum.
4. History of stellar populations
ULySS reconstructs stellar-population histories using SSP components, while Monte-Carlo simulations and χ2 maps expose degeneracies and solution validity. Tests on NGC 205 and simulated extended histories show useful recovery, but increasingly complex histories broaden uncertainties and create ambiguous minima.
- Stellar-population modelling: SSP-equivalent parameters provide luminosity-weighted estimates dominated by the burst contributing most of the light, while detailed SFH recovery is practically limited to 2 to 4 epochs.A single SSP fit uses age, [Fe/H], and possibly [Mg/Fe]; complex histories can instead be decomposed into several SSP components.
- Stellar-population modelling: Direct fitting of many SSPs is unstable because component degeneracies require regularization, so ULySS divides the time axis into bounded intervals and increases components progressively.As free parameters increase, local minima appear; global minimization and χ2 maps help characterize the parameter space.
- NGC 205: For NGC 205, the two-component fit finds a young population of approximately 130 Myr contributing about 25% of the light and 7% of the mass.The young and old populations are well separated in Monte-Carlo simulations, although the two-burst hypothesis is explicitly tested rather than established independently.
- NGC 205: Monte-Carlo solutions agree with the direct two-burst fit but have significantly larger errors because of degeneracies between ages and component weights.A secondary cloud containing 10% of solutions occurs near 3 Gyr and [Fe/H]=-1, and its associated local minimum cannot be statistically distinguished from the primary solution.
- Extended star-formation histories: For simulated constant and exponentially declining SFRs, reconstructed histories reproduce the inputs, while Monte-Carlo solutions agree with direct fits within uncertainties.The direct solution underestimates error bars because ages and component weights are degenerate; individual realizations can span the full allowed age ranges.
- Extended star-formation histories: The simulations recover more than 10% of the light in the youngest component even where the simulated SFR is lowest.This result applies to both tested extended star-formation scenarios.
5. Conclusion
The paper presents ULySS as a simple, efficient package for fitting stellar-atmosphere and stellar-population spectra. It supports parameter recovery, SFH decomposition, and detailed assessment of convergence, degeneracies, and uncertainties.
- Conclusion: ULySS determines stellar atmospheric parameters and reconstructs galaxy or cluster star-formation histories by decomposing spectra into SSPs.The package is presented as simple to use and efficient for both applications.
- Conclusion: Convergence regions and degeneracies can be examined in detail, allowing robust determination of estimated-parameter errors.The conclusion links this assessment to the package’s atmospheric-parameter application.
- Conclusion: Simultaneously fitting stellar kinematics and population mixtures can break degeneracies and improve reliability and precision for both kinematics and star-formation histories.This conclusion is stated for the joint analysis of kinematics and stellar mix.
- Conclusion: ULySS is publicly downloadable, includes additional components such as LINE for emission-line fitting, and can be extended to other applications.The package is available from the ULySS website.