Metal Pipe Algorithm for Stellar Spectroscopy

• Metal Pipe algorithm is a computational pipeline that integrates photometric SED fitting and spectral synthesis to derive detailed stellar parameters and elemental abundances.
• It leverages high-resolution spectra, curated line lists, and isochrone grids to enable homogeneous population-level stellar characterization across FGK stars.
• The method benchmarks against established catalogs, achieving precision in T_eff, log g, and elemental abundances, and incorporates NLTE corrections for refined accuracy.

The Metal Pipe algorithm refers to a robust, broadly applicable computational pipeline for deriving stellar atmospheric parameters and elemental abundances from high-resolution stellar spectra, in conjunction with photometric and astrometric data. The methodology integrates isochronal modeling for fundamental stellar properties and spectral synthesis for precision abundance analysis. Its design enables application across a wide range of spectral types and supports homogeneous population-level stellar characterization, which is critical for exoplanet host analysis and galactic archaeology. Metal Pipe combines atmospheric modeling, photometric SED fitting, MOOG-driven spectral synthesis, and a curated spectral linelist to output $T_{\rm eff}$, $\log g$, $M_*$, $R_*$, $L_*$, and abundances for C, O, Na, Mg, Al, Si, S, Ca, Ti, Fe. The algorithm has been benchmarked against established catalogs using hundreds of F, G, and K stars, demonstrating performance at the level of contemporary LTE spectroscopic analyses (Kolecki et al., 15 Jan 2026).

1. Algorithmic Workflow and Data Requirements

Metal Pipe requires as input:

• A normalized, high-resolution (R ≳ 45,000), high-S/N (≳70) spectrum in ASCII (λ, flux, σ_flux), with continuum normalized to unity.
• Multi-band photometry (UBVRI from SIMBAD, JHK from 2MASS, G, BP, RP, and parallaxes from Gaia DR3).
• Curated spectral linelists.
• MIST isochrone grids (age × EEP).
• A PHOENIX model-atmosphere grid reformatted for compatibility with MOOG.

The pipeline’s primary sequential steps are:

1. Spectrum normalization: Median and maximum filtering, morphological closing for continuum identification, convex-hull gap filling, smoothing, and division out to produce a normalized spectrum.
2. Photometric SED fitting: Conversion of apparent to absolute magnitudes using Gaia parallaxes, then computation of reduced $\chi^2$ over the isochrone grid for all bands:

$$\chi^2_\nu(x, y) = \frac{1}{n-2} \sum_b \frac{(M^b_{\rm obs} - M^b_{\rm mod}(x, y))^2}{\sigma_b^2}$$

Normalized probabilities $P(x, y)$ are computed as $P(x, y) = A\exp[-\frac{1}{2}\chi^2_\nu(x, y)]$. Isochronal parameters ($T_{\rm eff}$, $\log g$, $M_*$, $R_*$, $L_*$) are sampled (2,000 draws), with medians and 1σ uncertainties adopted.

1. Model-atmosphere interpolation: PHOENIX atmosphere models are interpolated using the estimated stellar parameters and initial [M/H], [$\alpha$/M].
2. Abundance loop: Iterative spectral synthesis and fitting using MOOG:
   • Fit [Fe/H] using individual Fe lines, minimizing the weighted

   $$\chi^2_\nu = \frac{1}{n_p-3} \sum_\lambda w(\lambda)\frac{(f_{\rm obs}-f_{\rm syn})^2}{\sigma^2_{\rm obs}+\sigma^2_{\rm syn}}$$

   Discard outlier lines ($\chi^2_\nu > 5$ or no minimum). - Iterate global [M/H] and [α/M] using Ca and Ti lines, repeating atmospheric interpolation as needed. - Fit abundances for C, O, Na, Mg, Al, Si, S, using the same formalism. - NLTE corrections are applied to O, Ca, Ti following published grids.

3. Final reporting: All best-fit parameters and uncertainties are output, including broadening velocity ($v_{\rm broad}$).

Pseudocode for the entire procedure is provided in condensed form in the original work (Kolecki et al., 15 Jan 2026).

2. Isochronal Parameter Estimation

Stellar parameters $T_{\rm eff}$ and $\log g$ are not assumed from spectroscopy but extracted by grid-based SED fitting over isochronal models (MIST):

• The isochronal grid is evaluated in $(x, y)$ for each photometric band, calculating $\chi^2_\nu$ as above.
• Probabilities $P(x, y)$ yield the likelihood surface.
• 2,000 points are sampled via rejection sampling from $P(x, y)$; distributions of $T_{\rm eff}$, $\log g$, $M_*$, $R_*$, $L_*$ are constructed. Medians are taken as best-fit values, with standard deviations as uncertainties.
• This step avoids nonlinear optimization beyond grid evaluation and naturally incorporates photometric uncertainties.

3. Spectral Line Selection and MOOG Coupling

Metal Pipe uses a carefully curated spectral line database:

• Starting from Linemake and VALD, lines are filtered by visual inspection and statistical criteria: retained if unblended in ≥80% of pilot spectra; then further selected by evaluating the [X/H] distribution for systematic offsets and outliers.
• Final linelists contain, e.g., ∼400 Fe I/II lines and as few as three lines for challenging species (e.g., O I).
• MOOG (2019, “synth” mode) is called as a backend solver for spectroscopic line fitting, with v_broad and element abundance as parameters. Microturbulence is fixed at 1.5 km s⁻¹.
• For each element, lines are fit independently, and medians of accepted lines (passing $\chi^2_\nu$ and convergence criteria) are reported as final abundances.

4. Uncertainty Modeling and Error Propagation

Uncertainties are quantified as follows:

• For isochronal parameters ($T_{\rm eff}$, $\log g$, $M_*$, $R_*$, $L_*$), uncertainties correspond to the standard deviations of the 2,000 SED samples.
• For elemental abundances [X/H], formal errors report the standard deviation across lines divided by $\sqrt{N_{\rm lines}}$:

$$\sigma([X/H]) = \frac{\rm std\ deviation}{\sqrt{N_{\rm lines}}}$$

with a catalog-wide error floor of $\sim$0.10 dex adopted.

• Covariances between parameters and abundances are not yet propagated; future versions will implement error mapping by perturbing input parameters and re-running abundance determination.

5. Performance Benchmarks and Validation

Performance of Metal Pipe has been assessed using a sample of 503 FGK stars with archival Keck/HIRES data:

• Parameter-level agreement with Brewer et al. (2016): $\Delta T_{\rm eff}$ median = +23 K, RMS = 88 K; $\Delta\log g$ median = 0.00 dex, RMS = 0.07 dex.
• Against Adibekyan et al. (2012): $\Delta T_{\rm eff}$ median = +11 K, RMS = 122 K; $\Delta\log g$ median = –0.01 dex, RMS = 0.12 dex.
• Abundance RMS scatter per element vs. catalogs: [C/H] ≈ 0.09 dex, [O/H] ≈ 0.12 dex, [Na/Mg/Al/Si/Ca/Ti/Fe] ≈ 0.07–0.12 dex.
• These discrepancies are within the expected uncertainty floor for LTE spectroscopic methods (∼50–100 K for $T_{\rm eff}$, ∼0.05–0.10 dex for abundances) (Kolecki et al., 15 Jan 2026).

6. Distinctive Features and Applicability Scope

Metal Pipe’s core methodological innovations are:

• Integration of isochronal photometry-based determination of fundamental properties, circumventing the bias inherent in pure spectroscopic parameter estimation.
• Iterated coupling with MOOG for converged global metallicity and alpha enhancement, followed by precision multi-element abundance extraction.
• Robust line selection pipeline to minimize blending and systematic elemental offsets.
• Direct applicability across a broader spectral-type range than previous catalog pipelines.
• Open workflow for adaptation to non-FGK spectral classes (e.g., late K, M dwarfs are anticipated in future work).

A plausible implication is that Metal Pipe’s isochronal anchoring may be better suited to population-level and cross-survey analyses than pipelines that rely exclusively on spectrum-only fitting.

7. Limitations and Future Prospects

The current version:

• Does not fully propagate parameter–abundance covariances.
• Adopts fixed microturbulence, which may limit accuracy for evolved/luminous stars.
• Is tuned and benchmarked primarily for FGK samples; further validation for M dwarfs and peculiar stars is upcoming.
• Outlier rejection in line selection relies on median statistics; pathological line blends or spectral artifacts may escape this filter.

Future developments indicated by the authors include improved covariance modeling, extended elemental grids, and calibration for cool dwarfs and giants. The underlying methodological formalism, equations, and pseudocode provided in (Kolecki et al., 15 Jan 2026) permit independent implementation and critique.