gr8stars II : judgement day for spectroscopic parameter model systematics

Technical Summary: gr8stars II: judgement day for spectroscopic parameter model systematics

Problem Statement
The search for Earth-like exoplanets around solar-like stars is currently limited not by the lack of such planets, but by the ability to disentangle small planetary signals from stellar noise. This requires precise and accurate stellar parameters (effective temperature $T_{\text{eff}}$, surface gravity $\log g$, and metallicity [Fe/H]). However, quoted precision uncertainties in spectroscopic analyses often fail to capture the extent of systematic biases introduced by differing methodologies, model atmospheres, analysis pipelines, and line lists. While previous studies have highlighted these systematics, there remains a gap in analyses where multiple state-of-the-art methods are applied to the exact same set of high-quality spectra to isolate methodological differences. Furthermore, the propagation of these atmospheric parameter systematics into derived stellar properties (mass, radius, age) and subsequently into exoplanetary parameters has not been fully quantified.

Methodology
This study utilizes a subset of 585 bright, single, slow-rotating FGK dwarfs from the gr8stars catalogue, observed exclusively with the SOPHIE spectrograph to eliminate instrumental scatter. The authors apply five independent spectroscopic analysis methods to each star:

1. ARES+MOOG: Equivalent width (EW) measurement combined with ionization/excitation equilibrium using Kurucz models.
2. FASMA: Spectral synthesis using MOOG and MARCS models with Levenberg–Marquardt optimization.
3. Metal Pipe: A C++/Python framework using line-by-line spectral synthesis with Phoenix models and MIST isochrones.
4. PAWS: A pipeline using the iSpec framework, combining EW (WIDTH code) and spectral synthesis (SPECTRUM code) with ATLAS9 models.
5. YARARA: A machine-learning approach mapping 16 extracted EWs to atmospheric parameters, calibrated on HARPS data.
6. webSME: Used specifically for a focused comparison of $\log g$ using the Mg I b triplet region.

For each star, the authors determine $T_{\text{eff}}$, $\log g$, and [Fe/H]. These spectroscopic inputs, combined with Gaia DR3 parallaxes and photometric magnitudes (2MASS, AllWISE), are fed into the isochrones package (using MIST models) to derive stellar mass, radius, and age. The study compares these results against the PASTEL catalogue, interferometric measurements for benchmark stars, and Spectral Energy Distribution (SED) fitting results from the first gr8stars paper.

Key Results

• Systematic Scatter: The study finds a typical scatter (RMSD) of 76 K in $T_{\text{eff}}$, 0.14 dex in $\log g$, and 0.07 dex in [Fe/H] across the different methods. These deviations significantly exceed the average internal precision errors quoted by the individual methods.
• Method-Specific Biases:
  • $\log g$: Spectroscopic methods generally underestimate $\log g$ compared to PASTEL values (mean difference of -0.09 dex). PAWS and ARES+MOOG show the largest underestimations (-0.22 and -0.16 dex, respectively). The authors find that isochrone-derived $\log g$ (constrained by photometry and parallax) provides more consistent agreement with literature values than spectroscopic $\log g$.
  • $T_{\text{eff}}$: FASMA shows the least bias relative to PASTEL. Metal Pipe and PAWS exhibit negative correlations between $T_{\text{eff}}$ differences and stellar temperature, suggesting calibration issues for hotter stars.
  • $v_{\text{mic}}$: Microturbulent velocity shows method-dependent systematics of 0.1–0.3 km s$^{-1}$, exceeding internal uncertainties.
• Derived Parameters: The scatter in spectroscopic inputs propagates to derived stellar parameters, resulting in typical scatters of 0.02 $R_{\odot}$ (1.22%) in radius, 0.05 $M_{\odot}$ (5.50%) in mass, and 2.08 Gyr in age.
• Exoplanetary Impact:
  • Radius and Mass: The induced fractional uncertainties in planetary radius ($\lesssim 3\%$) and mass ($\lesssim 5\%$) due to stellar parameter scatter are found to be lower than typical uncertainties reported in the literature.
  • Equilibrium Temperature: The study estimates a lower limit on the fractional uncertainty of planetary equilibrium temperature ($T_{\text{eq}}$) of $\approx 4\%$. This "noise floor" arises from the combination of $T_{\text{eff}}$ scatter and transit-derived radius uncertainties. The authors note that current literature often reports $T_{\text{eq}}$ uncertainties in the 2.0–3.5% range, suggesting these values are currently underestimated.

Significance and Claims
The paper asserts that the precision errors typically quoted for spectroscopic parameters are insufficient to reflect the true level of agreement between independent methods. The authors conclude that uncertainties for spectroscopically derived $T_{\text{eff}}$, $\log g$, and [Fe/H] must be inflated to at least 76 K, 0.14 dex, and 0.07 dex, respectively, to account for methodological systematics.

Regarding surface gravity, the study claims that spectroscopic $\log g$ is heavily influenced by methodological choices and model assumptions, and that isochrone-derived $\log g$ (incorporating photometric and parallax constraints) should be preferred for consistency.

While the scatter in stellar parameters does not currently dominate the uncertainty budget for exoplanetary mass and radius, the authors highlight that the 4% noise floor for planetary equilibrium temperature is a critical, often unrepresented limitation. As next-generation surveys (e.g., PLATO, EXPRES) push for higher precision in detecting Earth twins, these systematic floors will become increasingly significant. The paper modestly notes that its estimates represent a lower limit on systematics, as it relies on a single set of stellar evolution models (MIST); future work must incorporate multiple evolutionary grids to fully quantify systematic uncertainties.