Syntriod: A Robust Initial Parameter Estimator for Radial Velocity Curve Solutions Beyond Conventional Sampling Limits

Technical Summary of "Syntriod: A Robust Initial Parameter Estimator for Radial Velocity Curve Solutions Beyond Conventional Sampling Limits"

Problem Statement
Determining reliable orbital parameters for spectroscopic binary systems (SB1 and SB2) and exoplanet hosts is a critical challenge in modern astronomy, particularly as large-scale surveys like GAIA, SDSS, LAMOST, and TESS generate vast datasets with sparse and irregularly sampled radial velocity (RV) measurements. While full Bayesian inference frameworks (e.g., rvfit, RadVel, juliet) are powerful for exploring complex parameter spaces, their efficiency and convergence heavily depend on the availability of accurate initial parameter estimates.

Traditional period-search techniques, such as the Lomb–Scargle (LS) periodogram, often struggle in these regimes. LS assumes a sinusoidal signal model, making it prone to aliasing and incorrect period identification in eccentric systems or datasets with irregular temporal sampling. Furthermore, when the number of observations ($N_{obs}$) approaches the dimensionality of the orbital parameter space (e.g., $N_{obs} \le 6$ for a single-lined binary), the problem becomes formally underconstrained. In these sparse regimes, classical methods frequently fail due to aliasing, while full Bayesian approaches become computationally inefficient due to the vast parameter space required for exploration.

Methodology: The Syntriod Algorithm
The authors present Syntriod, an orbital-phase domain RV template-based algorithm designed to provide robust initial orbital parameter estimates across a wide range of sampling conditions. Rather than performing full orbital inference, Syntriod constrains the parameter space by comparing observations against a precomputed library of Keplerian radial-velocity templates.

Key methodological components include:

• Adaptive Workflow: The algorithm dynamically adjusts its strategy based on the number of available observations ($N_{obs}$).
  • $N_{obs} > 5$: Syntriod employs an adaptive period-search strategy combined with phase-domain template matching. It maximizes a penalized objective function ($\ln L_{eff}$) that includes terms for orbital likelihood, differential velocity consistency (for SB2), phase coverage, and penalties for unrealistic eccentricities or linear trends.
  • $N_{obs} \le 4$: When a unique Keplerian solution is impossible, the algorithm suppresses shape-dependent parameters ($P, e, \omega$) and transitions to linear dynamical relations. It uses Ordinary Least Squares (OLS) regression on the relation $RV_2 = -q RV_1 + \gamma(1+q)$ to constrain the mass ratio ($q$) and systemic velocity ($\gamma$), even with as few as two RV measurements.
• Template Library: The core of the method is a library of 292 distinct Keplerian RV morphologies spanning eccentricities $e \in [0, 0.8]$ and arguments of periastron $\omega \in [0, 360^\circ)$. Templates are normalized to a reference period and velocity amplitude, allowing for analytical scaling to physical values.
• Auxiliary Modules:
  • PSin: For sparse datasets, an auxiliary module performs a time-domain periodogram using an extended trigonometric model (including first-order harmonics) to identify a reliable period range, guiding the subsequent template refinement.
  • Robustness Assessment: For $N_{obs} \ge 7$, stability is tested via a randomized data-thinning (Jackknife-like) procedure. For $N_{obs} < 7$, Monte Carlo perturbation tests with injected Gaussian noise are used to evaluate solution sensitivity.
• Computational Efficiency: The algorithm utilizes Just-In-Time (JIT) compilation and deterministic template matching, avoiding iterative global optimization. This allows it to process datasets in seconds (typically <10s for sparse data), making it suitable as a pre-solver for more intensive pipelines.

Key Contributions and Results
The authors evaluated Syntriod using 10,000 synthetic Keplerian orbits and 12 real spectroscopic binary systems (including HD 160934, Capella A, Kepler 16, and various KIC systems).

1. Period Recovery Accuracy:

   • Well-sampled ($N_{obs} \ge 8$): Syntriod recovers orbital periods with relative accuracies of $\sim 10^{-3}$ and a success rate (within 10% tolerance) exceeding 99%.
   • Theoretical Limit ($N_{obs} = 6$): Syntriod maintains a success rate of $\sim 94\%$, significantly outperforming the Lomb–Scargle method, which drops to $\sim 62\%$ due to aliasing.
   • Underconstrained ($N_{obs} = 5$): Syntriod continues to recover the correct solution in $\sim 83\%$ of cases, whereas LS approaches a random-selection limit ($\sim 46\%$).
   • Degradation Mode: Unlike LS, which exhibits catastrophic failure and bimodal distributions driven by aliasing, Syntriod's performance degradation is characterized by a gradual broadening of uncertainty distributions while maintaining a unimodal, physically consistent solution.

2. Robustness to Orbital Geometry and Sampling:

   • Syntriod maintains high recovery rates across highly eccentric systems ($e > 0.3$) and large observational baselines ($T_{span}/P$), regimes where LS performance deteriorates rapidly.
   • The method successfully reproduces literature solutions for real systems even when datasets are randomly subsampled to sparse regimes ($N_{obs} = 5-6$).

3. Extraction of Physical Parameters in Sparse Regimes:

   • For double-lined systems (SB2) with $N_{obs} \le 4$, Syntriod successfully constrains the mass ratio ($q$) and systemic velocity ($\gamma$) with success rates exceeding 99% (within 10% accuracy for $N_{obs}=3$), even when the orbital period cannot be uniquely determined.

Significance and Claims
The paper claims that Syntriod provides a practical, computationally efficient "pre-solver" for modern orbit-fitting pipelines and large spectroscopic surveys. Its primary significance lies in its ability to:

• Bridge the Sampling Gap: Provide reliable initial estimates in regimes where classical period-search techniques fail due to aliasing and where full Bayesian inference is computationally prohibitive.
• Ensure Physical Consistency: By operating in the orbital-phase domain with a library of physically motivated templates, it avoids the non-physical solutions often generated by purely statistical periodograms in sparse data.
• Enable Downstream Analysis: By constraining the parameter space to physically plausible regions, it significantly reduces the initialization cost and convergence time for subsequent Bayesian inference methods (e.g., MCMC).

The authors explicitly state that Syntriod is not intended to replace full Bayesian inference tools for final uncertainty quantification, as it does not sample the posterior distribution. Instead, it serves as a robust mechanism to generate high-quality initial guesses, thereby enabling more efficient and reliable full orbital solutions in data-limited scenarios. The authors note that a full orbital modeling framework building upon Syntriod is currently under development.