Disentangling Multi-epoch Single-lined Binary Star Spectra to Reveal Interstellar Medium Features via MCMC

Abstract

Diffuse Interstellar Bands (DIBs) are interstellar absorption features with unidentified carriers in the optical and near-infrared spectra. When observed in cool star spectra, they are heavily contaminated by stellar lines. Traditional methods rely on template matching to remove the stellar component, but their accuracy is fundamentally limited by template quality. We present a template-independent algorithm that separates stellar and interstellar features in binary star spectra by exploiting orbital motion: stellar lines shift with changing radial velocity (RV), whereas DIBs remain stationary in the rest frame. The MCMC-RVSD algorithm, implemented as an RV-based Bayesian Markov Chain Monte Carlo framework, fits both components simultaneously. We validated our approach by applying it to mock spectra simulating LAMOST medium-resolution observations. The method precisely recovers the parameters of DIB $λ$6614. It performs reliably for spectra with $R=3000$-$7500$. We applied the method to 434 real binary sources from the LAMOST medium-resolution survey and introduced the MCMC autocorrelation time ($τ_{\mathrm{DIB}}$) to assess the fit quality. Our analysis shows that an RV span $>20$ km s$^{-1}$ is the primary driver of successful line separation. Denser orbital sampling provides further improvement, and the algorithm remains robust against typical observational noise at SNR $>50$ and RV errors $<5$ km s$^{-1}$. Under ideal conditions, our approach yields significantly smaller parameter uncertainties than traditional template matching. Additionally, the MCMC-RVSD algorithm can also handle multi-stellar line contamination, non-Gaussian DIBs and other DIBs by adjusting model and prior settings. This algorithm provides a reliable, general-purpose tool for precise measurement of interstellar features in time-domain spectroscopic surveys.