Disentangling Stellar Activity Noise for Exoplanet Detection

26 concepts15 papers1 proposal

The Stellar Noise Floor

When searching for an "Earth-twin"—a rocky planet orbiting within the habitable zone of a Sun-like star—astronomers look for an incredibly subtle gravitational tug. This tug induces a radial velocity (RV) shift in the host star of merely $\sim 10\text{ cm/s}$, a threshold highlighted by Richard Hall et al. in On the Feasibility of Intense Radial Velocity Surveys for Earth-Twin Discoveries.

However, stars are not static, uniformly bright spheres. Their surfaces are boiling cauldrons of plasma punctuated by dark starspots, bright faculae, granulation cells, and global acoustic oscillations. As A.-L. Maire et al. emphasize in Workshop Summary: Exoplanet Orbits and Dynamics, this intrinsic stellar jitter creates a noise floor that frequently exceeds the signals of low-mass planets by orders of magnitude.

Why does this surface activity mimic the Doppler shift of a planet? Consider a star rotating on its axis. One hemisphere moves toward us (blueshifted), while the other moves away (redshifted). In a perfectly quiet star, these contributions balance out, producing symmetric spectral lines. But if a dark starspot appears on the blueshifted hemisphere, it blocks some of the blueshifted light. The overall integrated spectrum of the star now appears artificially redshifted. As the spot rotates across the stellar disk, this apparent velocity shifts back and forth, creating a periodic or quasi-periodic signal that can easily be mistaken for the gravitational pull of an orbiting planet, as discussed by Umberto Simola et al. in Accounting for stellar activity signals in radial-velocity data by using change point detection techniques.

This confusion is magnified in young stars, where high magnetic activity, starspots, and flares introduce severe RV contamination. As shown by Roberto Tejada Arevalo et al. in Stability Constrained Characterization of the 23 Myr Old V1298 Tau System: Do Young Planets Form in Mean Motion Resonance Chains?, this noise makes it difficult to determine accurate planetary masses and eccentricities without invoking external dynamical stability constraints. Even worse, when the timescales of stellar rotation or spot lifetimes coincide with the observational sampling schedule, they can generate spurious periodicities. A famous example of this was analyzed by Vinesh Rajpaul et al. in Ghost in the time series: no planet for Alpha Cen B, where the authors demonstrated that a previously claimed Earth-mass planet around Alpha Centauri B was actually an artifact born from the interplay between stellar activity and data sampling.

Moving Beyond Scalar RVs: The Geometry of Line Profiles

To break the degeneracy between a true planetary wobble and a rotating starspot, we must look at how the spectral lines change. A planet shifts the entire spectral line back and forth without changing its shape. A starspot, by contrast, distorts the line profile itself, introducing asymmetries.

Traditionally, astronomers measured these asymmetries using scalar indicators like the Bisector Span or the Full Width at Half Maximum of the Cross-Correlation Function (CCF). However, compressing a complex 2D line profile into a few scalar metrics discards valuable information.

To fully leverage this geometric insight, Haochuan Yu et al. developed a framework in A Gaussian process model for stellar activity in 2D line profile time-series to model stellar activity directly within the 2D line profile time-series. Because a planet shifts the line uniformly across all velocity channels while a spot distorts specific parts of the profile depending on its position on the stellar disk, modeling the full 2D profile allows algorithms to cleanly separate the two effects, even when the planet's orbital period matches the star's rotation period.

Joint Modeling via Gaussian Processes

How do we mathematically capture the stochastic, evolving nature of stellar activity? Simple deterministic models fail because starspots emerge, evolve, and decay unpredictably.

The breakthrough solution is Gaussian Process (GP) regression. A GP treats the stellar activity as a latent, correlated random variable governed by a kernel function (often a quasi-periodic kernel that captures both the stellar rotation period and the lifetime of active regions).

The core intuition behind modern GP frameworks, pioneered by Vinesh Rajpaul et al. in A Gaussian process framework for modelling stellar activity signals in radial velocity data, is joint modeling. Instead of fitting the RV data alone, the GP simultaneously models the RVs alongside ancillary activity indicators (such as chromospheric emission or line-shape metrics). Because the underlying physical cause—the magnetic activity—drives both the RV variations and the activity indicators, they share the same latent GP framework.

Mathematically, if $V(t)$ is the observed radial velocity and $I(t)$ is an activity indicator, they can be expressed as linear combinations of a shared latent GP, $G(t)$, and its time derivative, $\dot{G}(t)$:

$$V(t) = A \cdot G(t) + B \cdot \dot{G}(t) + V_{\text{planet}}(t)$$ $$I(t) = C \cdot G(t)$$

The derivative term $\dot{G}(t)$ captures the fact that as a spot crosses the stellar meridian, its effect on the line asymmetry peaks or vanishes at different times than its effect on the total flux. By training the GP on multiple indicators simultaneously, the model learns the exact fingerprint of the stellar activity, leaving the purely Keplerian signal ($V_{\text{planet}}$) to be recovered with high fidelity. As shown by Eva-Maria Ahrer et al. in The HARPS search for southern extra-solar planets – XLV. Two Neptune mass planets orbiting HD 13808: a study of stellar activity modelling’s impact on planet detection, failing to use these sophisticated joint models can easily lead to false-positive planet discoveries or severely distorted orbital parameters.

Temporal Segmentation and High-Cadence Sampling

While GPs are powerful, their performance can degrade if a star's activity regime shifts abruptly. To address this, Umberto Simola et al. introduced Change Point Detection (CPD) techniques in Accounting for stellar activity signals in radial-velocity data by using change point detection techniques. CPD recognizes that the relationship between surface activity and apparent velocity is piecewise stationary; as old spot groups fade and new ones emerge at different latitudes, the mapping changes. Segmenting the time series into distinct, stationary epochs allows for more precise local corrections.

Complementing these algorithmic fixes is the observational strategy of high-cadence monitoring. As argued by Richard Hall et al. in On the Feasibility of Intense Radial Velocity Surveys for Earth-Twin Discoveries, observing a target nightly or multiple times per night allows astronomers to average out high-frequency stellar noise components like granulation and p-mode oscillations, which operate on timescales of minutes to hours. This high-cadence sampling provides the dense data streams required for GPs and CPD algorithms to map out the slower, rotation-driven stellar signals.

Contamination in Photometry and Spectroscopy

Stellar activity is not unique to radial velocity surveys; it equally plagues photometric transit searches and transmission spectroscopy. When a planet transits an active star, it may cross directly over a dark spot, causing a temporary bump in the transit light curve, or it may transit an unspotted region while unocculted spots elsewhere on the star artificially deepen the transit, an effect noted by J. Southworth et al. in High-precision photometry by telescope defocussing - II. The transiting planetary system WASP-4. In ultra-cool dwarfs, frequent high-energy flares and high variability further complicate the identification of true Earth-sized transiting worlds, as detailed by Florian Lienhard et al. in Global analysis of the TRAPPIST Ultra-Cool Dwarf Transit Survey.

The problem becomes acute in transmission spectroscopy, where astronomers look at starlight filtering through a planet's atmosphere to detect chemical signatures. As Olivia Lim et al. observed in Atmospheric Reconnaissance of TRAPPIST-1 b with JWST/NIRISS: Evidence for Strong Stellar Contamination in the Transmission Spectra, features like starspots and faculae imprint their own spectral signatures onto the data. Because these active regions have different temperatures than the rest of the stellar surface, their water vapor or metal oxide lines can perfectly mimic or completely mask the atmospheric signals of the exoplanet. This epoch-dependent contamination requires the same level of rigorous, joint stellar-surface modeling that has revolutionized radial velocity surveys.

Key Foundational Work

9+ frequencies for V534 Tauri, aδScuti variable in the Pleiades (2004) — contributed 1 concept elements to this direction.
High-precision photometry by telescope defocussing - II. The transiting planetary system WASP-4 (2009) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_stellar_activity_noise_40b7b54c,
  title = {stellar activity noise},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/stellar-activity-noise/rdr_40b7b54c420141b4a0a6a53c950ead57},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_stellar_activity_noise_40b7b54c_v1,
  title = {stellar activity noise},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/stellar-activity-noise/rdr_40b7b54c420141b4a0a6a53c950ead57/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake