Long-Baseline Radial Velocity Surveys

• Long-baseline radial velocity surveys are high-precision, multi-year observational programs designed to detect Earth-like planets and cold gas giants by monitoring subtle Doppler shifts in stellar spectra.
• They achieve sub-m/s precision using vacuum-enclosed, fiber-fed spectrographs with laser-frequency calibrators, dense cadence strategies, and advanced data reduction pipelines to mitigate stellar and instrumental noise.
• By combining extended time-series data with complementary astrometric measurements, these surveys robustly resolve decade-scale orbits and measure occurrence rates of Solar System analogs.

Long-baseline radial velocity (RV) surveys are high-precision, multi-year observational programs designed to detect and characterize exoplanets—especially true Solar System analogs—by tracking reflex Doppler shifts of nearby stars. These surveys deliver the time coverage and stability required to uncover low-amplitude signals from Earth-mass planets and to resolve decade-scale orbits of cold gas giants, both of which present amplitudes and timescales at the limit of current detection technology. By combining dense, homogeneous RV time series with complementary data (notably astrometry), long-baseline surveys can break key degeneracies, establish robust system architectures, and provide occurrence rates for Solar System-like planetary systems (Yahalomi et al., 2023).

1. Survey Design Principles and Observing Strategies

Long-baseline RV surveys are explicitly engineered to build extended, high-cadence time series for well-chosen target samples. The Terra Hunting Experiment (THE) exemplifies this approach: it implements nightly, queue-scheduled RV monitoring of at least 40 quiet, slowly rotating G/K dwarfs within 20 pc, using HARPS3 over a 10-year baseline (Yahalomi et al., 2023, Hall et al., 2018). The rationale is to optimize detection for both:

• Earth-analog planets: Orbits near 1 yr, with RV semi-amplitudes $K \approx 0.09$ m/s for 1 $M_\oplus$ at 1 AU, requiring a $\sim$0.3 m/s per-epoch precision and $N \sim 3000$ cadence to average down stellar granulation and resolve window aliases.
• Cold gas giants (CGGs): Saturn or Jupiter analogs at periods 10–12 yr ($K \sim 10$–20 m/s), necessitating $\gtrsim$10 yr continuous monitoring to constrain period, eccentricity, and phase.

Other legacy programs (California Legacy Survey, Keck/HIRES-APF-Lick, LCES/Keck, HET/HRS, GOT 'EM) span baselines of 9–33 yr and typically involve $N > 50$–$>300$ epochs per star. Cadence is tuned to target not only periodic sampling but also to fill gaps and mitigate annual and activity-induced aliases (Rosenthal et al., 2021, Butler et al., 2017, Niedzielski et al., 15 Oct 2025, Dalba et al., 2024).

Program	Baseline (yr)	Cadence	Median $\sigma_{RV}$ (m/s)	N (epochs per star)
THE (HARPS3)	10	Nightly	0.3	$\sim$3000
CLS (Keck/HIRES)	21 (3–33)	$\sim$Monthly	1.0 (post-2004)	41 (median)
HET/HRS	9	Sparse/seasonal	5–7	10–25

2. Precision Requirements and Error Budget

Detecting Earth-twins and CGGs hinges on sub-m/s instrumental stability and photon-limited precision:

• Photon-limited single-epoch precision:
  • THE: $\sigma_{RV} \approx 0.3$ m/s (S/N $\sim$200, V=7–9) (Yahalomi et al., 2023).
  • HARPS/ESPRESSO/EXPRES/NEID/EXPRES: Routinely $<1$ m/s, some nights achieving 10–30 cm/s (Burt et al., 3 Nov 2025).
• Systematic error control:
  • Modern surveys employ vacuum-enclosed, fiber-fed spectrographs with laser-frequency comb or Fabry–Pérot calibrators; daily and hourly calibration frames (LFC/FP); multi-layer temperature control ($\Delta T\lesssim1$ mK); and regular RV zero-point monitoring (e.g., using APOGEE DR17 standard stars) ensure global consistency below tens of cm/s over decadal timescales (Burt et al., 3 Nov 2025, Li et al., 2023).
• Astrophysical noise:
  • Granulation, oscillations ($\sim$0.5 m/s), and activity cycles (0.1–10 m/s) are mitigated by dense cadence, multi-year span, activity monitoring (S-index, BIS, FWHM, $<$B$_\ell$>$$), and advanced time-series modeling (Gaussian Processes, multi-output GP, PCA, deep learning) (<a href="/papers/2511.01954" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Burt et al., 3 Nov 2025</a>, <a href="/papers/2406.20023" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Rescigno et al., 2024</a>).</li> </ul></li> </ul> <h2 class='paper-heading' id='data-reduction-stellar-activity-mitigation-and-analysis-pipelines'>3. Data Reduction, Stellar Activity Mitigation, and Analysis Pipelines</h2> <p>Optimal extraction pipelines combine classical Doppler modeling (iodine cell forward modeling, cross-correlation with binary masks, or template-matching) with modern statistical and systematics-removal frameworks:</p> <ul> <li><strong>Pipeline elements:</strong> <ul> <li>Flat-relative optimal extraction; 2D polynomial wavelength solutions; fiber-scrambling corrections; and multi-epoch calibration tracking (<a href="/papers/2511.01954" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Burt et al., 3 Nov 2025</a>).</li> <li>Post-processing with tools such as YARARA (PCA at spectrum level) or GP modeling for activity decoupling, leveraging multi-index time series (S-index, BIS, FWHM,$$<$B$_\ell$>$).
  • For long-period signals, robust model selection via Bayesian evidence (Bayes factor, BIC) or periodograms with empirically calibrated false-alarm-probability thresholds ($<$0.1% typical) (Hall et al., 2018).
• Stellar activity discrimination:
  • Explicit correlation checks between candidate Keplerian signals and activity indicators; signals coincident in period/phase with activity proxies are rejected or explicitly modeled jointly (Butler et al., 2017, Yahalomi et al., 2023, Burt et al., 3 Nov 2025).
  • The mean longitudinal magnetic field $<$B$_\ell$>$$, measurable via polarimetry or disk-resolved magnetograms, provides a uniquely effective, planet-insensitive tracer of rotational period and cycle phase, enabling secure modeling/removal of both periodic and long-cycle RV noise via physically-informed GP priors (<a href="/papers/2406.20023" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Rescigno et al., 2024</a>).</li> </ul></li> </ul> <h2 class='paper-heading' id='detection-sensitivity-completeness-and-monte-carlo-assessment'>4. Detection Sensitivity, Completeness, and Monte Carlo Assessment</h2> <p>Long-baseline RV programs quantify detection sensitivity by injecting synthetic signals and measuring completeness as a function of mass ($$m_p$), period ($P$), and eccentricity ($e$$), under realistic survey cadence and noise:</p> <ul> <li><strong>Single-planet sensitivity:</strong> <ul> <li>For RV precision$$\sigma_{RV}=0.3$m/s,$N=3000$, THE achieves$\sigma (K)$well below$0.01$m/s—sufficient to detect 1$M_\oplus$at 1 AU with$S/N>5$$(<a href="/papers/2302.05064" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Yahalomi et al., 2023</a>, <a href="/papers/1806.00518" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Hall et al., 2018</a>).</li> <li>Multi-decade surveys (Keck/CLS) push Jupiter ($$1\,M_J$)-analog detectability to$P\approx$10–15 yr, $a\approx$5–7 AU, $K\approx$12 m/s, and are complete to $M_p \sin i \gtrsim 0.3 M_J$for$P\lesssim$$baseline (<a href="/papers/1603.08384" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Wright, 2016</a>, <a href="/papers/2105.11583" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Rosenthal et al., 2021</a>).</li> </ul></li> <li><strong>Joint RV-astrometry programs:</strong> <ul> <li>Astrometry with Gaia (10 yr,$$\sim$34 $\mu$as precision) and Roman (25 yr, 5–20$\mu$$as) combined with long-cadence <a href="https://www.emergentmind.com/topics/rotational-optimizer-variants-rvs" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">RVs</a> provides an order-of-magnitude improvement in mass and period precision for CGGs, and directly recovers orbital inclination, breaking the$$m_p \sin i$$degeneracy (<a href="/papers/2302.05064" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Yahalomi et al., 2023</a>).</li> </ul></li> <li><strong>Monte Carlo frameworks:</strong> <ul> <li>Complete end-to-end simulation, from synthetic system injection to joint <a href="https://www.emergentmind.com/topics/adaptive-markov-chain-monte-carlo-mcmc-algorithm" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">MCMC</a> retrieval, demonstrates that the inclusion of Roman astrometry raises survey completeness for Saturn analogs (at 10 pc) from$$\sim$20% to $\sim$90% (Yahalomi et al., 2023).

5. Empirical Results and Benchmark Surveys

Major long-baseline RV surveys have produced foundational astrophysical discoveries and statistical measurements:

• California Legacy Survey:
  • 719 FGKM stars, 178 planets (14 new/revised); baselines up to 33 yr; K and $P$constraints for$P$up to$>30$$yr (<a href="/papers/2105.11583" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Rosenthal et al., 2021</a>).</li> </ul></li> <li><strong>LCES HIRES/Keck:</strong> <ul> <li>20-yr survey, 357 significant periodicities (225 published planets), detection floor$$K\sim$2 m/s (post-2004) (Butler et al., 2017).
• HET/HRS (Niedzielski et al.):
  • 9-yr campaign, $\sigma_{RV}=5$–7 m/s for giants, detection of a 10.6$M_J$(P=5.17 yr,$e=0.59$) and a 0.55$M_J$(P=123 d,$e=0.73$) companion, illustrating necessity of baseline$>$ period and high-cadence periastron coverage for eccentric orbits (<a href="/papers/2510.13728" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Niedzielski et al., 15 Oct 2025</a>).</li> </ul></li> <li><strong>GOT &#39;EM Survey:</strong> <ul> <li>Long-term Doppler spectroscopy for 11 systems with long-period ($>$100 d) transiting giants, enabling dynamical confirmation, false-positive identification, and heavy-element abundance estimation, but still limited for faint hosts and lower-mass planets due to precision/cadence tradeoffs (Dalba et al., 2024).

6. Scientific Impact and Implications for Solar System Analogs

Long-baseline RV surveys are essential for empirical assessment of Solar System-like architectures:

• Solar System analog definition:
  • Host both an Earth-mass planet at 0.8–1.2 AU and a CGG (0.3–1 $M_J$) at 4–6 AU, both with$e<0.3$$(<a href="/papers/2302.05064" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Yahalomi et al., 2023</a>).</li> </ul></li> <li><strong>Expected discoveries:</strong> <ul> <li>At$$d \leq 10$$pc, a joint THE + Gaia + Roman campaign is expected to recover 3–10 such analogs out of 50 targets if the true occurrence rate is$$f_{SS} \sim$10–20% (Yahalomi et al., 2023).
• Architecture and occurrence rates:
  • Decades-long time series and completeness maps from surveys such as Keck+Lick underpin robust measurement of cold-Jupiter occurrence (empirically $f_{\rm Jup}\approx 4\pm1$$\%), test models of system hierarchy, and identify the best targets for imaging and atmospheric follow-up (<a href="/papers/1603.08384" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Wright, 2016</a>, <a href="/papers/2105.11583" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Rosenthal et al., 2021</a>).</li> </ul></li> </ul> <h2 class='paper-heading' id='best-practices-and-future-recommendations'>7. Best Practices and Future Recommendations</h2> <p>For future long-baseline RV campaigns targeting Earth-mass and Solar System analogs:</p> <ul> <li>Adopt nightly or near-nightly cadence with sustained (ideally$$\geq$10 yr) baseline (Yahalomi et al., 2023, Hall et al., 2018).
  • Implement queue-scheduled, multi-year observing for dynamic baseline fill and periastron coverage.
  • Use vacuum-enclosed, fiber-fed, LFC-calibrated spectrographs with routine zero-point monitoring through APOGEE-standard network (Burt et al., 3 Nov 2025, Li et al., 2023).
  • Regularly acquire activity indices (S-index, BIS, FWHM, $<$B$_\ell$>$) and apply joint multi-output GP activity modeling (Rescigno et al., 2024).
  • Publicly release all RV, activity, and calibration data for independent validation and occurrence-rate synthesis (Rosenthal et al., 2021).
  • Where possible, coordinate or cross-calibrate with astrometric campaigns (Gaia, Roman) for full 3D orbit reconstruction and degeneracy-breaking (Yahalomi et al., 2023).

  Long-baseline RV surveys, when fully exploiting cadence, error control, and joint astrometric modeling, define the empirical foundation for detecting and characterizing planetary system architectures that approach the complexity and diversity of the Solar System.