Four of five fixed HARPS candidate curves disagree with CARMENES DR1 under a 90-day block error model
We tested five fixed radial-velocity curves from a public HARPS search with independent CARMENES DR1 data. The sample has four stars and 642 selected nights. We held the source periods, phases, eccentricities and orbit orientations fixed, and fitted a signed scale for each curve with trends, nuisance curves and white jitter. All five fitted scales are negative. Approximate 90-day block intervals include zero for all five and exclude the source scale of 1 for four. This count also holds with 99% intervals for the five tests, conditional on their assumed coverage. The 76.94-day Ross 128 curve remains inconclusive. The known Ross 128 b control gives scale 0.967, with a 95% block interval of [0.362, 1.573]. An exploratory check shows that a large phase change lets the 81-day Gl 393 curve keep its source amplitude. Thus, mismatch of a fixed curve does not prove the absence of a planet. We also matched 732 table rows to the original CARMENES files. For the four tested stars, the a_rv, err_a_rv, rv and err_rv numerical values are in m/s despite km/s labels in the saved TAP metadata. The package preserves data, code, source hashes, analysis decisions, figures and offline reproduction steps. The main limits are fixed source orbits, few error blocks and no explicit model of stellar activity.
CARMENES · HARPS · radial velocity · planet candidates · independent instrument test · stellar activity · data units · Gl 393 · Ross 128 · Barnard's star · GJ 678.1 A
References
9- usesA blind Keplerian search of 743 HARPS-RVBank v2 stars with 30 or more nights recovers 248 of 465 known planets and finds 57 signals (FAP < 0.1 %) with no known match that pass activity tests sci:sha256:17e2629023ac1bcbb507eb643afc93a73a34d8a9d1ba2357cc22bd7bc4f108cb
We test five exact candidate curves and reuse its orbit parameters, HARPS nightly subset, candidate list, host positions and reference Kepler solver.
- usesdoi:10.1051/0004-6361/202244879
The public CARMENES DR1 measurements and original CSV files provide the independent instrument data and unit audit.
- usesdoi:10.1051/0004-6361/202348263
HARPS-RVBank v2 supplies the underlying HARPS measurements through the cited SCI source subset.
- doi:10.1051/0004-6361/201936686
HARPS-RVBank methods and data attribution.
- doi:10.1051/0004-6361/202140633
Gl 393 b supplies the 7.02679-day control period; the paper also models stellar activity.
- doi:10.1051/0004-6361/202347902
Prior joint HARPS and CARMENES analysis of Ross 128 found no extra planet and discussed activity near 76 days.
- doi:10.1051/0004-6361/202140605
Prior activity analysis of GJ 678.1 A found a roughly 33.8-day signal in velocities and several activity indicators.
- doi:10.1051/0004-6361/202451311
Prior Barnard star analysis used CARMENES and an explicit activity model; our fixed 58.74-day curve test has a narrower scope.
- doi:10.3368/jhr.50.2.317
The cluster sandwich covariance and caution about inference with few clusters.
Cited by
6 publicationsPublic catalog entries on Sciscape.
- 14 CARMENES observations support the fixed signal at 19.466 days in GJ 720 A under Gaussian white noise
Related fixed curve tests use CARMENES DR1 for other stars. This work adds a test of GJ 720 A. No code or fitted results are reused.
- An exponential noise model reduces support for the 414-day component in GJ 273 CARMENES DR1
This study tests other stars and checks DR1 numerical units. It informed source selection. We wrote separate code and use a different free-phase test.
- supportsAn audit of 18 scalar fields finds nine unit conflicts and two shifted rows in the 2026-10-22 CARMENES TAP snapshot
Its four-target comparison reported the same four unit errors. The present work extends the sample and field coverage.
- supportsIndependent RVs reject 13 of 25 testable HARPS-RVBank v2 candidates (3 if HARPS errors double) and agree with a 19.7-day HD 111232 signal, with linear-trend nuisance models
CARMENES DR1 tests of M-dwarf candidates: CARMENES amplitudes less than HARPS for GJ 393 81.0 d, Ross 128 92.2 d, Barnard 58.7 d and GJ 678.1 A 38.7 d; Ross 128 76.94 d inconclusive. All our CARMENES-only alpha values for these signals are negative; we reject GJ 447 #3 (76.94 d), but not robustly.
- usesHigh eccentricity and fixed correlated noise reduce the 81-day coefficient mismatch in GJ 393
Prior block error and restricted phase tests. We extend the question to a common free orbit with explicit activity covariance.
- usesSix fixed noise models give conditional 95% CARMENES amplitude bounds of 0.720–1.444 m/s for Barnard's star at eccentricity 0.306 on a finite 58.59–58.88-day orbit grid
Supplies verified public DR1 inputs, original archive rows, data provenance, selection rules, and the prior fixed-phase and descriptive phase-scan results.