GJ 887: M Dwarf Planetary System

• GJ 887 is a nearby M1 dwarf known for its compact multi-planet system, identified through high-precision radial velocity measurements.
• The system features four confirmed planets, including an Earth-mass inner world and a super-Earth in the habitable zone, plus a candidate signal at 2.22 days.
• Advanced signal extraction and Gaussian process modeling enabled separation of stable planetary signals from stellar activity, ensuring robust parameter estimation.

GJ 877, as it appears in the abstract of the RedDots study "RedDots: Multiplanet system around M dwarf GJ 887 in the solar neighborhood," is a typographical error for GJ 887. All analysis in that study concerns GJ 887 = Gl 887, also catalogued as HD 217987 and Lacaille 9352, a bright, nearby M1V dwarf at 3.2877 pc that hosts a compact system of low-mass, nontransiting planets detected by radial velocities. The reanalysis reported a preferred four-planet architecture consisting of the previously known planets b and c, the newly confirmed Earth-mass planet e, and the newly confirmed super-Earth d in the habitable zone, together with an additional 2.21661 d signal that remains a candidate rather than a confirmed planet (Hartogh et al., 9 Feb 2026).

1. Designation and object identity

The object studied is the nearby M dwarf commonly known as Gliese 887. Throughout the body of the RedDots paper the star is designated GJ 887, with SIMBAD aliases HD 217987 and Lacaille 9352. The occurrence of “GJ 877” in the abstract is explicitly identified as a typographical error; the analysis refers throughout to GJ 887 = Gl 887, described there as the well-known bright, nearby M1 dwarf at 3.29 pc.

This distinction is important because the planetary architecture, stellar properties, and observational inferences attributed to “GJ 877” in that abstract belong to GJ 887. A plausible implication is that any use of “GJ 877” in connection with the reported 9 d, 21 d, 50 d, and 4.4 d signals should be interpreted as shorthand for the GJ 887 system rather than as a separate stellar designation.

2. Stellar parameters and magnetic activity

GJ 887 is reported as an M1V star with fundamental parameters drawn from Mann et al. (2015) and Gaia. The adopted stellar parameters are: $T_{\rm eff} = 3688 \pm 86~\mathrm{K}$, $M_* = 0.495 \pm 0.049~M_\odot$, $R_* = 0.468 \pm 0.022~R_\odot$, $L_* = 0.0368 \pm 0.0040~L_\odot$, $[\mathrm{Fe/H}] = -0.06 \pm 0.08$, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$, $\log g = 4.156^{+0.040}_{-0.038}$, age $\approx 2.9^{+8.0}_{-2.2}$ Gyr, distance $d = 3.2877$ pc with parallax $304.135 \pm 0.020$ mas, and apparent brightnesses $M_* = 0.495 \pm 0.049~M_\odot$0 and TESS magnitude $M_* = 0.495 \pm 0.049~M_\odot$1.

The star is described as relatively quiet compared to typical early-M dwarfs, with low spot coverage, low photometric variability, and low Ca II H&K and H$M_* = 0.495 \pm 0.049~M_\odot$2 indices in earlier epochs. In the RedDots analysis, the stellar rotation period is measured from Gaussian-process modeling of the RV time series as $M_* = 0.495 \pm 0.049~M_\odot$3, with consistent values reported as 38.6–38.7 $M_* = 0.495 \pm 0.049~M_\odot$4 0.5–0.6 d across models (Hartogh et al., 9 Feb 2026).

That rotation period is independently corroborated by activity signals near 39 d in ASAS photometry and in several spectroscopic tracers: HARPS FWHM, BIS, differential line width (dLW), H$M_* = 0.495 \pm 0.049~M_\odot$5, and Na D1/D2. A harmonic near 19 d is also present, as expected. The study further reports an “active phase” in 2019, during which the 39 d periodicity appeared coherently in RV, Na D1, and dLW. This behavior supports the interpretation that the 38–39 d signal is rotationally modulated stellar activity rather than a planetary orbit.

3. Observational basis

The RedDots collaboration extended the earlier data set with 101 new HARPS and 12 new ESPRESSO radial velocities, obtained with a cadence intended to test the origin of the putative $M_* = 0.495 \pm 0.049~M_\odot$6 signal. The full HARPS material comprises 850 spectra total; after nightly binning, 277 RVs remained. These were segmented into three instrumental states: HARPSpre with 88 RVs and median internal uncertainty 0.44 m s$M_* = 0.495 \pm 0.049~M_\odot$7, HARPSpost with 164 RVs and median internal uncertainty 0.87 m s$M_* = 0.495 \pm 0.049~M_\odot$8, and HARPSpw with 25 RVs and median internal uncertainty 0.61 m s$M_* = 0.495 \pm 0.049~M_\odot$9. ESPRESSO contributed 19 spectra total and 12 nightly binned RVs spanning three years, with median internal uncertainty 0.17 m s$R_* = 0.468 \pm 0.022~R_\odot$0.

The analysis fit relative offsets and independent jitter parameters for each HARPS segment and for ESPRESSO. Additional public RVs from Keck HIRES, UCLES, and PFS were not used because their median uncertainties of 1–2 m s$R_* = 0.468 \pm 0.022~R_\odot$1 were lower precision than the HARPS and ESPRESSO data. Radial velocities were derived by template matching using serval, specifically because its order-by-order fitting performs better than cross-correlation for M dwarfs.

Activity diagnostics combined ESO-DRS CCF indicators—BIS and FWHM—with serval indices: H$R_* = 0.468 \pm 0.022~R_\odot$2, Na D1/D2, dLW, and the chromatic index (CRX). These observables were used to confirm $R_* = 0.468 \pm 0.022~R_\odot$3 and to distinguish activity-driven variability from Keplerian signals.

Photometric context came from both ground and space observations. ASAS provided 523 points from 2001–2009 with median photometric precision $R_* = 0.468 \pm 0.022~R_\odot$4. TESS observed the star in Sectors 2, 28, and 69, each of approximately 27 d, with median PDCSAP flux uncertainties of $R_* = 0.468 \pm 0.022~R_\odot$5. A Transit Least Squares (TLS) search found no transit detections at the RV periods, consistent with a nontransiting planetary architecture. The TESS sector window functions are explicitly described as suboptimal for detecting $R_* = 0.468 \pm 0.022~R_\odot$6 d stellar rotation.

4. Signal extraction and activity mitigation

The radial-velocity inference pipeline combined multiple complementary methods. Initial signal searches used generalized Lomb–Scargle (GLS), stacked Bayesian GLS (sBGLS), and $R_* = 0.468 \pm 0.022~R_\odot$7-periodograms based on compressed sensing. Keplerian modeling was performed with juliet, using RadVel for orbital fits and dynesty for dynamic nested sampling, yielding posterior distributions and Bayesian evidences $R_* = 0.468 \pm 0.022~R_\odot$8 for model comparison (Hartogh et al., 9 Feb 2026).

Stellar activity was modeled with Gaussian processes. The preferred covariance model was the quasi-periodic (QP) kernel implemented in george:

$R_* = 0.468 \pm 0.022~R_\odot$9

where $L_* = 0.0368 \pm 0.0040~L_\odot$0, $L_* = 0.0368 \pm 0.0040~L_\odot$1 is the GP amplitude for each instrument interval, $L_* = 0.0368 \pm 0.0040~L_\odot$2 sets the decay time of correlations with spot lifetime $L_* = 0.0368 \pm 0.0040~L_\odot$3, $L_* = 0.0368 \pm 0.0040~L_\odot$4 controls the strength of the periodic modulation, and $L_* = 0.0368 \pm 0.0040~L_\odot$5 is the activity periodic timescale. The QP kernel was preferred because it provided a better balance of parsimony and fidelity than SHO and double-SHO kernels in celerite. The dSHO model, in particular, is reported to have a tendency to overfit short-timescale variability and to partially absorb the $L_* = 0.0368 \pm 0.0040~L_\odot$6 d planetary signal.

Coherence tests were central to separating planets from activity. sBGLS tracked the temporal growth of signal significance, while apodized Gaussian-envelope fits probed the stability of the semi-amplitude over the observing baseline. Under this analysis, the signals at approximately 4.4249 d, 9.2619 d, 21.784 d, and 50.77 d behaved as coherent, stable planetary signals, whereas the 38–39 d rotation signature was incoherent across the full baseline and strongest during the active phase. This methodological distinction underpins the classification of the four principal periodicities as planetary and the 39 d signal as stellar activity.

5. Planetary system architecture

Bayesian model comparison identifies a preferred four-planet model, denoted D in the study, together with one additional short-period candidate, f, that does not yet satisfy the adopted threshold for strong evidence. The confirmed components are all nontransiting and are characterized through minimum masses $L_* = 0.0368 \pm 0.0040~L_\odot$7.

Object	Period	Minimum mass and status
GJ 887 e	$L_* = 0.0368 \pm 0.0040~L_\odot$8	$L_* = 0.0368 \pm 0.0040~L_\odot$9, confirmed
GJ 887 b	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$0	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$1, confirmed
GJ 887 c	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$2	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$3, confirmed
GJ 887 d	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$4	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$5, confirmed
GJ 887 f	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$6	$[\mathrm{Fe/H}] = -0.06 \pm 0.08$7, candidate

GJ 887 e is the newly confirmed Earth-mass planet. Its reported parameters are $[\mathrm{Fe/H}] = -0.06 \pm 0.08$8, $[\mathrm{Fe/H}] = -0.06 \pm 0.08$9 fixed to 0, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$0 BJD, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$1 AU, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$2, and $v \sin i < 2.5~\mathrm{km\,s^{-1}}$3 K for $v \sin i < 2.5~\mathrm{km\,s^{-1}}$4 and no redistribution. Its detection metrics are unusually strong for a sub-meter-per-second signal: GLS residual peak FAP $v \sin i < 2.5~\mathrm{km\,s^{-1}}$5, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$6 FAP $v \sin i < 2.5~\mathrm{km\,s^{-1}}$7, $v \sin i < 2.5~\mathrm{km\,s^{-1}}$8 relative to the two-planet model, and an amplitude detection of approximately $v \sin i < 2.5~\mathrm{km\,s^{-1}}$9.

GJ 887 b, previously known, is reported with $\log g = 4.156^{+0.040}_{-0.038}$0, $\log g = 4.156^{+0.040}_{-0.038}$1, $\log g = 4.156^{+0.040}_{-0.038}$2 rad, $\log g = 4.156^{+0.040}_{-0.038}$3 BJD, $\log g = 4.156^{+0.040}_{-0.038}$4 AU, $\log g = 4.156^{+0.040}_{-0.038}$5, and $\log g = 4.156^{+0.040}_{-0.038}$6 K. Its signal is supported by GLS FAP $\log g = 4.156^{+0.040}_{-0.038}$7, $\log g = 4.156^{+0.040}_{-0.038}$8 FAP $\log g = 4.156^{+0.040}_{-0.038}$9, and strong Bayesian evidence.

GJ 887 c, also previously known, has $\approx 2.9^{+8.0}_{-2.2}$0, $\approx 2.9^{+8.0}_{-2.2}$1, $\approx 2.9^{+8.0}_{-2.2}$2 rad, $\approx 2.9^{+8.0}_{-2.2}$3 BJD, $\approx 2.9^{+8.0}_{-2.2}$4 AU, $\approx 2.9^{+8.0}_{-2.2}$5, and $\approx 2.9^{+8.0}_{-2.2}$6–$\approx 2.9^{+8.0}_{-2.2}$7 K. The study notes that its coherence dips during the star’s active phase but recovers afterward, while apodized fits show a stable amplitude overall.

GJ 887 d is the newly confirmed super-Earth in the habitable zone. The favored model adopts a circular orbit with $\approx 2.9^{+8.0}_{-2.2}$8, $\approx 2.9^{+8.0}_{-2.2}$9 BJD, $d = 3.2877$0 AU, $d = 3.2877$1, and $d = 3.2877$2 K. The residual-periodogram and evidence-based metrics are more marginal than for the shorter-period planets but remain supportive: GLS residual peak FAP $d = 3.2877$3, $d = 3.2877$4 FAP $d = 3.2877$5, an amplitude detection of approximately $d = 3.2877$6, and $d = 3.2877$7 relative to the three-planet model before conservative prior correction. The paper explicitly notes that the dSHO GP can absorb part of the long-period power, yet the four-planet model remains preferred across kernels (Hartogh et al., 9 Feb 2026).

The additional signal, GJ 887 f, has $d = 3.2877$8, $d = 3.2877$9, $304.135 \pm 0.020$0 fixed to 0, $304.135 \pm 0.020$1–$304.135 \pm 0.020$2 BJD, $304.135 \pm 0.020$3 AU, $304.135 \pm 0.020$4, and $304.135 \pm 0.020$5 K. Its amplitude is detected at approximately $304.135 \pm 0.020$6, and the residuals of the four-planet GP model give GLS FAP $304.135 \pm 0.020$7 with $304.135 \pm 0.020$8 relative to the four-planet model. Because this lies below the adopted $304.135 \pm 0.020$9 threshold for strong evidence, and because prior-width correction makes the improvement negative, it is classified as promising but not yet confirmed.

6. Dynamical structure and habitability context

Long-term dynamical considerations were addressed with SPOCK following Tamayo et al. (2020). Posterior samples were weighted by predicted long-term stability; this down-weighted high-eccentricity samples for b and c, but the median parameters remained essentially unchanged. The adjacent pairs in the five-planet configuration febcd have mutual Hill separations of $M_* = 0.495 \pm 0.049~M_\odot$00, $M_* = 0.495 \pm 0.049~M_\odot$01, $M_* = 0.495 \pm 0.049~M_\odot$02, and $M_* = 0.495 \pm 0.049~M_\odot$03. These values are described as typical of compact multiplanet systems and consistent with stable orbital spacing (Hartogh et al., 9 Feb 2026).

The putative f:e pair is of particular dynamical interest. Their period ratio is $M_* = 0.495 \pm 0.049~M_\odot$04, close to a 2:1 mean-motion resonance. The posterior distribution of the resonant angle $M_* = 0.495 \pm 0.049~M_\odot$05 is reported as strongly clustered, with resultant length $M_* = 0.495 \pm 0.049~M_\odot$06, near $M_* = 0.495 \pm 0.049~M_\odot$07, and dispersion $M_* = 0.495 \pm 0.049~M_\odot$08 rad, which is suggestive of libration. The study nevertheless states that further ultra-precise RVs and dedicated N-body integrations are required to determine whether the angle truly librates rather than circulates.

Habitability considerations center on GJ 887 d. With $M_* = 0.495 \pm 0.049~M_\odot$09 and $M_* = 0.495 \pm 0.049~M_\odot$10–$M_* = 0.495 \pm 0.049~M_\odot$11 K for Bond albedo $M_* = 0.495 \pm 0.049~M_\odot$12 and zero heat redistribution, the planet is stated to lie comfortably within the Kopparapu et al. (2014) habitable zone for a low-mass planet orbiting an M1V star. The estimated habitable-zone boundaries correspond to orbital periods of approximately 43–122 d, placing the 50.77 d orbit squarely within the zone. By distance, GJ 887 d is described as the second-closest known habitable-zone planet after Proxima Cen b, given Proxima Cen b at 1.30 pc and GJ 887 at 3.29 pc.

The stellar environment is treated cautiously. The star’s low average magnetic activity is cited as favorable, and Mesquita et al. (2022) is noted as suggesting that the space weather environment may be Earth-like. At the same time, occasional flares have been observed according to Loyd et al. (2020), and the paper emphasizes that actual habitability depends on the planet’s magnetic field and atmosphere, both of which are presently unknown.

7. Observational constraints, ambiguities, and future work

The system is currently constrained primarily by radial velocities rather than transits. TLS analyses of TESS Sectors 2, 28, and 69 found no transits at any of the RV periods. Using probabilistic mass–radius relations from Chen & Kipping (2017), the expected transit depths span from approximately $M_* = 0.495 \pm 0.049~M_\odot$13 for c down to approximately $M_* = 0.495 \pm 0.049~M_\odot$14 for f; these depths would in principle be detectable given the reported TESS precision. Their absence therefore supports a non-transiting architecture, and all reported masses remain minimum masses $M_* = 0.495 \pm 0.049~M_\odot$15. The paper notes that true masses will require inclination constraints, for example from future astrometry (Hartogh et al., 9 Feb 2026).

The analysis also includes explicit sensitivity tests. Synthetic-signal injection and recovery showed that short-period signals at $M_* = 0.495 \pm 0.049~M_\odot$16 and $M_* = 0.495 \pm 0.049~M_\odot$17 were recovered with $M_* = 0.495 \pm 0.049~M_\odot$18 and $M_* = 0.495 \pm 0.049~M_\odot$19, respectively, consistent with the empirical behavior of e and f. A long-period 66 d signal at $M_* = 0.495 \pm 0.049~M_\odot$20, chosen to resemble d, was recovered strongly but weakened when the semi-amplitude was reduced to $M_* = 0.495 \pm 0.049~M_\odot$21, demonstrating the possibility of GP absorption near $M_* = 0.495 \pm 0.049~M_\odot$22. A formal detection-limit analysis in the five-planet residuals found a mean threshold of $M_* = 0.495 \pm 0.049~M_\odot$23 over 0.5–100 d, approximately 23 cm s$M_* = 0.495 \pm 0.049~M_\odot$24 near 2 d, and approximately 26 cm s$M_* = 0.495 \pm 0.049~M_\odot$25 beyond 20 d.

Several limitations remain material to interpretation. The paper explicitly cautions that GP kernels can absorb planetary power near the rotation period or its harmonics, and that the evidential strength of the 50.77 d signal depends somewhat on the activity model. It also stresses that the 2.2166 d candidate may represent a genuine resonant planet, but that an eccentric single-planet solution could mimic the pair at 2:1, following the alternative explanation discussed by Anglada-Escudé (2010). The rotational 38–39 d signature, by contrast, is regarded as activity because it is incoherent across the full baseline.

Future observing strategies are correspondingly targeted. The RedDots program emphasizes nightly observing to reduce aliasing and to capture activity cycles, and the paper identifies continued sub-m s$M_* = 0.495 \pm 0.049~M_\odot$26 RV monitoring with ESPRESSO, NIRPS, and future facilities as the principal route to confirming f, refining eccentricities, and constraining resonant dynamics. It also highlights coordinated photometry, such as with CHEOPS, to search for shallower or longer-duration transits missed by TESS.

GJ 887 d is additionally discussed as a possible target for future direct imaging. Assuming Earth-like albedo $M_* = 0.495 \pm 0.049~M_\odot$27, the expected reflected-light properties are angular separation $M_* = 0.495 \pm 0.049~M_\odot$28 mas and contrast $M_* = 0.495 \pm 0.049~M_\odot$29. For the Habitable Worlds Observatory (HWO) concept, the relevant benchmark cited is a contrast floor of approximately $M_* = 0.495 \pm 0.049~M_\odot$30–$M_* = 0.495 \pm 0.049~M_\odot$31 with inner working angle $M_* = 0.495 \pm 0.049~M_\odot$32 mas in Tier C. On those numbers, the paper states that the brightness ratio is well above the contrast floor, while the principal difficulty is that the separation lies almost exactly at the IWA. LIFE, the proposed mid-IR interferometer, is also identified as a relevant future facility because of the star’s brightness and proximity.