Evidence for LP 890-9d via Transit Timing Variations

Abstract: LP 890-9, also known as SPECULOOS-2 and TOI-4306, is a nearby late-M dwarf hosting two confirmed transiting rocky exoplanets. We analyze 20 JWST/NIRSpec PRISM transits of LP 890-9b and LP 890-9c obtained as part of GO program 7073 and detect statistically significant transit timing variations (TTVs), with peak-to-peak amplitudes of ~17 s and ~35 s, respectively. Using analytic linear TTV theory, we find that the known two-planet configuration cannot reproduce the measured TTV amplitudes or super-period, whereas three-planet models provide substantially better fits. The best-fit configuration places the candidate third planet, LP 890-9d, between planets b and c, with an orbital period of ~4.4 days; however, the current data do not uniquely determine its orbital architecture, and periods spanning 4.0-6.9 days remain plausible. TESS is insensitive to transits of LP 890-9d and we find no evidence for the candidate in JWST observations, although the phase coverage (ranging from ~50% to ~80%) depends strongly on the candidate orbital period. Additional high-precision transit observations of LP 890-9b and LP 890-9c are needed to refine their TTV solutions and further constrain the orbital properties of the third planet.

• - The paper identifies a third planet, LP 890-9d, in the LP 890-9 system using statistically significant transit timing variations (TTVs) measured by JWST/NIRSpec PRISM.
• - The favored three-planet model places the candidate planet between LP 890-9b and LP 890-9c, suggesting a period of approximately 4.4 days for LP 890-9d.
• - The study identifies limitations and open questions in current data, particularly the asteronomic validity of observed planet b timing deviations and the dependency on JWST timings for the three-planetary model.

The study presents a transit-timing analysis of LP 890-9, a nearby late-M dwarf hosting two confirmed transiting terrestrial-size planets. Using 20 JWST/NIRSpec PRISM transit observations, the authors report statistically significant deviations from linear ephemerides for both known planets and interpret the timing pattern as evidence for an additional, currently unconfirmed planet, designated LP 890-9d (2609.05312). The central result is not a unique orbital characterization, but rather a dynamical inconsistency between the observed timing signals and the known two-planet architecture. The favored three-planet solution places the candidate between LP 890-9b and LP 890-9c, although a broad range of orbital periods remains viable.

Observational basis and timing extraction

The analysis uses eight JWST transits of LP 890-9b and twelve of LP 890-9c. The data were independently reduced with the Tswift and Eureka! pipelines. Both reductions produce mutually consistent transit times, providing an important check against reduction-specific systematics. The timing uncertainties are nevertheless dominated by time-correlated noise rather than photon noise. For Tswift, the authors use circular residual-permutation bootstrap estimates, while the Eureka! analysis inflates formal uncertainties according to excess noise measured on ingress and egress timescales.

The resulting per-transit timing uncertainties are approximately $1.6$–$2.4$ seconds for planet b and $1.7$–$3.9$ seconds for planet c in the Tswift analysis. In the Eureka! reductions, the residual RMS on five-minute timescales exceeds the nominal white-noise expectation by factors of approximately $1.9$–$2.5$ for b and $1.1$–$2.2$ for c. These procedures are conservative relative to relying on formal MCMC posteriors alone, but the inference remains sensitive to how correlated noise is modeled. The agreement between independent pipelines mitigates, but does not eliminate, concerns that low-level instrumental or reduction systematics could mimic several-second timing offsets.

The measured TTVs have peak-to-peak amplitudes of approximately $17$ seconds for LP 890-9b and $35$ seconds for LP 890-9c. A sinusoidal analysis of the planet-c timings yields a semi-amplitude of roughly $2.4$0 seconds and a super-period near $2.4$1 days. These signals are substantially larger than the approximately $2.4$2–$2.4$3 second variations expected from the previously known two-planet system, which were inaccessible to earlier ground-based observations with typical timing errors of about one minute.

Testing the two-planet architecture

The authors fit the transit times using both analytic linear TTV theory implemented in TTV2Fast2Furious and direct $2.4$4-body integrations with TTVFast. The analytic model simultaneously fits the linear ephemerides and dynamical perturbations, thereby avoiding the covariance that arises when a timing signal is first detrended against a separately fitted ephemeris. The fitted parameters include periods, transit epochs, planet-to-star mass ratios, and the eccentricity-vector components $2.4$5 and $2.4$6.

The two-planet model, denoted the bc solution, fails in two related respects. It produces peak-to-peak TTV amplitudes of only $2.4$7 seconds for planet b and $2.4$8 seconds for planet c, and its super-period is approximately $2.4$9 days rather than the observed value near $1.7$0 days. Attempts to increase the planet-c timing amplitude require masses for b and c that are inconsistent with the adopted mass priors and still do not reproduce the overall timing structure. The two-planet model has $1.7$1, compared with $1.7$2 for the favored interior three-planet solution.

The absence of a clear anti-correlation between the b and c TTVs also disfavors a simple two-planet interpretation, because interacting two-planet systems generally exhibit approximately anti-correlated timing variations. This argument is explicitly limited by the short 43-day baseline for planet b, which covers only about 40% of the inferred super-period. Consequently, the lack of observed anti-correlation is supporting evidence rather than a decisive test.

The principal dynamical claim is therefore that the known two-planet configuration cannot simultaneously reproduce the observed amplitudes and super-period. This claim is stronger for planet c, whose timing signal is sampled more extensively, than for planet b. Whether the apparent planet-b TTVs are astrophysical is particularly important because they drive the preference for a perturber interior to planet c.

Degenerate orbital solutions for LP 890-9d

The inversion from TTVs to companion properties is highly non-unique in this dataset. The authors first identify candidate periods by matching the observed super-period to resonant TTV super-periods over a grid of period ratios and resonance indices. This procedure produces 26 candidate families: 12 interior solutions with periods from $1.7$3 to $1.7$4 days and 14 exterior solutions with periods from $1.7$5 to $1.7$6 days.

The global dynamical fits select representative solutions rather than exhausting the full posterior structure. For an interior perturber, the best-fit period is approximately $1.7$7 days. For an exterior perturber, the best-fit period is approximately $1.7$8 days. These should not be interpreted as competing sharply measured orbital solutions. The analysis explicitly finds that periods between approximately $1.7$9 and $3.9$0 days remain plausible for an interior planet, and the phase-space scan identifies additional exterior possibilities.

The representative interior, or bdc, solution provides the strongest fit. It produces peak-to-peak amplitudes of $3.9$1 seconds for planet b and $3.9$2 seconds for planet c, substantially improving the description of the observed timing pattern. Its inferred candidate-planet mass is $3.9$3 under the model assumptions. The independent TTVFast and NbodyGradient analyses yield similar periods near $3.9$4 days and candidate masses spanning approximately $3.9$5–$3.9$6, supporting the broad dynamical interpretation while not resolving the detailed posterior degeneracy.

The exterior bcd solution reproduces a substantial planet-c TTV amplitude, approximately $3.9$7 seconds, but leaves the planet-b amplitude nearly unchanged from the inadequate two-planet model. Its inferred candidate mass is lower, $3.9$8, with an Earth-like-density radius estimate of approximately $3.9$9. The poorer treatment of planet b yields $1.9$0 relative to the interior solution. Within the timing-only dynamical comparison, this is strong evidence for the ordering b–d–c if the planet-b timing measurements are reliable.

The posterior eccentricities of the known planets are not tightly constrained. In the favored bdc solution, the median eccentricities are approximately $1.9$1, $1.9$2, and $1.9$3, with uncertainties large enough that the detailed eccentricity structure should not be regarded as measured. The candidate’s mass and period are better viewed as correlated parameters along resonant TTV solution families than as independently determined quantities.

Transit and radial-velocity searches

The absence of a direct transit detection is consistent with, but does not establish, a non-transiting configuration. For candidate periods between approximately $1.9$4 and $1.9$5 days, the existing JWST observations cover about $1.9$6–$1.9$7 of orbital phase. Thus, assuming coplanarity and transitability, a transit would have had a substantial probability of occurring during the observations. However, a modest mutual inclination of roughly $1.9$8 could move a short-period planet outside the transit chord. The non-detection therefore does not strongly exclude the favored interior architecture.

The JWST data impose a stringent conditional constraint on any transit occurring in the searched baseline. For a representative $1.9$9-day orbit, the authors derive a $2.5$0 upper limit of $2.5$1 ppm on the transit depth, corresponding to a radius below $2.5$2. This limit is much smaller than the radii inferred from the TTV mass estimates, approximately $2.5$3 for the interior solution and $2.5$4 for the exterior representative solution. Consequently, a candidate with the predicted size would have been readily detectable if it had transited during the observed JWST baseline. The implication is conditional: the non-detection primarily constrains orbital phase and inclination, not the existence of the dynamical perturber.

TESS is insensitive to the expected sub-Earth radius. The phase-folded light curves have binned scatter of approximately $2.5$5–$2.5$6 ppm, corresponding to detection thresholds of roughly $2.5$7–$2.5$8. The predicted transit depths of approximately $2.5$9–$1.1$0 ppm fall below the standard $1.1$1 detection threshold. No candidate signal exceeds $1.1$2, but this null result cannot rule out the TTV-inferred planet.

The independent joint fits to TESS, ground-based transit photometry, and infrared radial velocities do not support a third planet when the JWST timings are excluded. In fact, the two-planet model has a BIC approximately 50 lower than the preferred three-planet configurations. Fits centered near $1.1$3 and $1.1$4 days are statistically indistinguishable from one another, while the $1.1$5-day configuration is less favorable by $1.1$6. The reported sensitivity to initialization, including redistribution of transit-like power among planets, indicates that these data are not independently informative enough to identify LP 890-9d.

Incorporating the JWST transit observations changes the model comparison substantially. The three-planet fits reproduce the measured planet-c timing pattern, whereas the two-planet fit is strongly disfavored with $1.1$7 relative to the three-planet models. The $1.1$8-day solution is preferred over the $1.1$9-day solution by only $2.2$0, while the $2.2$1-day solution is disfavored by $2.2$2. This contrast demonstrates that the evidential basis for LP 890-9d is specifically the JWST timing information, not an independent photometric or RV detection.

Dynamical stability and interpretation

The authors assess long-term secular stability using the angular momentum deficit criterion. Approximately $2.2$3 of posterior samples in the bdc configuration are AMD-stable, compared with $2.2$4 for the bcd configuration. The remaining samples are AMD-unstable, but AMD instability does not imply rapid disruption; it indicates that orbit crossing can be driven by secular evolution and requires further direct dynamical evaluation. Conversely, AMD stability does not account for all resonant or chaotic short-period effects. The stability analysis therefore excludes neither representative architecture, although the larger stable fraction of the exterior solution does not compensate for its poorer fit to the planet-b timings.

The inferred equilibrium temperature of the interior candidate is approximately $2.2$5 K under zero albedo and complete heat redistribution. This estimate follows from the representative orbital solution and is not a direct atmospheric or radius measurement. More importantly, the candidate’s expected mass and radius place it in the sub-Earth regime, explaining both the absence of a TESS detection and the difficulty of obtaining independent radial-velocity confirmation around a low-mass M dwarf.

Limitations and open questions

The principal limitation is the limited temporal baseline and sparse sampling. Planet b is observed over only 43 days, and the planet-c observations do not provide sufficient information to distinguish among the numerous resonant period families. The lack of detected chopping signals further weakens uniqueness. As emphasized by the paper, a sinusoidal TTV signal can be generated by multiple combinations of perturber period, mass, eccentricity, and phase.

The second limitation concerns the reliability of the planet-b timing signal. The favored interior configuration is selected primarily because it reproduces the nonzero timing deviations of b. Future observations of planet c alone can refine the super-period and candidate-planet parameter space, but they cannot directly establish whether the planet-b offsets are astrophysical. Additional high-precision transits of b are therefore necessary to validate the ordering b–d–c.

The mass estimates also depend on informative priors. The masses of b and c are not directly measured; their priors are inferred from radius-based mass–density relations. The candidate mass is consequently subject to correlations with the assumed masses and eccentricities of the known planets. The adopted Earth-like density relation used to convert candidate mass into radius is likewise an interpretive assumption rather than an observational constraint.

Finally, the two data analyses provide different conclusions depending on whether JWST timing measurements are included. Photometry and RVs alone favor the established two-planet model, while JWST timings strongly favor a three-planet dynamical model. This is not a contradiction in the statistical procedures, but it does mean that the third-planet interpretation rests on a signal that must be independently confirmed through additional timing measurements or a direct detection.

Conclusion

The paper identifies a statistically significant dynamical discrepancy between the known two-planet model of LP 890-9 and 20 high-precision JWST transit times. The observed TTV amplitudes of approximately $2.2$6 and $2.2$7 seconds, together with a super-period near $2.2$8 days, are not reproduced by the known planets. Three-planet models provide substantially better timing fits, with the strongest representative solution placing a sub-Earth-mass planet on an orbit near $2.2$9 days between LP 890-9b and LP 890-9c.

The evidence supports LP 890-9d as a planet candidate rather than a uniquely characterized planet. Periods from approximately $17$0 to $17$1 days remain viable, direct transit searches are conditional on orbital phase and inclination, and independent TESS, ground-based, and RV data do not favor a third planet without the JWST timing information. The decisive observational test is therefore additional high-precision timing of both known planets, particularly LP 890-9b, together with targeted searches for shallow transits across the remaining candidate period families.