betaFeedback

WorksPhysics and Astronomy

OGLE-2018-BLG-0829L: A Microlensing System with a Planet Orbiting a Star Possessing a Brown-Dwarf Companion

Abstract

We analyze the microlensing event OGLE-2018-BLG-0829, whose light curve exhibits a weak short-term anomaly in addition to the prominent caustic-crossing features of a binary lens. The anomaly cannot be explained by a binary-lens single-source (2L1S) model. We therefore investigate both binary-lens binary-source (2L2S) and triple-lens single-source (3L1S) interpretations and find that both models reproduce the additional anomaly, with the 3L1S model being favored by Δχ2=15.1Δχ^2=15.1. Under the favored 3L1S interpretation, the anomaly arises from a caustic induced by a planetary-mass companion, whereas the prominent caustic-crossing features are produced by the underlying binary lens. Combining the measured event timescale, tE=7.55±0.30t_{\rm E}=7.55\pm0.30 days, and angular Einstein radius, θE=0.325±0.059θ_{\rm E}=0.325 \pm 0.059 mas, with a Bayesian analysis, we infer a lens system consisting of an M dwarf of mass 0.193−0.098+0.300 M⊙0.193^{+0.300}_{-0.098}\,M_\odot, a possible brown-dwarf companion of mass 0.068−0.034+0.105 M⊙0.068^{+0.105}_{-0.034}\,M_\odot, and a Saturn-mass planet of mass 0.337−0.171+0.522 MJ0.337^{+0.522}_{-0.171}\,M_{\rm J}. The companions have projected separations of approximately 3.6 au and 1.6 au from the primary, respectively. However, the physical interpretation remains tentative because potentially important higher-order effects have not been considered, and the dynamical stability of the proposed configuration remains uncertain. If confirmed, this would be the third star–brown dwarf–planet system identified through microlensing, highlighting the potential of microlensing to detect planets in complex multiple systems.

(Leading authors)

(KMTNet Collaboration)

(The OGLE Collaboration)

Corresponding author: [email protected]

Corresponding author: [email protected]

Introduction

Most stars are thought to form in binary or higher-order multiple systems rather than in isolation [16, 39]. Consequently, understanding planet formation requires investigating not only planets orbiting single stars, but also those residing in dynamically complex multiple systems. The presence of a stellar or substellar companion can substantially modify the structure and evolution of a protoplanetary disk through tidal truncation, enhanced dynamical perturbations, and accelerated disk dispersal, thereby influencing both the formation and long-term survival of planets [5, 33]. Planetary systems in multiple-star environments therefore provide important observational constraints on theories of planet formation under conditions that differ markedly from those around isolated stars.

Brown dwarfs occupy the mass range between stars and planets and are thought to form through a variety of mechanisms, including turbulent fragmentation, disk fragmentation, and dynamical ejection [42, 35]. Their role in planetary systems remains poorly understood. In particular, whether planets can efficiently form and survive around stars possessing brown-dwarf companions remains an open question. Such systems probe a poorly explored regime in which the gravitational influence of the brown dwarf may significantly alter the conditions for planet formation and subsequent planetary evolution.

Gravitational microlensing is uniquely sensitive to planets located beyond the snow line around low-mass stars and is largely independent of the luminosities of either the host or the planet [36, 21, 17]. Unlike other detection methods, microlensing can detect planets orbiting faint M dwarfs, brown dwarfs, stellar remnants, and even free-floating objects (e.g., [9, 6, 18, 7] (P. Mróz et al. 2017)). It is also capable of revealing planets in binary and higher-order multiple systems through the characteristic perturbations they produce in lensing caustics [34, 25].

To date, only two microlensing planetary systems containing both a stellar host and a brown-dwarf companion have been reported. The first, KMT-2020-BLG-0414Lb [45], consists of an M dwarf hosting an Earth-mass planet together with a brown-dwarf companion. More recently, C. Han et al. [26] reported KMT-2024-BLG-0404L, in which a Uranus-mass planet orbits a close binary composed of a late M dwarf and a brown dwarf. Although these discoveries demonstrate that planets can exist in such environments, the sample remains extremely limited. Each additional system therefore provides valuable constraints on planet formation and dynamical evolution in the presence of brown-dwarf companions.

Interpreting multiple-lens microlensing events, however, is challenging because their anomalies can exhibit a wide variety of morphologies arising from the interplay of multiple caustics [11, 12, 10, 24]. This diversity frequently gives rise to degenerate or non-unique solutions, making detailed light-curve modeling essential for identifying the correct lens configuration. In some cases, the presence of an additional planet is revealed only through a subtle perturbation to the caustic structure generated by the underlying binary lens, making its detection considerably more difficult than in ordinary planetary microlensing events. Expanding the sample of such systems is therefore important not only for understanding planet formation in dynamically perturbed environments but also for exploring the full discovery space of gravitational microlensing.

In this paper, we present an analysis of the microlensing event OGLE-2018-BLG-0829, whose light curve exhibits a weak anomaly in addition to the prominent caustic-crossing features of a binary lens. We investigate two possible explanations for the weak anomaly: a binary-source interpretation and a triple-lens interpretation. Our analysis favors the triple-lens model, in which the lens system consists of a stellar primary and two lower-mass companions. According to the favored model, Bayesian analysis suggests that the system comprises an M dwarf, a brown dwarf, and a Saturn-mass planet, although the inferred masses are subject to substantial uncertainties. If confirmed, this would be the third star–brown-dwarf–planet system discovered by microlensing. This event illustrates the potential of microlensing to detect planets through subtle perturbations in binary-lens light curves, while highlighting the challenges of uniquely characterizing such complex lens systems.

Observations and data

The source star of the microlensing event OGLE-2018-BLG-0829 is located at equatorial coordinates (α,δ)J2000=(17 ⁣: ⁣55 ⁣: ⁣19.74,−29 ⁣: ⁣47 ⁣: ⁣08.9)(\alpha, \delta)_{\mathrm{J2000}} = (17\!:\!55\!:\!19.74, -29\!:\!47\!:\!08.9), corresponding to Galactic coordinates (l,b)=(0.∘3539,−2.∘2520)(l, b) = (0\mathord{.}^{\circ}3539, -2\mathord{.}^{\circ}2520). The event lies in a field with an II-band extinction of AI=1.68A_I = 1.68 [38] and has a baseline magnitude of Ibase=18.53I_{\mathrm{base}} = 18.53.

The event was discovered by the Optical Gravitational Lensing Experiment (OGLE) survey [41]. The OGLE survey conducts wide-field photometric monitoring of the Galactic bulge stars using the 1.3 m Warsaw Telescope at Las Campanas Observatory in Chile. Although the Korea Microlensing Telescope Network [32] group did not identify or issue an alert for the event during the 2018 observing season, the event was located in the region covered by the overlapping KMTNet prime fields BLG02 and BLG03. This enabled us to recover the light curve by performing post-season photometric reductions of the archived KMTNet images. KMTNet consists of three identical 1.6 m telescopes, each equipped with a 4 deg2^2 camera, located at the Cerro Tololo Inter-American Observatory in Chile (KMTC), the South African Astronomical Observatory in South Africa (KMTS), and the Siding Spring Observatory in Australia (KMTA). This global network provides nearly continuous photometric coverage of lensing events toward the Galactic bulge.

Our initial analysis was based on the OGLE photometry together with the standard KMTNet pipeline photometry. These data clearly revealed the major caustic-crossing features characteristic of a binary-lens event, leading us to initially interpret the event as a conventional binary-lens event. However, the relatively large photometric scatter in the pipeline reduction obscured weaker deviations from the binary-lens light curve, preventing us from recognizing the presence of an additional anomaly.

The subtle anomaly became apparent only after the KMTNet images were reprocessed using the pySIS photometric package developed by H. Yang et al. [43] for high-precision photometry of KMTNet microlensing events. The resulting light curve is presented in Figure 1. At first glance, the light curve is dominated by the characteristic signatures of a binary-lens caustic crossing: a pair of sharp caustic-crossing spikes at t2t_2 (HJD′' ≡\equiv HJD −2450000∼8255.7-2450000 \sim8255.7) and t3t_3 (HJD′' ∼8256.3\sim8256.3), followed by a broad peak at t4t_4 (HJD′' ∼8257.3\sim8257.3) produced by the source’s close approach to a caustic cusp. In addition to these prominent features, however, the improved photometry reveals a weak but statistically significant bump at t1t_1 (HJD′' ∼8255.6\sim8255.6), occurring shortly before the first caustic-crossing feature. This previously unrecognized feature motivates our investigation of more complex lensing configurations to determine its possible origin.

Figure 1. Light curve of OGLE-2018-BLG-0829. The bottom panel shows the full light curve, while the upper panels present an enlarged view of the anomalous region around the caustic-crossing features and the residuals from the close and wide 2L1S models. The dotted and solid curves represent the close and wide 2L1S solutions, respectively. The arrows labeled t1t_1, t2t_2, t3t_3, and t4t_4 indicate the four anomaly features discussed in the text. The two insets in the bottom panel illustrate the lens configurations of the close and wide 2L1S models. In each inset, the cuspy closed curve represents the caustic, and the arrowed line denotes the source trajectory.

Binary-lens single-source (2L1S) interpretation

The prominent caustic-crossing features indicate that the lens system contains multiple masses. We therefore begin by modeling the light curve with a binary-lens single-source (2L1S) configuration, the simplest lens model capable of producing caustics. The modeling is performed by searching for the set of lensing parameters that most accurately reproduces the observed light curve.

For a 2L1S event, the light curve is described by seven principal parameters. Three of these characterize the source trajectory relative to the lens, (t0,u0,tE)(t_0, u_0, t_{\mathrm{E}}), where t0t_0 is the time of closest approach between the source and the lens, u0u_0 is the impact parameter in units of the angular Einstein radius, and tEt_{\mathrm{E}} is the Einstein timescale. Two parameters, (s,q)(s, q), describe the binary lens, where ss is the projected separation between the lens components normalized to the angular Einstein radius and qq is their mass ratio. The remaining two parameters are the angle of the source trajectory with respect to the binary axis, α\alpha, and the normalized source radius, ρ=θ∗/θE\rho= \theta_*/\theta_{\mathrm{E}}, where θ∗\theta_* and θE\theta_{\mathrm{E}} denote the angular source radius and angular Einstein radius, respectively. The normalized source radius is included to account for finite-source effects during caustic crossings, which smooth the sharp magnification variations predicted for a point source because different parts of the extended source experience different magnifications [40, 20, 15].

A standard 2L1S analysis yields a pair of local solutions arising from the well-known close–wide degeneracy [14, 3]. This degeneracy occurs when binary lenses with separations smaller and larger than the angular Einstein radius produce similar central caustic structures, resulting in light curves with similar anomaly patterns. In this case, the close solution has lensing parameters (s,q,tE)∼(0.55,0.12,11.9 days)(s, q, t_{\mathrm{E}}) \sim(0.55, 0.12, 11.9\,\mathrm{days}), while the wide solution has (2.32,0.19,13.4 days)(2.32, 0.19, 13.4\,\mathrm{days}). Although both solutions reproduce the major features of the observed light curve, the wide solution provides a better fit, with Δχ2∼46.4\Delta\chi^2 \sim46.4. The model light curves of the two solutions are overlaid on the data in Figure 1.

The lens-system configurations corresponding to the close and wide 2L1S solutions are shown in the two insets of the bottom panel of Figure 1. In both cases, the source traverses the right side of the central caustic associated with the primary lens component (M1M_1). The caustic entrance and exit produce the sharp spikes at t2t_2 and t3t_3, respectively, while the subsequent close approach to the lower off-axis caustic cusp gives rise to the broad peak at t4t_4.

Although both solutions adequately reproduce the three major anomaly features at t2t_2, t3t_3, and t4t_4, neither accounts for the weak bump at t1t_1. The persistence of this residual indicates that the anomaly cannot be explained by a simple binary-lens geometry alone and instead requires an additional lens or source component.

Motivated by this discrepancy, we explore more complex lensing configurations in the following sections to investigate the possible origin of the t1t_1 anomaly and identify plausible interpretations of the event.

Binary-lens binary-source (2L2S) interpretation

We first tested a binary-lens binary-source (2L2S) model, motivated by the similarity of the observed anomaly to those in the microlensing events OGLE-2016-BLG-1003 [31] and KMT-2021-BLG-0209 [27]. In both events, the short-term anomaly was interpreted as resulting from the caustic crossing of a secondary source star located close to the primary source.

Compared to the 2L1S configuration, the 2L2S model introduces an additional source star, S2S_2. Modeling the anomaly produced by S2S_2 requires four additional lensing parameters, (t0,2,u0,2,ρ2,qF)(t_{0,2}, u_{0,2}, \rho_2, q_F). The parameters t0,2t_{0,2} and u0,2u_{0,2} represent the time at which S2S_2 approaches closest to the reference position of the lens and the corresponding lens–source separation, respectively. The third parameter, ρ2\rho_2, is the normalized source radius of S2S_2, while the fourth parameter, qFq_F, represents the flux ratio between the secondary and primary (S1S_1) source stars.

The 2L2S modeling was initialized with the best-fit 2L1S solution, which successfully reproduces the over-all light curve except for the anomaly at t1t_1. We then searched for the trajectory of the secondary source that best accounts for the timing and amplitude of the unexplained anomaly at t1t_1.

The modeling yielded a pair of local solutions originating from the close and wide 2L1S baseline models. We designate these as the “close” and “wide” 2L2S solutions, respectively. The lensing parameters of the two solutions are listed in Table 1 together with their χ2\chi^2 values. We find that the wide solution is favored over the close solution by Δχ2=36.0\Delta\chi^2 = 36.0. The model light curve of

Parameter2L2S Close2L2S Wide3L1S Close3L1S Wide
χ2\chi^22107.12071.12147.72056.0
t0t_0 (HJD′')8256.540 ±\pm 0.0158256.780 ±\pm 0.0178256.332 ±\pm 0.0138256.650 ±\pm 0.020
u0u_00.0443 ±\pm 0.00150.0544 ±\pm 0.00120.0796 ±\pm 0.00110.0761 ±\pm 0.0027
tEt_{\mathrm{E}} (days)13.16 ±\pm 0.4810.46 ±\pm 0.336.57 ±\pm 0.107.55 ±\pm 0.30
s2s_20.536 ±\pm 0.0122.156 ±\pm 0.0520.6200 ±\pm 0.00342.160 ±\pm 0.043
q2q_20.1217 ±\pm 0.00910.195 ±\pm 0.0170.204 ±\pm 0.0090.338 ±\pm 0.019
α\alpha (rad)5.591 ±\pm 0.0165.730 ±\pm 0.0155.458 ±\pm 0.0135.624 ±\pm 0.016
s3s_3⋯\cdots⋯\cdots1.1184 ±\pm 0.00200.9328 ±\pm 0.0087
q3q_3 (10−310^{-3})⋯\cdots⋯\cdots0.675 ±\pm 0.0791.57 ±\pm 0.18
ψ\psi (rad)⋯\cdots⋯\cdots0.660 ±\pm 0.0140.725 ±\pm 0.018
t0,2t_{0,2} (HJD′')8256.040 ±\pm 0.0168255.939 ±\pm 0.012⋯\cdots⋯\cdots
u0,2u_{0,2}−0.0365±0.0014-0.0365 \pm0.0014−0.0390±0.0010-0.0390 \pm0.0010⋯\cdots⋯\cdots
ρ1\rho_1 (10−310^{-3})0.53 ±\pm 0.020.77 ±\pm 0.031.214 ±\pm 0.0591.534 ±\pm 0.071
ρ2\rho_2 (10−310^{-3})1.34 ±\pm 0.311.40 ±\pm 0.32⋯\cdots⋯\cdots
qFq_{\mathrm{F}}0.196 ±\pm 0.0150.166 ±\pm 0.012⋯\cdots⋯\cdots

Table 1. Best-fit lensing parameters for the 2L2S and 3L1S models. Note— HJD′' ≡\equiv HJD −- 2450000.

the preferred wide solution is shown as the dotted curve in Figure 2. The model approximately reproduces all the observed anomaly features, including the previously unexplained bump at t1t_1.

Figure 2. Enlarged view of the anomaly region around the caustic-crossing features with the best-fit 2L2S and 3L1S models overlaid. The labels t1t_1–t3t_3 correspond to the anomaly features marked in Figure 1.

The lens-system configurations of the two 2L2S solutions are presented in Figure 3. The caustic geometries are nearly identical to those of the corresponding 2L1S solutions, and the trajectory of the primary source relative to the caustic remains essentially unchanged. The principal difference is the presence of the secondary source, which passes over the upper cusp of the caustic. The secondary source is relatively faint, with a flux ratio of qF∼0.17q_{\mathrm{F}} \sim0.17 relative to the primary source. It leads the primary source by Δt∼0.84\Delta t \sim0.84 days, entering the caustic ahead of the primary and thereby producing the anomaly at t1t_1. According to the model, the secondary source exits the caustic at around HJD′' ∼8255.8\sim8255.8, while the primary source is traversing the interior of the caustic. This caustic exit of S2S_2 produces a weak caustic-crossing feature, but the feature is only sparsely sampled by the observations.

Figure 3. Lens-system configurations of the close and wide 2L2S solutions. The black and blue arrowed lines represent the trajectories of the primary and secondary source stars, respectively. The points labeled t1t_1, t2t_2, t3t_3, and t4t_4 mark the source positions corresponding to the anomaly features indicated in Figure 1. In both panels, the secondary source crosses the upper cusp of the caustic, producing the anomaly at t1t_1.

Although the 2L2S model provides a reasonable fit to the data, several considerations argue against this interpretation. First, the inferred source properties appear difficult to reconcile with those expected for a typical binary-source system. The normalized source radii of the primary and secondary source stars inferred from the 2L2S model are ρ1=(0.77±0.03)×10−3\rho_1 = (0.77 \pm0.03) \times10^{-3} and ρ2=(1.40±0.32)×10−3\rho_2 = (1.40 \pm0.32) \times10^{-3}, respectively, implying that S2S_2 is larger than S1S_1. However, the measured flux ratio

is only qF=0.166±0.012q_{\mathrm{F}} = 0.166 \pm0.012, indicating that S2S_2 is substantially fainter than S1S_1. This relationship is illustrated in Figure 4, which shows the distribution of the Markov Chain Monte Carlo (MCMC) samples on the ρ2/ρ1\rho_2/\rho_1–qFq_{\mathrm{F}} plane. The MCMC samples favor ρ2/ρ1>1\rho_2/\rho_1 > 1 despite the small flux ratio. Although such a configuration is possible if the larger source has a sufficiently lower effective temperature, it would require an unusual combination of stellar properties for two stars with similar ages and metallicities, as expected for a physically bound binary. We therefore regard this configuration as less plausible, although it cannot be ruled out solely on the basis of the inferred source properties. Second, the 2L2S model is statistically disfavored relative to the 3L1S model discussed in the following subsection, providing a poorer fit by Δχ2=15.1\Delta\chi^2 = 15.1.

Figure 4. Scatter plot of the MCMC samples on the ρ2/ρ1\rho_2/\rho_1–qFq_{\mathrm{F}} plane for the wide 2L2S solution. The points are color-coded according to their confidence levels: <1σ<1\sigma (red), 11–2σ2\sigma (yellow), 22–3σ3\sigma (green), 33–4σ4\sigma (cyan), and 44–5σ5\sigma (blue).

Triple-lens single-source (3L1S) interpretation

We also tested a triple-lens single-source (3L1S) model, in which the lens consists of three masses and the source is a single star. Compared to the 2L1S configuration, the 3L1S model introduces an additional lens component and therefore requires three extra lensing parameters, (s3,q3,ψ)(s_3, q_3, \psi). Here, s3s_3 is the projected separation between the third lens component (M3M_3) and the primary lens (M1M_1), normalized to the angular Einstein radius. The parameter q3q_3 denotes the mass ratio of M3M_3 to M1M_1, and ψ\psi is the position angle of M3M_3 measured with respect to the primary–secondary binary-lens axis. To distinguish these parameters from those of the first companion (M2M_2), we denote the separation and mass ratio of the latter by (s2,q2)(s_2, q_2).

To characterize the third lens component, we first conducted a grid search over the (s3,q3,ψ)(s_3, q_3, \psi) parameter space. The remaining lensing parameters were initially fixed at their best-fit 2L1S values, because the 2L1S model adequately reproduces the overall light-curve morphology except for the anomaly at t1t_1. We then constructed a χ2\chi^2 map over the grid parameter space to identify local minima. Each identified minimum was subsequently refined by allowing all lensing parameters to vary.

respect to the primary–secondary binary-lens axis. To distinguish these parameters from those of the first companion (M2M_2), we denote the separation and mass ratio of the latter by (s2,q2)(s_2, q_2).

To characterize the third lens component, we first conducted a grid search over the (s3,q3,ψ)(s_3, q_3, \psi) parameter space. The remaining lensing parameters were initially fixed at their best-fit 2L1S values, because the 2L1S model adequately reproduces the overall light-curve morphology except for the anomaly at t1t_1. We then constructed a χ2\chi^2 map over the grid parameter space to identify local minima. Each identified minimum was subsequently refined by allowing all lensing parameters to vary.

This procedure yielded a pair of local solutions originating from the close and wide 2L1S solutions. The lensing parameters of the two 3L1S solutions are listed in Table 1. The wide solution is favored over the close solution by Δχ2=91.7\Delta\chi^2 = 91.7, a sufficiently large difference to resolve the close–wide degeneracy. The model light curve of the preferred wide solution, together with the corresponding residuals, is shown in Figure 2.

For the preferred wide solution, the triple-lens parameters are (s2,q2,s3,q3)∼(2.16,0.34,0.93,1.6×10−3)(s_2, q_2, s_3, q_3) \sim(2.16, 0.34, 0.93, 1.6 \times10^{-3}). These values suggest that the lens is a triple system consisting of a binary with a projected separation of approximately twice the angular Einstein radius and a low-mass third companion located at a projected separation close to the angular Einstein radius from the primary. The small mass ratio of the third component, q3∼1.6×10−3q_3 \sim1.6 \times10^{-3}, suggests its planetary nature. For this mass ratio, the companion mass would fall below the conventional planet–brown-dwarf boundary of ∼13 MJ\sim13\,M_{\mathrm{J}} for a primary mass below ∼7.8 M⊙\sim7.8\,M_{\odot}. As discussed in § 7, the Bayesian analysis favors a low-mass stellar primary, supporting the planetary interpretation. The Einstein timescale of the event, tE∼7.6t_{\mathrm{E}} \sim7.6 days, is relatively short, which may indicate a low-mass lens system. However, a short timescale can also result from a small lens–source relative parallax or a high relative proper motion, and therefore does not necessarily imply low lens masses.

The lens-system configurations of the close and wide 3L1S solutions are presented in Figure 5. In both solutions, the planet generates an additional caustic above the central caustic produced by the M1M_1–M2M_2 binary. Consistent with the binary-superposition approximation proposed by V. Bozza [10] and C. Han et al. [24], the overall caustic topology is well described by the superposition of the binary caustics associated with the M1M_1–M2M_2 and M1M_1–M3M_3 lens pairs.

Figure 5. Lens-system configurations of the close (upper panels) and wide (lower panels) 3L1S solutions. For each solution, the left panel shows the full caustic structure, while the right panel presents an enlarged view of the source-crossing region. The inset in each left panel shows the lens components as filled blue circles and the Einstein ring as a solid circle of unit radius. In the right panels, the points labeled t1t_1, t2t_2, t3t_3, and t4t_4 along the source trajectory indicate the source positions corresponding to the anomaly features marked in Figure 1. The gray curves surrounding the caustics are equal-magnification contours.

According to the preferred model, the source first enters the planet-induced caustic, producing the anomaly at t1t_1, and subsequently enters the binary-induced caustic, generating the caustic spike at t2t_2. It then exits the planet-induced caustic before leaving the binary caustic, producing a weak caustic-crossing feature at around HJD′∼8255.9\mathrm{HJD}^{\prime} \sim8255.9 within the U-shaped trough. This predicted feature was not covered by the observations. The source subsequently exits the binary caustic at t3t_3, producing the second caustic-crossing spike, and finally passes close to the lower cusp of the binary caustic after leaving the caustic, generating the broad peak at t4t_4.

Source star and angular Einstein radius

Characterizing the source star is an essential step in measuring the angular Einstein radius, θE\theta_{\mathrm{E}}. Once the source type is determined, its angular radius, θ∗\theta_{\ast}, can be estimated from its de-reddened color and magnitude. The angular Einstein radius is then obtained from

θE=θ∗ρ(1)\theta_{\mathrm{E}} = \frac{\theta_{\ast}}{\rho} \tag*{(1)}

where ρ\rho is the normalized source radius measured from the light-curve modeling. The angular Einstein radius provides an important constraint on the physical properties of the lens because it is related to the lens mass

and distance through

θE=(κMπrel)1/2,πrel=au(1DL−1DS).(2)\theta_{\mathrm{E}} = \left(\kappa M \pi_{\mathrm{rel}}\right)^{1/2}, \qquad\pi_{\mathrm{rel}} = \mathrm{au}\left(\frac{1}{D_{\mathrm{L}}} - \frac{1}{D_{\mathrm{S}}}\right). \tag*{(2)}

Here, κ=4G/(c2au)≃8.144 mas M⊙−1\kappa= 4G/(c^2\mathrm{au}) \simeq8.144\ \mathrm{mas}\ M_{\odot}^{-1}, MM is the total lens mass, and DLD_{\mathrm{L}} and DSD_{\mathrm{S}} denote the distances to the lens and source, respectively.

The source properties were determined from its position in the instrumental color–magnitude diagram (CMD) constructed using stars in the vicinity of the event. We determined the instrumental color and magnitude of the source by regressing the VV- and II-band photometry against the best-fit light-curve model. For OGLE-2018-BLG-0829, however, it was difficult to measure the source color using the survey data because the VV-band observations were both sparse and of low signal-to-noise ratio. To overcome this limitation, we used the CMD constructed from Hubble Space Telescope. We first aligned the instrumental CMD obtained from the KMTC observations with the HST CMD using red giant clump (RGC) centroids in the two data sets. The source color was then estimated from the mean color of HST stars lying within the range of the measured source II-band magnitude. Figure 6 shows the source position in the combined HST and KMTC CMD.

Figure 6. Source position in the instrumental color–magnitude diagram. Also marked is the centroid of the red giant clump (RGC), which is used to calibrate the source color and magnitude. The gray and brown points represent the CMDs constructed from the KMTC and HST observations, respectively.

The source color and magnitude were then calibrated relative to the RGC centroid using the method of J. Yoo et al. [44]. Because the source and the RGC lie along nearly the same line of sight, they are expected to experience approximately the same extinction and reddening. The de-reddened source color and magnitude were therefore obtained from

(V−I,I)S,0=(V−I,I)RGC,0+Δ(V−I,I),(3)(V-I,I)_{S,0}=(V-I,I)_{\mathrm{RGC},0}+\Delta(V-I,I), \tag*{(3)}

where Δ(V−I,I)=(V−I,I)S−(V−I,I)RGC\Delta(V-I,I)=(V-I,I)_S-(V-I,I)_{\mathrm{RGC}}, (V−I,I)S(V-I,I)_S and (V−I,I)RGC(V-I,I)_{\mathrm{RGC}} are the positions of the source and the RGC centroid in the instrumental CMD, respectively, and (V−I,I)RGC,0(V-I,I)_{\mathrm{RGC},0} is the known de-reddened position of the RGC centroid (T. Bensby et al. [8]; D. M. Nataf et al. [38]).

The uncertainties in the de-reddened source color and magnitude were estimated by adding the source measurement uncertainties in quadrature with the adopted uncertainties of 0.04 mag in the intrinsic RGC color and 0.02 mag in the intrinsic RGC magnitude. The source color was estimated from the mean color of HST stars within the magnitude range corresponding to the source brightness, and its uncertainty was determined from the standard deviation of the colors of the selected stars. In estimating the angular source radius, we additionally adopted a 7% fractional uncertainty to account for residual systematic uncertainties associated with the color–surface-brightness relation and photometric transformations. This contribution was added in quadrature to the uncertainty estimated from the de-reddened source color.

The measured and calibrated color and magnitude of the source, together with those of the RGC centroid, are listed in Table 2. The calibrated color and magnitude indicate that the source is an early K-type main-sequence star located in the Galactic bulge.

ParameterValue
(V−I,I)S(V-I,I)_S(1.878±0.157,20.930±0.069)(1.878 \pm0.157, 20.930 \pm0.069)
(V−I,I)RGC(V-I,I)_{\mathrm{RGC}}(2.065,16.103)(2.065, 16.103)
(V−I,I)RGC,0(V-I,I)_{\mathrm{RGC},0}(1.060,14.426)(1.060, 14.426)
(V−I,I)S,0(V-I,I)_{S,0}(0.873±0.162,19.253±0.072)(0.873 \pm0.162, 19.253 \pm0.072)

Table 2. Source parameters.

Using the estimated de-reddened source color and magnitude, we determined the angular source radius from the empirical color–surface-brightness relation of A. D. Adams et al. [1]. Applying this relation yields

θ∗=0.525±0.093 μas.(4)\theta_* = 0.525 \pm0.093~\mathrm{\mu as}. \tag*{(4)}

Combining the measured angular source radius with the normalized source radius using Equation (1), we obtained the angular Einstein radius,

θE=0.325±0.059 mas.(5)\theta_{\mathrm{E}} = 0.325 \pm0.059~\mathrm{mas}. \tag*{(5)}

Finally, the relative lens–source proper motion was estimated from the angular Einstein radius and the event timescale,

μ=θEtE=16.5±3.1 mas yr−1.(6)\mu= \frac{\theta_{\mathrm{E}}}{t_{\mathrm{E}}} = 16.5 \pm3.1~\mathrm{mas~yr}^{-1}. \tag*{(6)}

Physical lens parameters

The physical parameters of the lens cannot be uniquely determined from the light curve alone because the microlens-parallax parameter, πE\pi_{\mathrm{E}}, was not measured. We therefore estimated the properties of the lens system by conducting a Bayesian analysis using the constraints provided by the measured event timescale, tEt_{\mathrm{E}}, and angular Einstein radius, θE\theta_{\mathrm{E}}.

For the Bayesian analysis, we generated a large ensemble of artificial microlensing events based on a Galactic model incorporating the spatial distributions, kinematics, and mass functions of disk and bulge lens populations. We adopted the Galactic model of Y. K. Jung et al. [29] together with the lens mass function of Y. K. Jung et al. [30]. For each simulated event, we computed the lensing observables, tE,it_{\mathrm{E},i} and θE,i\theta_{\mathrm{E},i}, from the simulated lens mass, distance, and transverse velocity, and assigned a likelihood weight

wi=exp⁡{−12[(tE,i−tEσ(tE))2+(θE,i−θEσ(θE))2]}.(7)w_i = \exp\left\{-\frac{1}{2}\left[\left(\frac{t_{\mathrm{E},i}-t_{\mathrm{E}}}{\sigma(t_{\mathrm{E}})}\right)^2+\left(\frac{\theta_{\mathrm{E},i}-\theta_{\mathrm{E}}}{\sigma(\theta_{\mathrm{E}})}\right)^2\right]\right\}. \tag*{(7)}

The posterior distributions of the physical lens parameters were then constructed from the weighted ensemble of simulated events.

From these posterior distributions, we estimated the primary-lens mass, M1M_1, and lens distance, DLD_{\mathrm{L}}. For each parameter, we adopted the median of its posterior distribution as the representative value and the 16th and 84th percentiles as the lower and upper bounds of the uncertainty range, respectively. The masses of the two companions were subsequently derived using the measured mass ratios,

M2=q2M1,M3=q3M1.(8)M_2 = q_2 M_1,\qquad M_3 = q_3 M_1. \tag*{(8)}

The projected separations of the companions from the primary were calculated as

aL,2=s2DLθE,aL,3=s3DLθE.(9)a_{\mathrm{L},2}=s_2D_{\mathrm{L}}\theta_{\mathrm{E}},\qquad a_{\mathrm{L},3}=s_3D_{\mathrm{L}}\theta_{\mathrm{E}}. \tag*{(9)}

Figure 7 presents the Bayesian posterior distributions of the host-star mass, M1M_1 (upper panel), and the lens distance, DLD_{\mathrm{L}} (lower panel). The black histograms represent the total posterior distributions, while the red and blue histograms show the contributions from Galactic bulge and disk lenses, respectively. In each panel, the three vertical lines mark the 16th, 50th, and 84th percentiles of the total posterior distribution, with the solid line indicating the median.

Figure 7. Bayesian posterior distributions of the host-star mass, M1M_1 (upper panel), and the lens distance, DLD_{\mathrm{L}} (lower panel). The black histograms show the total posterior distributions, while the red and blue histograms represent the contributions from Galactic bulge and disk lenses, respectively. For each distribution, the solid vertical line indicates the median (50th percentile), while the two dotted vertical lines mark the 16th and 84th percentiles of the total posterior distribution.

The posterior mass distributions of the secondary and tertiary lens components are not shown separately because their masses are related to the primary mass by M2=q2M1M_2=q_2M_1 and M3=q3M1M_3=q_3M_1, respectively. Given the relatively small uncertainties in the mass ratios, their posterior distributions in logarithmic mass are essentially shifted versions of the primary mass distribution, with offsets of log⁡q2\log q_2 and log⁡q3\log q_3, respectively.

The estimated masses of the individual lens components, the lens distance, and the projected separations are summarized in Table 3. The quoted values correspond to the medians of the posterior distributions, while the uncertainties represent the 16th and 84th percentiles. The table also lists the relative probabilities that the lens lies in the Galactic disk (pdiskp_{\mathrm{disk}}) or bulge (pbulgep_{\mathrm{bulge}}).

ParameterValue
M1M_1 (M⊙M_{\odot})0.193−0.098+0.3000.193^{+0.300}_{-0.098}
M2M_2 (M⊙M_{\odot})0.068−0.034+0.1050.068^{+0.105}_{-0.034}
M3M_3 (MJM_{\mathrm{J}})0.337−0.171+0.5220.337^{+0.522}_{-0.171}
DLD_{\mathrm{L}} (kpc)6.54−1.33+1.036.54^{+1.03}_{-1.33}
a⊥,2a_{\perp,2} (au)3.58−0.73+0.573.58^{+0.57}_{-0.73}
a⊥,3a_{\perp,3} (au)1.56−0.32+0.251.56^{+0.25}_{-0.32}
pdiskp_{\mathrm{disk}}29%
pbulgep_{\mathrm{bulge}}71%

Table 3. Physical lens parameters estimated through Bayesian analysis.

Bases on the median values, the Bayesian estimates indicate that the primary lens, M1M_1, is a low-mass star, while its binary companion, M2M_2, lies in the brown-dwarf mass regime. Because of the large range of uncertain-

ties, M1M_1 has a probability of brown dwarf of ∼14%\sim14\% with a joint probability that both M1M_1 and M2M_2 are brown dwarfs is ∼11%\sim11\%. The third component, M3M_3, has a mass in the planetary regime, approximately comparable to that of Saturn. The brown-dwarf companion is located at a projected separation of a⊥,2=3.58−0.73+0.57 aua_{\perp,2}=3.58^{+0.57}_{-0.73}\,\mathrm{au} from the primary, whereas the planet lies at a projected separation of a⊥,3=1.56−0.32+0.25 aua_{\perp,3}=1.56^{+0.25}_{-0.32}\,\mathrm{au}.

The relatively high lens–source proper motion, μ=16.5±3.1 mas yr−1\mu=16.5\pm3.1\,\mathrm{mas\,yr}^{-1}, favors a nearby or kinematically fast lens. Nevertheless, because neither the microlens parallax nor the lens flux was measured, the individual masses and distance remain subject to substantial uncertainties. Future high-angular-resolution imaging may resolve the lens from the source and thereby measure the lens flux and relative proper motion independently, providing a direct test of the inferred star–brown dwarf–planet interpretation.

Discussion

We analyzed the microlensing event OGLE-2018-BLG-0829, whose light curve exhibits a weak anomaly in addition to prominent caustic-crossing features characteristic of a binary lens. Although the overall light-curve morphology is well reproduced by a 2L1S model, the weak anomaly preceding the first caustic crossing cannot be explained within this framework. We therefore explored both 2L2S and 3L1S interpretations and found that both models can reproduce the additional anomaly. However, the 2L2S model implies an unusual combination of source properties and is statistically disfavored relative to the 3L1S model by Δχ2=15.1\Delta\chi^2=15.1. We therefore regard the 3L1S interpretation as the more plausible explanation among the models considered.

Under the favored 3L1S interpretation, the lens may be a hierarchical system consisting of a low-mass M dwarf, a brown-dwarf candidate, and a Saturn-mass planet. Bayesian analysis suggests that the two companions lie at projected separations of approximately 3.6 au and 1.6 au from the primary, respectively. However, the inferred component masses and orbital configuration remain uncertain, as discussed below. If this interpretation is confirmed, OGLE-2018-BLG-0829L would represent the third star–brown dwarf–planet system identified through microlensing.

The potential significance of this system lies in the insight it may provide into planet formation and survival in star–brown dwarf environments. Together with KMT-2020-BLG-0414Lb (W. Zang et al. [45]) and KMT-2024-BLG-0404L (C. Han et al. [26]), OGLE-2018-BLG-0829L would contribute to a small but diverse sample of planetary systems containing brown-dwarf companions. These three systems would span nearly three orders of magnitude in planetary mass, ranging from an Earth-mass planet to a Saturn-mass planet. Although the sample remains far too small to support statistical conclusions, the diversity of these systems suggests that planet formation may occur under a range of conditions in systems containing brown-dwarf companions, motivating further observational and theoretical investigations.

Beyond the possible identification of a planet in a binary system, this event is also notable from a methodological perspective. The planetary signal manifests as a subtle deviation in the light curve superposed on the prominent caustic-crossing features produced by the underlying binary lens. Although the planet substantially modifies the caustic structure, its photometric signature is relatively weak and would likely have remained undetected without the high-precision re-reduction of the KMTNet data. This demonstrates that improved photometric precision can reveal planetary signals hidden in apparently well-understood binary-lens events. As photometric reduction techniques continue to improve and increasingly large microlensing data sets become available, similar weak perturbations may reveal additional planets in complex multiple-lens systems.

The relatively high lens–source relative proper motion makes this event a promising target for high-angular-resolution follow-up observations. With a measured relative proper motion of μ∼17\mu\sim17 mas yr−1\mathrm{yr}^{-1}, the lens and source are expected to be separated by approximately 130 mas eight years after the event, potentially allowing them to be resolved with current adaptive-optics facilities such as Keck. A direct measurement of the lens flux would provide additional constraints on the mass and distance of the luminous lens component, thereby offering an independent test of the physical parameters inferred under the favored 3L1S interpretation.

Limitations of the Current Analysis

Although the 3L1S model provides the best fit among the models considered, the present analysis should be regarded as a preliminary exploration of plausible interpretations rather than a definitive characterization of the lens system. Several effects not included in the modeling could influence the inferred lens parameters and potentially alter the preferred interpretation.

First, the modeling did not include microlens parallax ([19] (J. H. An et al. 2002)) or orbital motion of the lens components ([13] (M. D. Albow et al. 2000)). Although the relatively short event timescale may limit the sensitivity to parallax, the presence of multiple sharp light-curve features could provide additional constraints on this effect. Similarly, orbital motion could modify the caustic geometry during the event and affect the relative goodness of fit of the competing solutions. Including these effects could therefore change the best-fit parameters and increase their uncertainties.

Second, the substantial blending may introduce a degeneracy between the blending flux and the Einstein timescale, potentially increasing the uncertainties in the inferred lens parameters ([22, 23]). If a significant fraction of the blended light originates from the lens system, its brightness could provide an additional constraint on the lens mass and distance, thereby affecting the Bayesian interpretation. An astrometric analysis of the blended light, together with future high-angular-resolution observations, could help determine its origin and refine the inferred physical properties of the lens system.

Another important consideration is the dynamical stability of the proposed system. The inferred projected separations of the brown-dwarf candidate and the planet are approximately 3.6 au and 1.6 au, respectively, corresponding to a separation ratio of only ∼2.3\sim2.3. For a planet orbiting one component of a binary, long-term stability generally requires the planetary orbit to be substantially smaller than the binary orbit ([28]). The relatively small projected separation ratio therefore raises questions about the dynamical stability of the proposed configuration. However, the projected separations do not directly determine the orbital semi-major axes, which depend on the unknown orbital orientations, phases, and eccentricities. A detailed dynamical analysis would therefore be required to assess the long-term stability of the proposed system.

In addition, although the favored 3L1S model reproduces the principal features of the observed light curve, small residuals remain around the planetary anomaly. These residuals may indicate limitations of the static-lens approximation or the presence of alternative solutions that were not identified in the present search. More extensive modeling, incorporating additional physical effects and a broader exploration of parameter space, would be needed to establish the uniqueness and physical plausibility of the favored interpretation.

Taken together, these considerations indicate that the proposed star–brown dwarf–planet configuration remains tentative. Nevertheless, the present analysis identifies a plausible interpretation of the observed anomaly and provides a basis for further investigation.

Acknowledgements

C.H. was supported by the National Research Foundation grant (RS-2025-21073000) funded by the Korea government. This research has made use of the KMTNet system operated by the Korea Astronomy and Space Science Institute (KASI) at three host sites of CTIO in Chile, SAAO in South Africa, and SSO in Australia. Data transfer from the host site to KASI was supported by the Korea Research Environment Open NETwork (KREONET). This research was supported by KASI under the R&D program (project No. 2026-1-904-01) supervised by the Ministry of Science and ICT. Q.Q., W.Z., H.Y., H.L., J.Z. and S.M. acknowledge support by the National Natural Science Foundation of China (Grant No. 12133005). The OGLE project has received funding from the Polish National Science Centre grant OPUS-28 2024/55/B/ST9/00447 awarded to AU.

References

  1. [1]Adams, A. D., Boyajian, T. S., & von Braun, K. 2018, MNRAS, 473, 3608, doi: 10.1093/mnras/stx2367DOI
  2. [3]An, J. H. 2005, MNRAS, 356, 1409, doi: 10.1111/j.1365-2966.2004.08559.x
  3. [5]Artymowicz, P., & Lubow, S. H. 1994, ApJ, 421, 651, doi: 10.1086/173679DOI
  4. [6]Beaulieu, J.-P., Bennett, D. P., Fouqué, P., et al. 2006, Nature, 439, 437, doi: 10.1038/nature04441DOI
  5. [7]Bennett, D. P., Batista, V., Bond, I. A., et al. 2014, ApJ, 785, 155, doi: 10.1088/0004-637X/785/2/155DOI
  6. [8]Bensby, T., Yee, J. C., Feltzing, S., et al. 2013, A&A, 549, A147, doi: 10.1051/0004-6361/201220678DOI
  7. [9]Bond, I. A., Udalski, A., Jaroszyński, M., et al. 2004, ApJ, 606, L155, doi: 10.1086/420928DOI
  8. [10]Bozza, V. 2000, A&A, 355, 423–432, doi: 10.48550/arXiv.astro-ph/9910535DOI
  9. [11]Daněk, K., & Heyrovský, D. 2015, ApJ, 806, 99, doi: 10.1088/0004-637X/806/1/99DOI
  10. [12]Daněk, K., & Heyrovský, D. 2019, ApJ, 880, 72, doi: 10.3847/1538-4357/ab2982DOI
  11. [13]Dominik, M. 1998, A&A, 329, 361
  12. [14]Dominik, M. 1999, A&A, 349, 108
  13. [15]Dominik, M. 2004, MNRAS, 353, 69, doi: 10.1111/j.1365-2966.2004.08046.xDOI
  14. [16]Duquennoy, A., & Mayor, M. 1991, A&A, 248, 485
  15. [17]Gaudi, B. S. 2012, ARA&A, 50, 411, doi: 10.1146/annurev-astro-081811-125518DOI
  16. [18]Gaudi, B. S., Bennett, D. P., Udalski, A., et al. 2008, Science, 319, 927, doi: 10.1126/science.1151947DOI
  17. [19]Gould, A. 1992, ApJ, 392, 442, doi: 10.1086/171443DOI
  18. [20]Gould, A. 1994, ApJL, 421, L71, doi: 10.1086/187190DOI
  19. [21]Gould, A., & Loeb, A. 1992, ApJ, 396, 104, doi: 10.1086/171700DOI
  20. [22]Han, C. 1997, ApJ, 490, 51, doi: 10.1086/304877DOI
  21. [23]Han, C. 2000, MNRAS, 312, 807, doi: 10.1046/j.1365-8711.2000.03171.xDOI
  22. [24]Han, C., Chang, H.-Y., An, J.-H., & Chang, K. 2001, MNRAS, 328, 986, doi: 10.1046/j.1365-8711.2001.04973.xDOI
  23. [25]Han, C., Udalski, A., Gould, A., et al. 2017, AJ, 154, 223, doi: 10.3847/1538-3881/aa9179DOI
  24. [26]Han, C., Udalski, A., Lee, C.-U., et al. 2025, A&A, 700, A181, doi: 10.1051/0004-6361/202452760DOI
  25. [27]Han, C., Lee, C.-U., Udalski, A., et al. 2026, ApJ, 1004, 183, doi: 10.3847/1538-4357/ae6ef2DOI
  26. [28]Holman, M. J., & Wiegert, P. A. 1999, AJ, 117, 621, doi: 10.1086/300695DOI
  27. [29]Jung, Y. K., Han, C., Udalski, A., et al. 2021, AJ, 161, 293, doi: 10.3847/1538-3881/abf8bdDOI
  28. [30]Jung, Y. K., Zang, W., Han, C., et al. 2022, AJ, 164, 262, doi: 10.3847/1538-3881/ac9c5cDOI
  29. [31]Jung, Y. K., Udalski, A., Bond, I. A., et al. 2017, ApJ, 841, 75, doi: 10.3847/1538-4357/aa7057DOI
  30. [32]Kim, S.-L., Lee, C.-U., Park, B.-G., et al. 2016, JKAS, 49, 37, doi: 10.5303/JKAS.2016.49.1.37DOI
  31. [33]Kraus, A. L., Ireland, M. J., Hillenbrand, L. A., & Martinache, F. 2012, ApJ, 745, 19, doi: 10.1088/0004-637X/745/1/19DOI
  32. [34]Lee, D.-W., Lee, C.-U., Park, B.-G., et al. 2008, ApJ, 672, 623, doi: 10.1086/523662DOI
  33. [35]Luhman, K. L. 2012, ARA&A, 50, 65, doi: 10.1146/annurev-astro-081811-125528DOI
  34. [36]Mao, S., & Paczyński, B. 1991, ApJL, 374, L37, doi: 10.1086/186066DOI
  35. [38]Nataf, D. M., Gould, A., Fouqué, P., et al. 2013, ApJ, 769, 88, doi: 10.1088/0004-637X/769/2/88DOI
  36. [39]Raghavan, D., McAlister, H. A., Henry, T. J., et al. 2010, ApJS, 190, 1, doi: 10.1088/0067-0049/190/1/1DOI
  37. [40]Schneider, P., & Weiss, A. 1987, A&A, 171, 49
  38. [41]Udalski, A., Szymański, M. K., & Szymański, G. 2015, Acta Astron., 65, 1, doi: 10.48550/arXiv.1504.05966DOI
  39. [42]Whitworth, A., Bate, M. R., Nordlund, Å., Reipurth, B., & Zinnecker, H. 2007, in Protostars and Planets V, ed. B. Reipurth, D. Jewitt, & K. Keil (Tucson: University of Arizona Press), 459–476
  40. [43]Yang, H., Yee, J. C., Hwang, K.-H., et al. 2024, Monthly Notices of the Royal Astronomical Society, 528, 11, doi: 10.1093/mnras/stad3672DOI
  41. [44]Yoo, J., DePoy, D. L., Gal-Yam, A., et al. 2004, ApJ, 603, 139, doi: 10.1086/381241DOI
  42. [45]Zang, W., Han, C., Kondo, I., et al. 2021, Research in Astronomy and Astrophysics, 21, 239, doi: 10.1088/1674-4527/21/9/239DOI

Paper details

Contents