- Split View
-
Views
-
Cite
Cite
Y. Shvartzvald, D. Maoz, S. Kaspi, T. Sumi, A. Udalski, A. Gould, D. P. Bennett, C. Han, F. Abe, I. A. Bond, C. S. Botzler, M. Freeman, A. Fukui, D. Fukunaga, Y. Itow, N. Koshimoto, C. H. Ling, K. Masuda, Y. Matsubara, Y. Muraki, S. Namba, K. Ohnishi, N. J. Rattenbury, To. Saito, D. J. Sullivan, W. L. Sweatman, D. Suzuki, P. J. Tristram, K. Wada, P. C. M. Yock, J. Skowron, S. Kozłowski, M. K. Szymański, M. Kubiak, G. Pietrzyński, I. Soszyński, K. Ulaczyk, Ł. Wyrzykowski, R. Poleski, P. Pietrukowicz, MOA-2011-BLG-322Lb: a ‘second generation survey’ microlensing planet, Monthly Notices of the Royal Astronomical Society, Volume 439, Issue 1, 21 March 2014, Pages 604–610, https://doi.org/10.1093/mnras/stt2477
- Share Icon Share
Abstract
Global ‘second-generation’ microlensing surveys aim to discover and characterize extrasolar planets and their frequency, by means of round-the-clock high-cadence monitoring of a large area of the Galactic bulge, in a controlled experiment. We report the discovery of a giant planet in microlensing event MOA-2011-BLG-322. This moderate-magnification event, which displays a clear anomaly induced by a second lensing mass, was inside the footprint of our second-generation microlensing survey, involving MOA, OGLE and the Wise Observatory. The event was observed by the survey groups, without prompting alerts that could have led to dedicated follow-up observations. Fitting a microlensing model to the data, we find that the time-scale of the event was tE = 23.2 ± 0.8 d, and the mass ratio between the lens star and its companion is q = 0.028 ± 0.001. Finite-source effects are marginally detected, and upper limits on them help break some of the degeneracy in the system parameters. Using a Bayesian analysis that incorporates a Galactic structure model, we estimate the mass of the lens at |$0.39^{+0.45}_{-0.19}\,{\rm M}_{\odot }$|, at a distance of 7.56 ± 0.91 kpc. Thus, the companion is likely a planet of mass |$11.6^{+13.4}_{-5.6}\, M_{\rm J}$|, at a projected separation of |$4.3^{+1.5}_{-1.2}$|au, rather far beyond the snow line. This is the first pure-survey planet reported from a second-generation microlensing survey, and shows that survey data alone can be sufficient to characterize a planetary model. With the detection of additional survey-only planets, we will be able to constrain the frequency of extrasolar planets near their systems’ snow lines.
1 INTRODUCTION
The discovery of thousands of extrasolar planets ranks among the most exciting scientific developments of the past decade. The majority of those exoplanets were detected and characterized by the transit and radial velocity methods, which favour the detection of massive planets in close orbits around their hosts, stars at distances within a few hundred parsec. Microlensing, in contrast, can reveal planets down to Earth mass and less (Bennett & Rhie 1996), at larger orbits – about 1–10 au, which is where the ‘snowline’ is located, and beyond which giant planets are expected to form according to planet formation models (Ida & Lin 2005). Microlensing enables the detection of planets around all types of stars at distances as far as the Galactic centre, and even planets unbound from any host star (Sumi et al. 2011). Although planets discovered with microlensing still number in the few tens, several teams (e.g. Gould et al. 2010; Sumi et al. 2010; Cassan et al. 2012) have attempted to estimate the frequency of planets at these separations. Gould et al. (2010) estimated from high-magnification events an ∼1/6 frequency of Solar-like systems. Sumi et al. (2010) found that Neptune-mass planets are three times more common than Jupiters beyond the snowline. Cassan et al. (2012) concluded that, on average, every star in the Galaxy hosts a snowline-region planet. Moreover, the fact that 2 out of the 20 planetary systems discovered by microlensing, OGLE-2006-BLG-109Lb,c (Gaudi et al. 2008) and OGLE-2012-BLG-0026Lb,c (Han et al. 2013), host two planets, suggests that multiple systems are common, as also indicated at smaller separations by transit data from Kepler (Howard 2013). Over the past decade, microlensing planet discoveries have largely come from observing campaigns in which specific, high-magnification (A ≳ 100), events are followed intensively by networks of small telescopes, in order to detect and characterize planetary anomalies in the light curves. High-magnification events are very sensitive to planets near their snowlines, but they are rare events. Furthermore, the inhomogeneous social process through which potentially high-magnification events are alerted and followed up complicates their use for statistical inferences on planet frequency.
Microlensing surveys have been in transition into the so-called second generation phase (Gaudi et al. 2009), wherein a large area of the Galactic bulge is monitored continuously, with cadences high enough to detect planetary anomalies without any follow-up observations or changes to the observing sequence (e.g. switching to a higher cadence). The first such survey began in 2011 and combines three groups: OGLE (Optical Gravitational Lensing Experiment) – observing from Chile (Udalski 2009), MOA (Microlensing Observations in Astrophysics) – observing from New-Zealand (Sumi et al. 2003) and the Wise survey observing from Israel (Shvartzvald & Maoz 2012). Shvartzvald & Maoz (2012) have simulated the results that can be expected from this ‘controlled experiment’, and its potential to measure the abundance of planetary systems. They found that the overall planet detection efficiency for the survey is ∼20 per cent. In the 2011 season, there was a total of 80 events inside the high-cadence survey footprint that is common to all three groups (an additional 218 events were observed by only two of the groups). Of those 80 events, at least three showed a clear planetary anomaly: MOA-2011-BLG-293 (Yee et al. 2012), MOA-2011-BLG-322 (this paper) and OGLE-2011-BLG-0265 (in preparation). Accounting for our detection efficiency, these results imply ∼1/5 frequency of planetary systems, which is in line with previous estimates by Gould et al. (2010).
In this paper, we present the analysis of MOA-2011-BLG-322Lb. This is the first planetary microlensing event that is detected and analysed based solely on second-generation survey data. In principle, OGLE-2003-BLG-235/MOA-2003-BLG-53Lb (Bond et al. 2004) and MOA-2007-BLG-192Lb (Bennett et al. 2008) were also discovered and characterized based only on MOA and OGLE data. This was fortuitously possible despite the sparse sampling of first generation surveys. MOA-2011-BLG-293Lb (Yee et al. 2012) was also characterizable by survey-only data, but it did include a large amount of non-survey data that were prompted by alerts, following early realization of the event's high magnification. We describe the observations by the three survey groups in Section 2. In Section 3, we present the binary microlensing model fitted to the event. A Bayesian analysis estimating the physical properties of the system is presented in Section 4, and we discuss our results in Section 5.
2 OBSERVATIONAL DATA
The microlensing event MOA-2011-BLG-322 was first detected on 2011 June 30 16:15 ut by MOA, who operate the 1.8 m MOA-II telescope at the Mt John Observatory in New Zealand. The source is located inside the second-generation survey footprint, at RA = 18:04:53.6, Dec. = −27:13:15.4 (J2000.0). Thus, the event was also observed by the Wise team with the 1 m telescope at the Wise Observatory in Israel, and by the OGLE team with the 1.3 m Warsaw telescope at the Las Campanas Observatory in Chile. It was discovered independently by the OGLE early warning system (Udalski 2003) and designated as OGLE-2011-BLG-1127. Since the event was not identified as interesting in real time (although MOA noted some anomalous behaviour), the survey teams continued their regular observing cadences for this field. The observational information for each group (filter, cadence, exposure time) is summarized in Table 1. The event was not observed by most of the microlensing follow-up groups such as RoboNet-II (Tsapras et al. 2009), MiNDSTEp (Dominik et al. 2010) or PLANET (Albrow et al. 1998) since it was not very bright at baseline, and had an apparently low magnification (∼1 mag, although we show in Section 3, below, that this is due to blending, and the true magnification was a moderate A = 23). We note that Farm Cove Observatory from μFun (Gould et al. 2010) tried to observe the event for a couple of nights after the anomaly, but the data were too noisy to be of use.
Group . | Filter . | Cadence . | Exp. time . |
---|---|---|---|
. | . | (min) . | (s) . |
MOA | MOA-Reda | 18 | 60 |
Wise | I | 30 | 180 |
OGLE | I | 60 | 100 |
Group . | Filter . | Cadence . | Exp. time . |
---|---|---|---|
. | . | (min) . | (s) . |
MOA | MOA-Reda | 18 | 60 |
Wise | I | 30 | 180 |
OGLE | I | 60 | 100 |
aEquivalent to R + I.
Group . | Filter . | Cadence . | Exp. time . |
---|---|---|---|
. | . | (min) . | (s) . |
MOA | MOA-Reda | 18 | 60 |
Wise | I | 30 | 180 |
OGLE | I | 60 | 100 |
Group . | Filter . | Cadence . | Exp. time . |
---|---|---|---|
. | . | (min) . | (s) . |
MOA | MOA-Reda | 18 | 60 |
Wise | I | 30 | 180 |
OGLE | I | 60 | 100 |
aEquivalent to R + I.
OGLE and MOA data were reduced by their standard difference image analysis (DIA) procedures (Bond et al. 2001; Udalski 2003). The Wise data were reduced using the pysis DIA software (Albrow et al. 2009). The MOA and Wise fluxes were aligned to the OGLE I-band magnitude scale, and inter-calibrated to the microlensing model (see Section 3). As further described below, this event includes a large amount of ‘blended light’, which could be due to unrelated stars projected near the source and lens stars, and/or due to the lens star itself. Re-reduction of the pipeline-reduced data from each observatory, including centroid alignment and correction for a trend of DIA flux with seeing width for MOA and Wise data (seen for this source in baseline data, before the event), corrected a number of measurements with systematic errors in the observatories’ pipeline reductions.
Fig. 1 shows the observed light curve, with a clear deviation from symmetric, point-mass microlensing (Paczynski 1986). We therefore proceed to the next level of complexity and attempt to model this event as a binary lens.
3 MICROLENSING MODEL
The possibility of finite source effects due to the non-zero size of the source star is included by allowing a source of angular radius ρ, relative to θE, assuming a limb-darkened profile with the ‘natural’ coefficient Γ (Albrow et al. 1999). Although the source colour cannot be measured directly, it is likely a main-sequence G-type star (see below). We therefore estimate the limb-darkening coefficients from Claret (2000), using effective temperature Teff = 5750 K, and gravity logg = 4.5, to be ΓI = 0.43 for OGLE and Wise, and ΓR/I = 0.47 for MOA.
In order to solve for the magnification of the binary-lens model, we use the ray-shooting light-curve generator described in Shvartzvald & Maoz (2012). Briefly, we construct a trial model of the binary lens with a given choice of the parameters in the problem. We divide the lens plane on to a grid and use the lens equation directly to map the lens plane on to the source plane. By calculating the entire source trajectory at once, and by using an adaptive grid that increases the lens-plane resolution around the image positions, we achieve fast computation times. The magnification is then the ratio of the summed solid angles subtended by all the images in the lens plane to that of the solid angle subtended by the source.
Fig. 1 shows the inter-calibrated light-curve of the event and our best-fitting model. The best-fitting parameters are given in Table 2, along with the re-normalization coefficients, ci, for each group (emin = 0 for all groups). The time-scale of the event, tE = 23.2 ± 0.8 d, suggests a sub-solar-mass primary lens. The mass ratio between the primary and the secondary is q = 0.028 ± 0.001, about 30 times the Jupiter/Sun ratio, near the brown-dwarf/planetary border for an M-type host star. As mentioned above, almost 95 per cent of the flux is blended light due to a star near the line of sight. We discuss this result and its implications in Section 5.
t0 (HJD] | 245 5774.2729(54) |
u0 | 0.046 86(17) |
tE (d) | 23.17(80) |
s | 1.822(10) |
α (rad) | 0.3662(36) |
q | 0.0284(10) |
ρ | <0.007 |
fs/fb (OGLE) | 0.0573(29) |
Isource (mag) | 19.83(5) |
Iblend (mag) | 16.69(2) |
cmoa | 1.50 |
cwise | 1.25 |
cogle | 1.57 |
t0 (HJD] | 245 5774.2729(54) |
u0 | 0.046 86(17) |
tE (d) | 23.17(80) |
s | 1.822(10) |
α (rad) | 0.3662(36) |
q | 0.0284(10) |
ρ | <0.007 |
fs/fb (OGLE) | 0.0573(29) |
Isource (mag) | 19.83(5) |
Iblend (mag) | 16.69(2) |
cmoa | 1.50 |
cwise | 1.25 |
cogle | 1.57 |
t0 (HJD] | 245 5774.2729(54) |
u0 | 0.046 86(17) |
tE (d) | 23.17(80) |
s | 1.822(10) |
α (rad) | 0.3662(36) |
q | 0.0284(10) |
ρ | <0.007 |
fs/fb (OGLE) | 0.0573(29) |
Isource (mag) | 19.83(5) |
Iblend (mag) | 16.69(2) |
cmoa | 1.50 |
cwise | 1.25 |
cogle | 1.57 |
t0 (HJD] | 245 5774.2729(54) |
u0 | 0.046 86(17) |
tE (d) | 23.17(80) |
s | 1.822(10) |
α (rad) | 0.3662(36) |
q | 0.0284(10) |
ρ | <0.007 |
fs/fb (OGLE) | 0.0573(29) |
Isource (mag) | 19.83(5) |
Iblend (mag) | 16.69(2) |
cmoa | 1.50 |
cwise | 1.25 |
cogle | 1.57 |
We note the existence of a small, ∼0.02 mag, ‘bump’ in the light curve, at ∼t0 + 50 d, with duration of about 20 d. Fig. 1(b) shows the 5-d-binned OGLE and MOA data over this period. The possibly similar behaviour in OGLE and MOA data lends some credence to the reality of this feature. On the other hand, its amplitude is similar to those of several other baseline fluctuations, and thus no definite conclusions can be drawn about it.
The trajectory of the source relative to the caustic structure is shown in Fig. 2. Since the source passes near the central caustic, we have checked for the so-called s ↔ s−1 degeneracy, that often occurs with anomalies dominated by the central caustic in the case of high-magnification events (Griest & Safizadeh 1998). However, the close solution is disfavoured by ▵χ2 = 85, and we adopt the wide solution.
In many anomalous microlensing events (e.g. Udalski et al. 2005; Gaudi et al. 2008; Muraki et al. 2011; Han et al. 2013; Kains et al. 2013), high-order effects, such as microlens parallax and finite-source effects, can break (or partially break) the degeneracies among the physical parameters that determine tE. However, since the event duration was short and with moderate magnification, the amplitude of the microlens parallax effect is small, and including it does not improve the fit significantly (Δχ2 = 7). Our modelling sets an upper limit on the Einstein-radius-normalized source radius of ρ < 0.007 (3σ level). The best-fitting model has ρ = 0.002, but the data are also consistent with a point source at the ∼1σ level (Δχ2 = 1.4). The probability distribution for the finite-source size, recovered from the MCMC chain, is shown in Fig. 3.
To try to set further constraints on the source and lens properties, we construct a colour–magnitude diagram (CMD) of objects within 90 arcsec of the event's position (Fig. 4), using OGLE-III (Udalski et al. 2008) calibrated V-band and I-band magnitudes of images, taken before the event (there were no V-band images during the event). We estimate the position of the ‘red clump’ at (V − I, I)cl = (1.94, 15.45) and compare it to the unreddened values derived by Nataf et al. (2013) for the Galactic coordinates of the event, (l,b) = (3.6,−2.8). Nataf et al. (2013) and Bensby et al. (2011) find (V − I, I)cl, 0 = (1.06, 14.36), i.e. a line-of-sight I-band extinction to the red clump of AI = 1.09, and a distance modulus of 14.48 ± 0.24 mag, or 7.9 ± 0.9 kpc.
In principle, some or all of the bright ‘blend light’ in this event could be due to the lens star itself. Its isolated position on the CMD, at (V − I, I)blend = (1.45, 16.7) (see Fig. 4), ∼1 mag brighter than the general track of the stars in this region of the diagram, suggests that it might be a foreground disc star that suffers from relatively less extinction. On the other hand, our Bayesian analysis, below, suggests such a nearby lens is highly unlikely, and therefore, the blend may be due to one or more unrelated stars seen in projection. In addition, based on long-term OGLE data, the proper motion of the blend star relative to the majority bulge stars in the field is very small, 2 ± 1 mas yr− 1, thus also suggesting it is located in the bulge. If so, it might be a horizontal branch star. Another possibility is that this blend is not the lens itself, but a companion to the lens.
4 PHYSICAL PARAMETERS – BAYESIAN ANALYSIS
The physical parameters of the lens and its companion are connected to three of the model parameters: tE – which involves the lens mass and distance, and q and s which, given the lens properties, give the companion mass and projected distance to the lens star, respectively. Given the limited constraints we can set, in this case, using high-order effects and the CMD, we use Bayesian analysis to estimate probabilistically the lens distance and mass, and the Einstein radius of the event, following previous analyses of microlensing events (e.g. Batista et al. 2011; Yee et al. 2012). The prior distribution includes a Galactic stellar structure model, which sets the rate equation for lensing events. As noted, due to the large amount of blended light, likely unrelated to the lens, we cannot place strong constraints on the lens mass from the observed flux.
We adopt the Galactic model of Han & Gould (1995, 2003), which reproduces well the observed statistical distribution of properties of microlensing events. The stellar density, n(x, y, z), includes a cylindrically symmetric disc, a bulge and a central bar (for specific model parameters see table 2 in Batista et al. 2011).
For the mass function, we follow Dominik (2006) and use different mass functions for the Galactic disc and bulge, consisting of power laws and a log-normal distributions in (M/M⊙) adopted from Chabrier (2003).
The posterior probability distributions for the lens mass and distance are found by marginalizing over all other parameters, and for the Einstein radius by summing the probability for the appropriate combination of DL, DS and M, marginalizing over μ. Fig. 5 shows the results of our Bayesian analysis. The inferred lens mass is |$M=0.39^{+0.45}_{-0.19}\,{\rm M}_{\odot }$|, and thus the companion is a planet with mass of |$11.6^{+13.4}_{-5.6}\ M_{\rm J}$|. The uncertainties are the 68 per cent probability range about the median of the probability distribution, which we take as the most likely value. The system's distance is 7.56 ± 0.91 kpc. We find that the Einstein radius of the lens star is |$2.38^{+0.81}_{-0.65}$|au, so the projected separation of the companion, r⊥ = s · rE, is |$4.3^{+1.5}_{-1.2}$|au, rather far beyond the location of the snowline at |$R_{\rm SL}=1.1^{+1.2}_{-0.6}$|au [assuming a relation |$R_{\rm SL}=2.7(M/{\rm M}_{\odot })$|au].
5 DISCUSSION
We have presented the detection, via microlensing, of a Jovian planet orbiting around a likely M-type star. There are three additional microlensing events that have detected Jupiters around M stars (Udalski et al. 2005; Batista et al. 2011; Poleski et al. 2013), and together they constitute ∼1/5 of the planetary systems detected to date through microlensing. In all of them, the planet is located beyond the snowline, at distances ≲5 au from the host star. However, within the uncertainties, for two out of the four events (Poleski et al. 2013 and this work), the primary could also be a K- or G-type star. If snowline-region massive planets around M stars are indeed common, this may be in conflict with the two leading planetary formation scenarios. According to the ‘core accretion’ scenario (e.g. Ida & Lin 2005), Jovian planets form beyond the snowlines of their parent stars, but massive planets around M-type stars should be rare (Laughlin, Bodenheimer & Adams 2004), since their formation times are longer than the typical disc lifetime. In the disc instability planet-formation scenario (e.g. Boss 2006) massive planets do form around M stars, but at distances ≳7 au.
Future high-resolution imaging could confirm our results by isolating the light from the lens. In addition, a set of such images, spread over several years after the event, could measure the relative proper motion between the lens and the source star, and set stronger constraint on the system parameters. Since this is a general problem for many of the planets detected by microlensing, a dedicated observing programme following all microlensing planets is essential for the interpretation of those planets, and could give better priors for future Bayesian analysis of such events.
Unlike most of the microlensing-detected planets to date, the planet presented here was not detected in real time, but in a post-season analysis, illustrating the essence and elegance of the second-generation survey principle. The other planetary events that were inside the collaboration footprint in the 2011 season could have also been fully characterized by the survey data alone. The final results of this controlled experiment can thus eventually help us determine the frequency of snowline planets in the Galaxy.
We thank J.C. Yee for stimulating discussions on Bayesian analysis and M. Albrow for kindly providing the pysis software, which was used for the DIA of the Wise survey data. The anonymous referee is thanked for comments that improved the presentation. This research was supported by the I-CORE programme of the Planning and Budgeting Committee and the Israel Science Foundation, Grant 1829/12. DM and AG acknowledge support by the US-Israel Binational Science Foundation. The OGLE project has received funding from the European Research Council under the European Community's Seventh Framework Programme (FP7/2007-2013)/ERC grant agreement no. 246678 to AU. TS acknowledges funding from JSPS 23340044 and JSPS 24253004.