OGLE-2014-BLG-0289: Precise Characterization of a Quintuple-Peak Gravitational Microlensing Event
A. Udalski, C. Han, V. Bozza, A. Gould, I.A. Bond, P. Mróz, J. Skowron, Ł. Wyrzykowski, M. K. Szymański, I. Soszyński, K. Ulaczyk, R. Poleski, P. Pietrukowicz, S. Kozłowski, F. Abe, R. Barry, D.P. Bennett, A. Bhattacharya, M. Donachie, P. Evans, A. Fukui, Y. Hirao, Y. Itow, K. Kawasaki, N. Koshimoto, M.C.A. Li, C.H. Ling, K. Masuda, Y. Matsubara, S. Miyazaki, H. Munakata, Y. Muraki, M. Nagakane, K. Ohnishi, C. Ranc, N. Rattenbury, T. Saito, A. Sharan, D.J. Sullivan, T. Sumi, D. Suzuki, P.J. Tristram, T. Yamada, A. Yonehara, E. Bachelet, D.M. Bramich, G. DÁgo, M. Dominik, R. Figuera Jaimes, K. Horne, M. Hundertmark, N. Kains, J. Menzies, R. Schmidt, C. Snodgrass, I.A. Steele, J. Wambsganss, R.W. Pogge, Y.K. Jung, I.-G. Shin, J. C. Yee, W.-T. Kim, C. Beichman, S. Carey, S. Calchi Novati, W. Zhu
aa r X i v : . [ a s t r o - ph . S R ] J a n D RAFT VERSION S EPTEMBER
11, 2018
Preprint typeset using L A TEX style AASTeX6 v. 1.0
OGLE-2014-BLG-0289: PRECISE CHARACTERIZATION OF A QUINTUPLE-PEAK GRAVITATIONALMICROLENSING EVENT
A. U
DALSKI , C. H AN , V. B OZZA , A. G
OULD , I. A. B
OND
AND
P. M
RÓZ , J. S KOWRON , Ł. W YRZYKOWSKI M. K. S
ZYMA ´NSKI , I. S OSZY ´NSKI , K. U LACZYK , R. P OLESKI ,P. P
IETRUKOWICZ , S. K OZŁOWSKI ,(T HE OGLE C
OLLABORATION )F. A BE , R. B ARRY , D. P. B ENNETT , A. B
HATTACHARYA , M. D ONACHIE , P. E VANS , A. F UKUI , Y. H IRAO , Y. I TOW ,K. K AWASAKI , N. K OSHIMOTO , M. C. A. L I , C. H. L ING , K. M ASUDA , Y. M ATSUBARA , S. M IYAZAKI , H. M UNAKATA ,Y. M URAKI , M. N AGAKANE , K. O HNISHI , C. R ANC , N. R ATTENBURY , T. S AITO , A. S HARAN , D. J. S ULLIVAN ,T. S UMI , D. S UZUKI , P. J. T RISTRAM , T. Y AMADA , A. Y ONEHARA ,(T HE MOA C
OLLABORATION )E. B
ACHELET , D. M. B RAMICH , G. DÁ GO , M. D OMINIK , R. F IGUERA J AIMES , K. H
ORNE , M. H UNDERTMARK ,N. K
AINS , J. M ENZIES , R. S CHMIDT , C. S NODGRASS , I. A. S
TEELE , J. W AMBSGANSS (R OBO N ET C OLLABORATION )R. W. P
OGGE , Y. K. J UNG , I.-G. S HIN , J. C. Y EE , W.-T. K IM ,(T HE µ FUN C
OLLABORATION )C. B
EICHMAN , S. C AREY , S. C ALCHI N OVATI , W. Z HU (T HE Spitzer T EAM ) Warsaw University Observatory, Al. Ujazdowskie 4, 00-478 Warszawa, Poland Department of Physics, Chungbuk National University, Cheongju 28644, Korea Dipartimento di Fisica "E. R. Caianiello", Universitá di Salerno, Via Giovanni Paolo II, I-84084 Fisciano (SA), Italy Istituto Nazionale di Fisica Nucleare, Sezione di Napoli, Via Cintia, I-80126 Napoli, Italy Korea Astronomy and Space Science Institute, Daejon 34055, Korea Department of Astronomy, Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA Institute of Natural and Mathematical Sciences, Massey University, Auckland 0745, New Zealand Institute for Space-Earth Environmental Research, Nagoya University, 464-8601 Nagoya, Japan Code 667, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA Deptartment of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 46556, USA Dept. of Physics, University of Auckland , Private Bag 92019, Auckland, New Zealand Okayama Astrophysical Observatory, National Astronomical Observatory of Japan, Asakuchi,719-0232 Okayama, Japan Department of Earth and Space Science, Graduate School of Science, Osaka University, Toyonaka, Osaka 560-0043, Japan Nagano National College of Technology, 381-8550 Nagano, Japan Tokyo Metroplitan College of Industrial Technology, 116-8523 Tokyo, Japan School of Chemical and Physical Sciences, Victoria University, Wellington, New Zealand Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Kanagawa 252-5210, Japan Mt. John University Observatory, P.O. Box 56, Lake Tekapo 8770, New Zealand Department of Physics, Faculty of Science, Kyoto Sangyo University, 603-8555 Kyoto, Japan Las Cumbres Observatory Global Telescope Network, 6740 Cortona Drive, Suite 102, Goleta, CA 93117, USA New York University Abu Dhabi, Saadiyat Island, Abu Dhabi, PO Box 129188, United Arab Emirates European Southern Observatory, Karl-Schwarzschild-Str. 2, 85748 Garching bei M unchen, Germany Istituto Nazionale di Fisica Nucleare, Sezione di Napoli, Napoli, Italy School of Physics & Astronomy, University of St Andrews, North Haugh, St Andrews KY 16 9SS, UK Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100, Kobenhavn, Denmark SUPA, School of Physics & Astronomy, University of St Andrews, North Haugh, St Andrews KY16 9SS, UK Space Telescope Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA South African Astronomical Observatory, PO Box 9, Observatory 7935, South Africa Astronomisches Rechen-Institut, Zentrum für Astronomie der Universit at Heidelberg (ZAH), 69120 Heidelberg, Germany Planetary and Space Sciences, Dept of Physical Sciences, The Open University, Milton Keynes, MK7 6AA, UK Max Planck Institute for Solar System Research, Justus-von-Liebig-Weg 3, 37077 Gotingen, Germany Astrophysics Research Institute Liverpool John Moores University, Liverpool L3 5RF, UK Harvard-Smithsonian Center for Astrophysics, 60 Garden St., Cambridge, MA, 02138, US Department of Physics & Astronomy, Seoul National University, Seoul 151-742, Korea NASA Exoplanet Science Institute, California Institute of Technology, Pasadena, CA 91125, USA H AN ET AL . Spitzer Science Center, MS 220-6, California Institute of Technology, Pasadena, CA, USA IPAC, Mail Code 100-22, California Institute of Technology, 1200 E. California Boulevard, Pasadena, CA 91125, USA Department of Astronomy, Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA Canadian Institute for Theoretical Astrophysics, 60 St George Street, University of Toronto, Toronto, ON M5S 3H8, Canada
OGLE Collaboration
The MOA Collaboration
The RoboNet collaboration
The µ FUN Collaboration
Corresponding author
ABSTRACTWe present the analysis of the binary-microlensing event OGLE-2014-BLG-0289. The event light curve ex-hibits very unusual five peaks where four peaks were produced by caustic crossings and the other peak wasproduced by a cusp approach. It is found that the quintuple-peak features of the light curve provide tight con-straints on the source trajectory, enabling us to precisely and accurately measure the microlensing parallax π E .Furthermore, the three resolved caustics allow us to measure the angular Einstein radius θ E . From the combi-nation of π E and θ E , the physical lens parameters are uniquely determined. It is found that the lens is a binarycomposed of two M dwarfs with masses M = 0 . ± . M ⊙ and M = 0 . ± . M ⊙ separated in projectionby a ⊥ = 6 . ± . D L = 3 . ± . Spitzer data is that the time of observation corresponds to theflat region of the light curve.
Keywords: gravitational lensing: micro – binaries: general INTRODUCTIONSince commencing in the early 1990s (Udalski et al. 1994;Alcock et al. 1995; Aubourg et al. 1995), massive surveyshave detected numerous microlensing events. The detectionrate of microlensing events, which was of order 10 yr − in theearly stage of the surveys, has greatly increased and currentlymore than 2000 events are annually detected.However, determinations of lens masses have been pos-sible for very limited cases. The difficulty of the lens massmeasurement arises because the event timescale, which is theonly measurable quantity related to the lens mass for generallensing events, is related to not only the mass M but also tothe relative lens-source proper motion µ and the lens-sourceparallax π rel , i.e. t E = √ κ M π rel µ ; π rel = au (cid:18) D L − D S (cid:19) , (1)where κ = 4 G / ( c au) and D L and D S represent the distancesto the lens and source, respectively. For the unique deter-mination of the lens mass, one needs to measure two addi-tional observables of the angular Einstein radius θ E and themicrolens parallax π E , i.e. M = θ E κπ E , (2)where θ = κ M π rel (Gould 2000).The angular Einstein radius can be measured by detect-ing light curve deviations caused by finite-source effects(Gould 1994; Nemiroff & Wickramasinghe 1994). For lens-ing events produced by single masses, finite-source effects [email protected] can be detected when a lens crosses the surface of a sourcestar (Pratt et al. 1996; Choi et al. 2012). However, the ratioof the angular source radius θ ∗ to the angular Einstein ra-dius is of order 10 − for a main-sequence source star and oforder 10 − even for a giant star. Therefore, the chance to de-tect finite-source effects for a single-lens event is very low.For events produced by binary objects, on the other hand,the probability of θ E measurement is relatively high becausebinary-lens events usually produce caustic-crossing featuresfrom which finite-source effects can be detected.One can measure the microlens parallax from the lightcurve deviation induced by the acceleration of the source mo-tion caused by the Earth’s orbital motion: ‘annual microlensparallax’ (Gould 1992). One can also measure the microlensparallax by simultaneously observing a lensing event fromground and from a satellite in a heliocentric orbit: ‘space-based microlens parallax’ (Refsdal 1966; Gould 1994). Con-sidering that the physical Einstein radius r E = D L θ E of typi-cal Galactic lensing events is of order a few au, for satelliteswith a projected Earth-satellite separation of order au, thelight curves seen from the Earth and the satellite usually ex-hibit considerable differences, e.g. OGLE-2015-BLG-0124(Udalski et al. 2015), OGLE-2015-BLG-0966 (Street et al.2016), OGLE-2015-BLG-1268, and OGLE-2015-BLG-0763(Zhu et al. 2016), and this enables a precise measurement of π E . In contrast, deviations in lensing light curves induced byannual microlens-parallax effects are in most cases very sub-tle due to the small positional change of the Earth during ∼ ( O )10 day durations of typical lensing events. Furthermore,parallax-induced deviations can often be confused with devi-ations caused by other higher-order effects such as the orbitalmotion of the lens (Batista et al. 2011; Skowron et al. 2011;Han et al. 2016b). As a result, measurements of annual mi-crolens parallaxes are in many cases subject to large uncer-GLE-2014-BLG-0289 3tainty both in precision and accuracy.It was pointed out by An & Gould (2001) that the chanceto determine the lens mass by measuring both π E and θ E ishigh for a subclass of binary lensing events with three well-measured peaks where two peaks are produced by causticcrossings and the other is produced by a cusp approach. Thisis because the individual peaks provide tight constraints onthe source trajectory, enabling one to measure the microlensparallax. Furthermore, the angular Einstein radius is mea-surable from the analysis of almost any well-resolved causticcrossing, making triple-peak events good candidates for lensmass measurements.In this paper, we present the analysis of the binary-lensevent OGLE-2014-BLG-0289. The light curve of the eventexhibits very unusual five peaks. Among these peaks, fourwere produced by caustic crossings and the other was pro-duced by a cusp approach. The angular Einstein radius isprecisely measured by detecting finite-source effects fromthe resolutions of 3 caustic crossings. Furthermore, the well-resolved multiple peaks enable us to measure the microlensparallax, leading to an accurate and precise measurement ofthe lens mass. OBSERVATION AND DATAThe source star of the microlensing event OGLE-2014-BLG-0289 is located toward the Galactic bulge field. Theequatorial coordinates of the source star are (RA , DEC)
J2000 =(17:53:51.66, -29:05:05.6), which correspond to Galacticcoordinates ( l , b ) = (0 . ◦ , − . ◦ ). The magnification ofthe source flux caused by lensing was found on 17 March2014 (HJD ′ = HJD − ∼ . Data from the OGLEand MOA surveys were acquired in the the standard Cousins I and the customized MOA R passband, respectively. Fig-ure 1 shows the light curve of the event.The source flux was already magnified before the 2014Bulge season started. Just one day before the event was iden-tified, i.e. HJD ′ ∼ Figure 1 . Light curve of OGLE-2014-BLG-0289. The upper panelsshow the zoom of the regions enclosed by boxes in the lower panel.The numbered arrows designate the peaks in the light curve. Thenumbered arrows indicate the locations of the five peaks. pattern of a binary lensing event until another spike appearedat HJD ′ ∼ ′ ∼ ∼ θ E . Real-time modeling wasimportant in preparing follow-up observations to resolve thecaustic crossing, which yields the angular Einstein radius. Italso helped to prepare space-based observations using Spitzer telescope, which was separated ∼ Spitzer observations because one can mea-sure the space-based microlens parallax and the measured π E combined with θ E leads to the measurement of the lensmass. Due to these considerations, real-time modeling wasconducted more frequently as the source approached closerto the caustic exit.The caustic exit occurred at HJD ′ ∼ µ FUN) and RoboNet. The µ FUN group observed the event using the 1.3 m SMARTtelescope at the CTIO Observatory in Chile. The RoboNetobservations were conducted with the 1 m robotic tele-scopes at South African Astronomical Observatory (SAAO)in South Africa and Siding Spring Observatory (SSO) inAustralia. µ FUN observations were conducted in standardCousins I band and several V -band images were obtained tomeasure the source color. RoboNet data were taken in SDSS- H AN ET AL .i band. From these follow-up observations, the second spikeproduced by the source star’s caustic exit was captured withsufficient resolution to determine the angular source size. Wenote that the caustic exit was also covered by both OGLE andMOA surveys. See the upper middle panel of Figure 1.The event continued after the caustic exit and so did real-time modeling. Modeling conducted several days after thecaustic exit revealed two important findings. First, it was pre-dicted that there would be another pair of caustic crossings.Second, it was found that considering the microlens-parallaxeffect is important for the precise description of the observedlight curve. With the progress of the event, the time of thenext caustic crossing was refined. The predicted time ofthe caustic crossing was informed to the microlensing com-munity and follow-up observations were prepared accord-ingly. The third and fourth caustic crossings occurred suc-cessively at HJD ′ ∼ I ∼ . Spitzer observations (Calchi Novati et al. 2015) were con-ducted during the period 6814 < HJD ′ < Spitzer data, however, itis found that there exists no noticeable lensing signal, i.e. novariation of the source brightness. We discuss the reason forthe absence of the
Spitzer lensing signal in Section 3.A very unusual characteristic of the event is that the lightcurve exhibits 5 peaks. The individual peaks occurred atHJD ′ ∼ µ FUN data were reducedwith the DANDIA pipeline (Bramich 2008) and the pySIS(Albrow et al. 2009), respectively. For the µ FUN CTIO data,photometry were additionally done with DoPHOT software(Schechter et al. 1993) in order for the source color measure-ment and color-magnitude diagram construction. We notethat the quality of the MOA data at the baseline is not good,but their coverage of the caustic crossings is important inmeasuring θ E . We, therefore, use MOA data taken when thesource was magnified.For the use of multiple data sets that are obtained with dif-ferent telescopes and detectors and processed with different photometry softwares, it is required to readjust the errorbarsof the data sets. For this readjustment, we follow the stan-dard procedure of Yee et al. (2012), where the error bars arerenormalized by σ = k ( σ + σ ) / , (3)where σ is the uncorrected error bar from the automatedpipelines. We set the factor σ min based on the scatter of data.The factor k is set so that χ per degree of freedom (dof)becomes unity, i.e. χ / dof = 1. We list the error-bar readjust-ment factors in Table 1 along with the number of data points, N data . Table 1 . Error bar readjustment factorsData set k σ min N data OGLE 1.808 0.002 3700MOA 1.252 0.003 808 µ FUN CTIO 1.122 0.005 59RoboNet SSO (Dome A) 0.692 0.025 61RoboNet SSO (Dome C) 0.824 0.005 32RoboNet SAAO (Dome A) 0.663 0.020 47RoboNet SAAO (Dome C) 0.583 0.020 823.
LIGHT CURVE MODELINGFrom the spike features, it is obvious that the event wasproduced by a lens composed of multiple components. We,therefore, start modeling of the observed light curve based onthe binary-lens interpretation. For the simplest case wherethe relative lens-source motion is rectilinear, one needs 7principal parameters in order to describe the light curve ofa binary-lens event. The first three parameters ( t , u , t E ) areneeded to describe the source approach to the lens and theyrepresent the time of the closest lens-source separation, theseparation at that time, and the event timescale, respectively.Another three parameters ( s , q , α ) are used to describe thebinary lens and they denote the binary separation, mass ra-tio between the lens components, and the angle between thesource trajectory and the line connecting the binary compo-nents, respectively. The caustic-crossing parts of a binary-lens event are affected by finite-source effects and the lastparameter ρ is used to describe the deviation.The light curve of the event exhibits caustic-crossing fea-tures and thus we consider finite-source effects. We computelensing magnifications affected by finite-source effects us-ing the inverse ray-shooting technique. In computing finite-source magnifications, we take the surface brightness vari-ation caused by limb-darkening into consideration. Thesurface-brightness profile is approximated by a linear model,i.e. Σ λ ∝ − Γ λ (cid:18) −
32 cos φ (cid:19) , (4)where λ denotes the observed passband, Γ λ is the linear limb-darkening coefficient, and φ represents the angle between theGLE-2014-BLG-0289 5 Figure 2 . Model light curves of the “standard” (blue curve) and“orbit” (red curve) solutions. The middle and lower panels showthe residuals from the individual models. line of sight and the normal to the surface of the source star.We determine the limb-darkening coefficients based on thesource star’s stellar type. It turns out that the source is anearly K-type main-sequence star. See Section 4 for the de-tailed procedure of the source type determination. Basedon the stellar type, we adopt the limb-darkening coefficientsfrom the Claret (2000) catalog. The adopted I - and V -bandcoefficients are Γ I = 0 .
485 and Γ V = 0 . R -band data, we use Γ MOA = ( Γ I + Γ R ) / . Γ R = 0 .
585 is the R -band coefficient.We search for the solution of the lensing parameters intwo steps. In the first step, we divide the lensing parametersinto two groups. We select ( s , q , α ) as grid parameters sincelensing magnifications can vary dramatically with the smallchange of these parameters. We choose the other parame-ters, i.e. ( t , u , t E , ρ ), as downhill parameters because lensingmagnifications vary smoothly with the changes of the param-eters. For the individual sets of the grid parameters, we thensearch for the set of the downhill parameters yielding the best χ using the Markov Chain Monte Carlo (MCMC) method.The total computation time for the grid search is ∼
24 hoursusing 176 CPUs. This initial search provides a χ map in the s - q - α parameter space, from which we identify local minima.We then refine each local minimum by allowing all parame-ters to vary. We note that the initial grid search is impor-tant to identify degenerate solutions where different combi-nations of lensing parameters result in similar lensing lightcurves. For the case of OGLE-2014-BLG-0289, we identifya unique solution and find no solution with χ comparable tothe best-fit solution.In Figure 2, we present the model light curve (blue curve inthe upper panel) of the solution obtained under the assump-tion of the rectilinear lens-source motion (“standard model”).The middle panel shows the residual from the model. For bet-ter visual comparison of the fit with data, we plot data points of only the OGLE, MOA, and µ FUN CTIO data sets. Thebinary-lens parameters estimated by the model are s ∼ . q ∼ .
9. Although the standard model basically describesthe overall light curve, it leaves considerable residuals. Themajor residuals occur near the first peak at HJD ′ ∼ ′ ∼ Table 2 . Comparison of ModelsModel χ Standard 17950.3Orbit 5011.8Parallax ( u >
0) 4847.2Parallax ( u <
0) 4867.0Parallax+Orbit ( u >
0) 4829.9Parallax+Orbit ( u <
0) 4848.0
Incorporating higher-order effects requires to include addi-tional lensing parameters. In order to consider the microlens-parallax effect, one needs 2 parameters π E , N and π E , E . Theydenote the north and east components π E that represents themicrolens-parallax vector projected onto the sky in the equa-torial coordinate systems. The direction of π E is the same asthat of the relative lens-source motion (Gould 2000, 2004).Under the first-order approximation that the projected binaryseparation s and the source trajectory angle α vary in constantrates, the lens-orbital effect is described by 2 parameters of ds / dt and d α/ dt (Albrow et al. 2000). With these parame-ters, we conduct additional modeling to check the improve-ment of the fit with the higher-order effects. In this modeling,we first separately consider the microlens-parallax (“paral-lax model”) and lens-orbital effects (“orbit model”) and sec-ond simultaneously consider both effects (“parallax + orbit”model).For events affected by microlens-parallax effects, theremay exist a pair of degenerate solutions with u > u <
0: ‘ecliptic degeneracy’ (Skowron et al. 2011). Thisdegeneracy arises because the source trajectories of the twodegenerate solutions are in the mirror symmetry with respectto the binary-lens axis. For the pair of the solutions subjectto this degeneracy, the lensing parameters are approximatelyrelated by ( u , α, π E , N , d α/ dt ) ↔ − ( u , α, π E , N , d α/ dt ). Wecheck this degeneracy whenever microlens-parallax effectsare considered in modeling. H AN ET AL . Figure 3 . The cumulative distributions of χ as a function of timefor the tested models. To better show the differences between mod-els considering higher-order effects, we present the zoom of the dis-tributions in the upper panel. We note that the χ difference betweenthe “parallax” and “parallax+orbit” models is so small that the twodistributions are difficult to be distinguished within the line width. We find that higher-order effects, particularly themicrolens-parallax effect, are important in explaining theresiduals from the standard model. In Table 2, we presentthe χ values of the tested models. In Figure 3, we alsopresent the cumulative distributions of χ as a function oftime for the individual models. From the comparison of mod-els, it is found that the fit improves by ∆ χ ∼ . ∆ χ = 164 . ∆ χ = 17 .
3) is minor.In Figure 2, we present the model light curve the orbitmodel (red curve in the upper panel) and the residual fromthe model (lower panel). It is found the model can describethe fifth peak, that could not be explained by the standardmodel, but it still cannot describe the first peak. To check thepossibility that the assumption of the constant change ratesof ds / dt and d α/ dt do not sufficiently describe lens-orbitaleffects, we conduct an additional modeling by fully consid-ering the Keplerian orbital motion of the lens. This mod-eling requires 2 more parameters of s k and ds k / dt . Theseparameters represent binary separation (in units of θ E ) alongthe line of sight and the rate of separation change, respec-tively (Skowron et al. 2011; Shin et al. 2011). This modelingresults in almost an identical χ to that of the linear orbital- motion solution. This confirms that the major cause of thedeviation is the microlens-parallax effect.In Table 3, we list the lensing parameters of the u > u < F s , and the blendedlight, F b , that are measured based on the OGLE data. Fromthe comparison of u > u < u > u < ∆ χ = 18 .
1. In Figure 4, we present the model light curveof the best-fit solution (“parallax+orbit” with u > Table 3 . Best-fit Lensing ParametersParameter Value u > u < χ t (HJD’) 6820 . ± .
327 6802 . ± . u . ± . − . ± . t E (days) 144 . ± .
24 162 . ± . s . ± .
01 1 . ± . q . ± .
01 1 . ± . α (rad) 2 . ± . − . ± . ρ (10 − ) 0 . ± .
01 0 . ± . π E , N . ± . − . ± . π E , E − . ± . − . ± . ds / dt (yr − ) − . ± . − . ± . d α / dt (yr − ) 0 . ± . − . ± . F s / F b ) OGLE
Figure 5 shows the lens-system geometry. In the geometry,we present the source trajectory (solid curve with an arrow)with respect to the caustic (cuspy closed curve) and the lenscomponents (marked by M and M ) for the best-fit solution,i.e. u > ′ ∼ Spitzer data is due to the fact that the space-based light curve duringthe
Spitzer observation accidentally corresponds to a regionwhere the light curve is very flat. In the middle panel of Fig-ure 4, we present the light curve expected to be observed inGLE-2014-BLG-0289 7
Figure 4 . Best-fit model light curve (black solid curve). Upperpanels show the model fits around the regions enclosed by boxesin the middle panel. The lower panel shows the residual from themodel. The blue curve in the middle panel represents the light curveexpected to be observed in space using
Spitzer telescope. The re-gion represented by a left-right arrow and marked by ‘
Spitzer obser-vation’ denotes the period during which
Spitzer observations wereconducted (2456814 < HJD < space using the Spitzer telescope (blue curve). We note thatthe
Spitzer light curve is constructed based on the microlensparallax parameters determined from the ground-based data.The region represented by a left-right arrow and marked by‘
Spitzer observation’ denotes the period during which
Spitzer observations were conducted. It shows that this region of thelight curve is very flat and thus there is no noticeable lens-ing signal in the
Spitzer data. In the lower panel of Figure 5,we present the source trajectory (dotted curve with an arrow)that is expected to be seen from the
Spitzer telescope. Theconsistency of the predicted model with the flat
Spitzer datafurther supports the correctness of the solution determinedfrom the ground-based data.If the source is a binary, the orbital motion of the sourcecan also induce long-term deviations in lensing light curves:“xallarap effect” (Poindexter et al. 2005; Rahvar & Dominik2009). We, therefore, check the xallarap possibility of thedeviation. Considering xallarap effects requires five param-eters in addition to the principal parameters. These includethe north and east components of the xallarap vector, ξ E , N and ξ E , E , the orbital period, the phase angle and inclination of theorbit. See the appendix of Han et al. (2016a) for details aboutthe xallarap parameters. We find a best-fit xallarap modelwith an orbital period P ∼ . ∆ χ = 118 .
3, which issignificant enough to exclude the xallarap interpretation. CHARACTERIZING THE LENS
Figure 5 . Lens-system geometry showing the source trajectory(solid curve with an arrow) with respect to the caustic (cuspy closedcurve) and the lens components (marked by M and M ). The upperpanel shows the enlargement of the lower left region of the caustic.The caustics at 4 different times are presented in different colors.The dotted curve represents the source trajectory seen in space fromthe Spitzer telescope. The thick line on the
Spitzer source trajec-tory represents the time during which the event was observed by the
Spitzer telescope.
Physical Lens Parameters
To uniquely determine the lens mass, it is needed to esti-mate the angular Einstein radius in addition to the microlensparallax. The angular Einstein radius is determined by θ E = θ ∗ ρ . (5)We measure the normalized source radius ρ from the analy-sis of the caustic-crossing parts of the light curve. We notethat the third peak was resolved with a sufficient coveragefor the ρ measurement. See the upper panels of Figure 4. Todetermine θ E , then, one needs to estimate the angular sourceradius θ ∗ .We determine the angular source radius based on thedereddened color ( V − I ) and brightness I of the sourcestar. For the color and brightness determinations, we usethe method of Yoo et al. (2004). In this method, ( V − I ) and I are determined from the offsets in color ∆ ( V − I ) andbrightness ∆ I with respect to the centroid of the red giantclump (RGC), for which the intrinsic color and brightness areknown. In Figure 6, we present the color-magnitude diagramof stars in the neighboring region around the source star.The color-magnitude diagram is constructed based on theDoPHOT photometry of the µ FUN CTIO data. It is alignedto the OGLE-III photometric system by shifting the clumpmagnitude according to the extinction, A I = 1 .
77, and the red-dening, E ( V − I ) = 1 .
46, toward the field based on the OGLE-III extinction map (Nataf et al. 2013). We mark the positions H
AN ET AL . Figure 6 . Source location with respect to the centroid of the red gi-ant clump (RGC) in the color-magnitude diagram. Also marked arethe position of the blend. The filled triangle and square dots denotethe lens positions under the assumptions of no and full extinction,respectively. of the RGC centroid and the source by a red and blue dots.From the offsets in color ∆ ( V − I ) = − .
39 and magnitude ∆ I = 4 .
70 and the known dereddened values of the RGC,( V − I , I ) RGC , = (1 . , .
41) (Bensby et al. 2013; Nataf et al.2013), we find that the dereddeded color and brightness ofthe source star are ( V − I , I ) = ( V − I , I ) RGC + [ ∆ ( V − I ) , ∆ I ] =(0 . ± . , . ± . V − I ) into ( V − K ) using the V − I / V − K relation of Bessell & Brett(1988) and employ the color/surface brightness relation ofKervella et al. (2004) to find θ ∗ = 0 . ± . µ as. We esti-mate that the angular Einstein radius is θ E = 1 . ± .
09 mas . (6)Here we adopt the source distance that is estimated usingthe relation D S = D GC / (cos l + sin l / tan φ ) (Nataf et al. 2013),where D GC = 8160 pc is the galactocentric distance, l is thegalactic longitude, and φ ∼ ◦ is the angle between thesemimajor axis of the bulge and the line of sight. With l = 0 . ◦ , the adopted source distance is D S = 8011 pc. Incombination of the event timescale, the measured angularEinstein radius yields the relative lens-source proper motionof µ = θ E t E = 2 . ± .
21 mas yr − . (7)With both measured π E and θ E , the total mass M = M + M is determined using the relation in Equation (2) and themasses of the individual components are determined by M = M + q ; M = qM + q . (8) The distance to the lens is determined by the relation D L = au π E θ E + π S , (9)where π S = au / D S represents the parallax of the source star.The projected separation between the lens components is de-termined by a ⊥ = sD L θ E .In Table 4, we list the physical parameters of the lens. Wefind that the lens is a binary composed of two M dwarfs withmasses M = 0 . ± . M ⊙ (10)and M = 0 . ± . M ⊙ . (11)The estimated distance to the lens is D L = 3 . ± . . (12)The projected separation between the lens components is a ⊥ = 6 . ± . . (13)Also presented in Table 4 is the ratio of the transverse kinetic-to-potential energy ratio (KE/PE) ⊥ . The ratio is computedfrom the measured lensing parameters by (cid:18) KEPE (cid:19) ⊥ = ( a ⊥ / au) π ( M / M ⊙ ) "(cid:18) s dsdt (cid:19) + (cid:18) d α dt (cid:19) . (14)The ratio should be less than unity to be a bound system, i.e.(KE / PE) ⊥ ≤ KE / PE < .
0. It is found that the determinedvalue (KE / PE) ⊥ = 0 .
03 meets this requirement. Due to thesmall lens-orbital effect, the ratio is small, probably due tothe alignment of the lens components along the line of sight.
Table 4 . Physical lens parametersParameter ValuePrimary mass 0 . ± . M ⊙ Companion mass 0 . ± . M ⊙ Projected separation 6 . ± . . ± . ⊥ Is the blend the lens?
In Figure 6, we mark the location of the blend in the color-magnitude diagram. Then, a question is whether the blendis the lens itself. The intrinsic color corresponding to themass of the primary lens, ∼ . M ⊙ , is ( V − I ) L1 , ∼ . V − I ) L1 ∼ ( V − I ) L1 , ∼ . I L1 = M I , + D L − ∼ .
6. Here M I , ∼ . V − I , L ) L ∼ (2 . , . Figure 7 . ∆ χ distributions of MCMC chains obtained from mod-eling runs based on different data sets. The distribution in the up-per panel is obtained based on all data. The distribution in themiddle panel is based on the data where data points in the region6890 < HJD ′ < < HJD ′ < ∆ χ < other hand, it would have experienced the same amount ofextinction A I ∼ . E ( V − I ) ∼ . D L ∼ . the blend is not the lens .This line of reasoning is supported by the astrometric offset ∆ θ ∼ .
14” between the source position (measured from thedifference image near the peak of the event) and the OGLEcatalog position (which is dominated by the blend because itis ∼ DISCUSSIONThe event OGLE-2014-BLG-0289 is very unusual in thesense that its light curve exhibits 5 peaks among which 2were partially covered and the others are densely resolved.In this section, we demonstrate that the quintuple peaks helpto determine the microlens parallax with improved accuracyand precision.For this demonstration, we conduct additional modelingruns with data sets where parts of the data points are ex-cluded. In the first run, we exclude data points in the re-gion 6890 < HJD ′ < < HJD ′ < ∆ χ distributions of MCMCchains in the π E , E – π E , N parameter space obtained from themodeling runs with three different data sets. Dots markedin different colors represent chains with ∆ χ < ∆ ( π E , N , π E , E ) ∼ (0 . , .
01) for the ‘case 1’ and ∼ (0 . , .
02) for the ‘case2’. This indicates that the coverage of the peaks affects theaccuracy of the π E determination. Furthermore, the uncer-tainties of the determined microlens-parallax parameters in-crease as fewer caustics are resolved, suggesting that the peakcoverage also affects the precision of the π E determination.These results demonstrate that the resolution of the individualpeaks provides important constraints on the determinationsof the lens parameters.We note that there was a discovery of an additional lensingevent with quintuple peaks. The event, Gaia16aye, showeda complex light curve with 4 caustic crossings and a fifthbrightening likely due to a cusp approach (Mróz et al. 2016;Wyrzykowski et al. 2017). CONCLUSIONWe analyzed the binary-microlensing event OGLE-2014-BLG-0289. The light curve of the event exhibited very un-usual five peaks where four peaks were produced by causticcrossings and the other peak was produced by a cusp ap-proach. We found that the quintuple-peak features of thelight curve enabled us to precisely and accurately measurethe microlensing parallax π E . The three resolved caustics al-lowed us to precisely measure the angular Einstein radius θ E .From the combination of π E and θ E , the physical parame-ters of the lens were uniquely determined. We found that thelens was a binary composed of two M dwarfs with masses M = 0 . ± . M ⊙ and M = 0 . ± . M ⊙ separated inprojection by a ⊥ = 6 . ± . D L = 3 . ± . AN ET AL .Work by C. Han was supported by the grant(2017R1A4A1015178) of National Research Founda-tion of Korea. The OGLE project has received fundingfrom the National Science Centre, Poland, grant MAESTRO2014/14/A/ST9/00121 to A. Udalski. OGLE Team thanksProfs. M. Kubiak and G. Pietrzy´nski for their contributionto the OGLE photometric data set presented in this paper.The MOA project is supported by JSPS KAKENHI GrantNumber JSPS24253004, JSPS26247023, JSPS23340064,JSPS15H00781, and JP16H06287. Work by A. Gould wassupported by JPL grant 1500811. Work by J. C. Yee wasperformed in part under contract with the California Institute of Technology (Caltech)/Jet Propulsion Laboratory (JPL)funded by NASA through the Sagan Fellowship Programexecuted by the NASA Exoplanet Science Institute. Weacknowledge the high-speed internet service (KREONET)provided by Korea Institute of Science and TechnologyInformation (KISTI).
Software: