A terrestrial-mass rogue planet candidate detected in the shortest-timescale microlensing event
P. Mroz, R. Poleski, A. Gould, A. Udalski, T. Sumi, M.K. Szymanski, I. Soszynski, P. Pietrukowicz, S. Kozlowski, J. Skowron, K. Ulaczyk, M.D. Albrow, S.-J. Chung, C. Han, K.-H. Hwang, Y.K. Jung, H.-W. Kim, Y.-H. Ryu, I.-G. Shin, Y. Shvartzvald, J.C. Yee, W. Zang, S.-M. Cha, D.-J. Kim, S.-L. Kim, C.-U. Lee, D.-J. Lee, Y. Lee, B.-G. Park, R.W. Pogge
DDraft version October 21, 2020
Typeset using L A TEX twocolumn style in AASTeX62
A terrestrial-mass rogue planet candidate detected in the shortest-timescale microlensing event
Przemek Mr´oz,
1, 2
Rados(cid:32)law Poleski, Andrew Gould,
3, 4
Andrzej Udalski, and Takahiro Sumi —Micha(cid:32)l K. Szyma´nski, Igor Soszy´nski, Pawe(cid:32)l Pietrukowicz, Szymon Koz(cid:32)lowski, Jan Skowron, andKrzysztof Ulaczyk
6, 2 (OGLE Collaboration)Michael D. Albrow, Sun-Ju Chung,
8, 9
Cheongho Han, Kyu-Ha Hwang, Youn Kil Jung, Hyoun-Woo Kim, Yoon-Hyun Ryu, In-Gu Shin, Yossi Shvartzvald, Jennifer C. Yee, Weicheng Zang, Sang-Mok Cha,
8, 14
Dong-Jin Kim, Seung-Lee Kim,
8, 9
Chung-Uk Lee, Dong-Joo Lee, Yongseok Lee,
8, 14
Byeong-Gon Park,
8, 9 andRichard W. Pogge (KMT Collaboration) Division of Physics, Mathematics, and Astronomy, California Institute of Technology, Pasadena, CA 91125, USA Astronomical Observatory, University of Warsaw, Al. Ujazdowskie 4, 00-478 Warszawa, Poland Max Planck Institute for Astronomy, K¨onigstuhl 17, D-69117 Heidelberg, Germany Department of Astronomy, Ohio State University, 140 W. 18th Ave., Columbus, OH 43210, USA Department of Earth and Space Science, Graduate School of Science, Osaka University, Toyonaka, Osaka 560-0043, Japan Department of Physics, University of Warwick, Coventry CV4 7 AL, UK University of Canterbury, Department of Physics and Astronomy, Private Bag 4800, Christchurch 8020, New Zealand Korea Astronomy and Space Science Institute, Daejon 34055, Republic of Korea University of Science and Technology, Korea (UST) Gajeong-ro, Yuseong-gu, Daejeon 34113, Republic of Korea Department of Physics, Chungbuk National University, Cheongju 28644, Republic of Korea Department of Particle Physics and Astrophysics, Weizmann Institute of Science, Rehovot 76100, Israel Center for Astrophysics | Harvard & Smithsonian, 60 Garden St., Cambridge, MA 02138, USA Department of Astronomy and Tsinghua Centre for Astrophysics, Tsinghua University, Beijing 100084, China School of Space Research, Kyung Hee University, Yongin, Kyeonggi 17104, Republic of Korea (Received January 1, 2018; Revised January 1, 2018; Accepted January 1, 2018)
ABSTRACTSome low-mass planets are expected to be ejected from their parent planetary systems during earlystages of planetary system formation. According to planet formation theories, such as the core accretiontheory, typical masses of ejected planets should be between 0 . . M ⊕ . Although in practicesuch objects do not emit any light, they may be detected using gravitational microlensing via theirlight-bending gravity. Microlensing events due to terrestrial-mass rogue planets are expected to haveextremely small angular Einstein radii ( (cid:46) µ as) and extremely short timescales ( (cid:46) . t E ≈ . . θ E = 0 . ± . µ as,making the event the most extreme short-timescale microlens discovered to date. Depending on itsunknown distance, the lens may be a Mars- to Earth-mass object, with the former possibility favoredby the Gaia proper motion measurement of the source. The planet may be orbiting a star but we ruleout the presence of stellar companions up to the projected distance of ∼ . Corresponding author: Przemek Mr´[email protected] a r X i v : . [ a s t r o - ph . E P ] O c t Mr´oz et al.
Keywords:
Gravitational microlensing (672); Gravitational microlensing exoplanet detection (2147);Finite-source photometric effect (2142); Free floating planets (549) INTRODUCTIONThousands of extrasolar planets have been discoveredto date. Although many of the known exoplanets do notresemble those in our solar system, they have one thingin common—they all orbit a star. However, theories ofplanet formation and evolution predict the existence offree-floating (rogue) planets, gravitationally unattachedto any star.Exoplanets may be ejected from their parent planetarysystems as a result of planet–planet scattering (Rasio &Ford 1996; Weidenschilling & Marzari 1996; Lin & Ida1997; Chatterjee et al. 2008). It is estimated that atleast 75% of systems with giant planets must have expe-rienced planet–planet scattering in the past (Raymond& Morbidelli 2020, and references therein). Dynamicalinteractions between giant planets inevitably lead to dis-ruptions of orbits of inner smaller (rocky) planets (e.g.,Veras & Armitage 2005; Matsumura et al. 2013; Car-rera et al. 2016) and may lead to their ejection. In theirpopulation synthesis calculations (which are based onthe core accretion theory of planet formation; Ida et al.2013), Ma et al. (2016) found that typical masses ofejected planets are between 0 . . M ⊕ . Accordingto their model, rogue planets are more likely to formaround massive stars, which are in turn more likely tohost giant planets. A similar conclusion was reached byBarclay et al. (2017) who carried out N -body simula-tions of terrestrial planet formation around solar-typestars. They found that in the presence of giant plan-ets in such systems, a large fraction of the protoplane-tary material is ejected, partly in the form of Mars-massbodies ( ∼ . − . M ⊕ ). Planets may also be liberatedas a result of interactions in multiple-star systems (e.g.,Kaib et al. 2013) and stellar clusters (e.g., Spurzem et al.2009), stellar flybys (e.g., Malmberg et al. 2011), or thepost-main-sequence evolution of the host star (e.g., Ve-ras et al. 2011).Dark compact objects, such as rogue planets, maybe in principle detected in gravitational microlensingevents—microlensing does not depend on the brightnessof a lensing object. However, typical Einstein timescalesof microlensing events due to sub-Earth-mass objects areextremely short: t E = θ E µ rel = 1 . (cid:18) M . M ⊕ (cid:19) / (cid:16) π rel . (cid:17) / (cid:18) µ rel − (cid:19) − (1) rendering their detection difficult. (Here, θ E is the an-gular Einstein radius, µ rel – relative lens-source propermotion, M – mass of the lens, and π rel – relative lens-source parallax.) If the radius of the source star is largerthan the Einstein radius, the duration of microlensingevents is extended thanks to finite-source effects (Gould1994; Nemiroff & Wickramasinghe 1994; Witt & Mao1994). For sub-Earth-mass lenses, finite-source effectsbecome important if the angular radius of the source, θ ∗ , is of the order of the Einstein radius, θ E = 0 . µ as (cid:18) M . M ⊕ (cid:19) / (cid:16) π rel . (cid:17) / . (2)So far, only four short-timescale microlensing events ex-hibiting finite-source effects were identified (i.e., OGLE-2012-BLG-1323, t E = 0 . ± .
005 day, θ E = 2 . ± . µ as; OGLE-2016-BLG-1540, t E = 0 . ± .
003 day, θ E = 9 . ± . µ as; OGLE-2019-BLG-0551, t E = 0 . ± .
017 day, θ E = 4 . ± . µ as; KMT-2019-BLG-2073, t E = 0 . ± .
026 day, θ E = 4 . ± . µ as; Mr´oz et al.2018, 2019, 2020; Kim et al. 2020). These events may becaused by unbound or wide-orbit ( (cid:38)
10 au) planets sincemicrolensing observations alone are not able to rule outthe presence of a distant stellar companion (as discussedin more detail by Mr´oz et al. 2020). These detections, to-gether with short-timescale events found by Mr´oz et al.(2017), provide strong evidence for a large population offree-floating or wide-orbit planets in the Milky Way.In this Letter, we present the discovery of the shortest-timescale microlensing event detected to date ( t E =0 . +0 . − . day, θ E = 0 . ± . µ as), which waslikely caused by a Mars- to Earth-mass object. DATAMicrolensing event OGLE-2016-BLG-1928 occurredon 2016 June 18 (HJD (cid:48) = HJD − . I = 17 . V − I = 1 .
91) locatedat the equatorial coordinates of RA = 18 h m . s − ◦ (cid:48) . (cid:48)(cid:48) l, b ) =(1 . ◦ , − . ◦ )). The event was found in data fromthe fourth phase of the Optical Gravitational LensingExperiment (OGLE; Udalski et al. 2015) as part of thesearch for wide-separation planetary systems (Poleski etal., in preparation) but it has been observed by OGLE This event was not detected in real-time by the OGLE EarlyWarning System (Udalski 2003). For consistency with previousworks, we assigned it the name OGLE-2016-BLG-1928. terrestrial-mass planet detected in the shortest-timescale microlensing event . . . . . . . . . M ag n i t ud e OGLEKMT CTIOKMT SAAO t E = 0 . ± .
003 d ρ = 3 . ± . . . . . − . − . . . . R e s i du a l Figure 1.
Upper panel: 23 yr long OGLE light curve ofthe microlensing event OGLE-2016-BLG-1928 reveals onlyone brightening that occurred on 2016 June 18 and lastedabout 0.2 days. Lower panel: close-up of the magnified partof the light curve with the best-fitting microlensing modeloverplotted. since 1997. The event was located near the area that wasextensively monitored during the Campaign 9 of the K2 mission (Henderson et al. 2016) – both by K . < HJD (cid:48) < .
17 by the Mi-crolensing Observations in Astrophysics (MOA) survey(Bond et al. 2001) under adverse weather conditions(poor seeing and clouds), which prevented us from ex-tracting useful data. All analyzed data were collectedin the I -band. The photometry was extracted usingcustom implementations of the difference image analy- sis (Alard & Lupton 1998) method by Wo´zniak (2000)(OGLE) or Albrow et al. (2009) (KMTNet). SINGLE-LENS MODELSThe light curve of the event (Figure 1) can be well de-scribed by an extended-source point-lens model, whichhas four parameters: t and u – time and projectedseparation during the closest approach between the lensand the center of the source, t E – Einstein timescale,and ρ = θ ∗ /θ E – which is the angular radius of thesource θ ∗ expressed in θ E units. The approximate val-ues of these parameters can be estimated from thelight curve without the need of sophisticated model-ing (cf. Mr´oz et al. 2020): the maximum magnifica-tion A and duration ∆ t of the event are related to ρ ≈ (cid:112) / ( A − ≈ . t E ≈ ∆ t/ ρ = 0 .
03 days. Indeed,in our best-fitting model we measure ρ = 3 . +0 . − . and t E = 0 . +0 . − . days (Table 1). Microlensing magni-fications are computed using the method described byBozza et al. (2018), we assume a linear limb-darkeninglaw with Γ = 0 .
46 (as appropriate for the effective tem-perature of the source of 5000 ±
200 K, see Section 5;Claret & Bloemen 2011). The best-fitting parametersand their uncertainties are estimated using the MarkovChain Monte Carlo sampler of Foreman-Mackey et al.(2013).During the modeling we fix the value of the dimen-sionless blending parameter f s = 1, that is, we assumethat the entire flux comes from the source star. Usually,for every data set, there are two additional parametersthat describe the source flux F s and unmagnified blendflux F b . We define f s = F s / ( F s + F b ). When both F s and F b were allowed to vary, the best-fitting solutionshad large negative blending flux ( f s (cid:38) f s = 1is disfavored by only ∆ χ = 4 . θ E provided thatthe blend and source have similar colors. Thus, in ourfinal models, we kept F b = 0 (that is, f s = 1) constant,but we also added in quadrature 0.05 mag to the un-certainty of the source brightness. For comparison, thebest-fit parameters for the free-blending fit (assuming f s ≤
1) are also presented in Table 1. Blending may af-fect the characterization of the lens only if the blend andsource have significantly different colors, as discussed indetail by Mr´oz et al. (2020) and Kim et al. (2020). BINARY-LENS MODELSThe light curve of OGLE-2016-BLG-1928 shows aclear signal from a low-mass planet, but it does not show
Mr´oz et al.
Table 1.
Microlensing parameters for best-fit solutionsPoint Lens Binary LensParameter Value Value Parameter Value( f s = 1) ( f s ≤ t (HJD (cid:48) ) 7557 . +0 . − . . ± . t (HJD (cid:48) ) 7593 . ± . t E (days) 0 . +0 . − . . ± . t E (days) 1 . +0 . − . u . +0 . − . . +0 . − . u . +0 . − . ρ . +0 . − . . +0 . − . ρ . ± . I s . ± .
05 17 . +0 . − . s . +1 . − . f s . . +0 . − . q . +0 . − . × − α (deg) 188 . ± . Note —HJD (cid:48) =HJD-2450000. f s = F s / ( F s + F b ) is the dimensionless blending param-eter. an obvious signal from a host of the planet. To search fora host, we fitted a binary-lens model to the data. Thebinary-lens model has three parameters more than thesingle-lens model and these are: s – the projected sepa-ration between the planet and host expressed in Einsteinradii (of the total mass of the system), q – planet to hostmass ratio, and α – angle between the binary axis andthe source trajectory. We started the search for binary-lens models by defining t , u , and t E relative to theplanet (Han 2006; Mr´oz et al. 2020), because then thevalues of the four parameters ( t , u , t E , ρ ) are well con-strained by the light curve. In order to speed-up calcu-lations, we neglected limb-darkening of the source. Themagnification of the finite-source binary-lens model wasevaluated using the method presented by Bozza (2010)and Bozza et al. (2018). The fitting was done using theMulensModel code by Poleski & Yee (2019). After theseinitial fits converged, we reran the fits in standard pa-rameterization ( t , u defined relative to the center ofmass, and t E relative to the total mass of the system)and including limb-darkening of the source.The parameters of the best-fit binary-lens model arepresented in Table 1. When compared to the single-lensmodel, the χ improves by 44.2. The main differencebetween single-lens and binary-lens models is the pres-ence of a low-amplitude bump at HJD (cid:48) ≈ χ improvement, however, comes mostly from one observa-tory (OGLE) from one night and the KMT data fromthat night do not corroborate the signal (Figure 2). Thebump could have an origin other than the microlensingof planet host: it could be produced by low-level fluctu-ations in the light curve (either of instrumental origin,intrinsic variability, or a combination of both). We can-not judge the reliability of the binary-lens fit using theBayesian approach because we cannot present a mean- . . . . . M ag n i t ud e singlebinaryOGLEKMT CTIOKMT SAAO C u m u l a t i v e ∆ χ ∆ χ = χ (single) − χ (binary) Figure 2.
Upper panel: comparison between the single-(dashed line) and binary-lens (solid line) models. Lowerpanel: cumulative distribution of ∆ χ between these models. ingful prior on such a binary-lens model. Instead, wedecided to check ∆ χ relative to a constant brightnessmodel and determine whether similar bumps are presentin other seasons of the OGLE-IV data. We fitted point-source point-lens models to each of the observing seasons2010–2015 and 2017–2019. For each season ( i ), we cal-culated χ difference between the above fit and a modelof constant brightness (∆ χ i ). In order to compare thesevalues with the bump in 2016 we normalized them bythe number of epochs in the given season: ∆ χ i N /N i ,where N = 1621. The results of these calculations terrestrial-mass planet detected in the shortest-timescale microlensing event χ between point-lensand binary-lens models for 2016 data is 32.5 when onlyOGLE data are considered. We see that in four outof nine other seasons the ∆ χ i N /N i is higher than32.5, hence, the probability that the bump detected inbinary-lens analysis is just the manifestation of low-levelfluctuations as seen in other seasons is 4 / . s ≈
19 is almost four times largerthan the widest separation microlensing planet currentlyknown (OGLE-2008-BLG-092, s = 5 .
3; Poleski et al.2014).Since there is no strong evidence for a host star fromthe microlensing light curve, we use the method of Mr´ozet al. (2018, 2019) to estimate lower limits on the pro-jected star–planet separation of a putative host star. Weconsider a 0 . M (cid:12) host located either in the Galacticdisk ( π rel = 0 . π rel = 0 .
016 mas),which correspond to θ E , host = 0 .
49 mas or 0.20 mas, re-spectively. Then, we simulate synthetic OGLE lightcurves (spanning from 2010 March 5 through 2019 Oc-tober 30) assuming q = ( θ E /θ E , host ) , 1 ≤ s ≤
10, and0 ≤ α ≤ π . For each pair of ( q, s ) we calculate thefraction of light curves that show signatures of the hoststar. We find a 90% lower limit on the projected hostseparation of 3.6 Einstein radii for the lens located inthe disk ( π rel = 0 . π rel = 0 .
016 mas). These limits translate to8.0 au and 4.6 au, respectively.We also searched for binary-lens solutions in which theobserved brightening is due to a cusp crossing. We dida grid search with − ≤ log q ≤ . ≤ s ≤
3. Weconsidered only trajectories that cross the caustic twiceduring the night of HJD (cid:48) = 7557 and parameterized themodels using the Cassan (2008) approach. In the best-fitting model ( t E = 0 .
020 day, ρ = 2 . s = 4 . q = 0 . χ = 10 . t E < .
05 days forall binary-lens models in the grid search. PHYSICAL PARAMETERSThe light curve of the event OGLE-2016-BLG-1928(Figure 1) exhibits prominent finite-source effects thatenable us to measure the angular Einstein radius of thelens provided that the angular radius of the source star is
Table 2.
Statistics of trends seen in OGLE data in seasons2010–2015 and 2017–2019.year N i ∆ χ i N /N i Table 3.
Physical parameters for the sourceand lens for point lens ( f s = 1 . I s , . ± . V − I ) s , . ± . V − K ) s , . ± . T eff (K) 5000 ± I band) 0.46 θ ∗ ( µ as) 2 . ± . µ l (mas yr − ) − . ± . µ b (mas yr − ) − . ± . θ E ( µ as) 0 . ± . µ rel (mas yr − ) 10 . ± . known: θ E = θ ∗ /ρ . We use the color–surface brightnessrelation of Pietrzy´nski et al. (2019) to calculate θ ∗ . Todetermine the dereddened color ( V − I ) s , and brightness I s , of the source, we use the standard method of Yooet al. (2004). We measure that the source is ∆( V − I ) = − . ± .
02 bluer and ∆ I = 1 . ± .
09 fainter thanthe red clump centroid in the color–magnitude diagram(Figure 3). Because the dereddened color (( V − I ) RC , =1 .
06) and brightness ( I RC , = 14 .
38) of red clump starsin this direction are known (Bensby et al. 2011; Natafet al. 2013), we measure ( V − I ) s , = 0 . ± .
02 and I s , = 15 . ± .
09. Subsequently, we determine ( V − K ) s , = 2 . ± .
07 using the color–color relations ofBessell & Brett (1988) and θ ∗ = 2 . ± . µ as usingthe color–surface brightness relation of Pietrzy´nski et al.(2019).The calculation above is based on two assumptions.First, we assume that the source star and red clump Mr´oz et al. V − I I OGLE-2016-BLG-1928 red clumpsource
Figure 3.
Color–magnitude diagram of stars located within2 (cid:48) × (cid:48) of the microlensing event OGLE-2016-BLG-1928. stars are reddened by the same amount. Because the Gaia proper motion of the source relative to the meanproper motion of red clump stars is only 0 .
18 mas yr − (Gaia Collaboration et al. 2018) (Figure 4), the sourceis likely located in the Galactic bulge and so the first as-sumption holds true. Second, we assume that the colorof the source is equal to that of the baseline object. Nei-ther OGLE nor KMTNet observed the magnified partof the event in the V -band filter, which prevents us fromdirectly calculating the color of the source. Because thebest-fitting model evinces no evidence for blended lightfrom unresolved ambient stars, our best estimate of thecolor of the source is the color of the baseline object.Having the angular radius of the source star measured,we can calculate the angular Einstein radius: θ E = θ ∗ ρ = 0 . ± . µ as (3)and the relative lens-source proper motion (in the geo-centric frame): µ rel = θ E t E = 10 . ± . − . (4) DISCUSSIONWith the Einstein timescale of t E = 0 . +0 . − . d =41 . +3 . − . min and the angular Einstein radius of θ E =0 . ± . µ as, OGLE-2016-BLG-1928 is the most ex-treme short-timescale microlensing event discovered to − − − − µ l (mas yr − ) − − − µ b ( m a s y r − ) main sequencegiantssource Figure 4.
Gaia proper motions of stars located within5 (cid:48) of the event (Gaia Collaboration et al. 2018). Red gi-ant stars (representing the Galactic bulge population) aremarked by red dots, while main-sequence stars (Galactic diskpopulation) are marked by blue dots. Solid contours mark(1 , , σ error ellipses based on the scatter in the distribu-tion. The proper motion of the source and its position onthe color–magnitude diagram are consistent with those ofGalactic bulge stars. The relative lens-source proper motionis µ rel = 10 . ± . − , so the lens should be locatedon a dashed circle. date. The mass of the lens cannot be determined be-cause the relative lens-source parallax cannot be mea-sured: M = θ κπ rel , (5)where κ = 8 .
144 mas M (cid:12)− . If the lens is located in theGalactic disk ( π rel ≈ . M ≈ . M ⊕ (whichis approximately three Mars masses). The lens locatedin the Galactic bulge (typically π rel ≈ .
016 mas) wouldbe more massive ( M ≈ M ⊕ ).The Gaia proper motion of the source (Table 3) favorsthe interpretation that the lens is located in the Galacticdisk, so the lens should be a sub-Earth-mass object. Theproper motion of the source is consistent with that ofred clump stars (Figure 4), and the relative lens-sourceproper motion is µ rel = 10 . ± . − . In order tohave this relative proper motion, the lens should lie inthe region around a dashed circle marked in Figure 4 andthere are virtually no Galactic bulge stars in this region(while there exist some disk stars). To quantify this, we terrestrial-mass planet detected in the shortest-timescale microlensing event Gaia data (Figure 4). Thisdistribution can be approximated as a Gaussian withdispersions of 2 . ± .
052 and 2 . ± .
049 mas yr − inthe l and b directions, respectively. Thus the probabilitythat the proper motion of the lens is consistent with thatof bulge stars is smaller than 2 × − .The lens in OGLE-2016-BLG-1928 is likely a sub-Earth-mass object, one of the lowest-mass objects everfound by microlensing. As in the case of other short-timescale microlensing events (Sumi et al. 2011; Mr´ozet al. 2018, 2019, 2020; Kim et al. 2020), we cannotrule out the presence of a distant stellar companion.We conducted an extensive search for possible binary-lens models – we found that the best-fitting binary-lensmodel is preferred by ∆ χ = 44 . π rel = 0 . Nancy GraceRoman Space Telescope , formerly known as
WFIRST ,Johnson et al. 2020). Despite the fact that the eventwas located in high-cadence survey fields, only 15 datapoints were magnified (11 from OGLE and 4 from KMT-Net), rendering the event difficult to detect. In partic-ular, the declining part of the light curve is not fullycovered with observations (Figure 1).This raises the question of whether the observed lightcurve is due to a genuine microlensing event in the firstplace. The source star is located in the red giant branchin the color–magnitude diagram (Figure 3), and somegiants are known to produce stellar flares (e.g., VanDoorsselaere et al. 2017; Iwanek et al. 2019). However,the properties of the event (its duration, amplitude, andlight-curve shape) do not match those of flaring stars.For example, Balona (2015) compiled an atlas of stellarflares observed by the
Kepler satellite in short-cadencemode. They found that 97.8% of stellar flares are shorterthan 0.2 day and that the light-curve shapes and am-plitudes of the remaining 2.2% do not match those ofOGLE-2016-BLG-1928. We also note that the event hasbeen observed by OGLE since 1997 and there is no ev- idence for other flares (nor periodic variability due tostar spots) in the archival data, suggesting the object isunlikely to be a flaring star.Another issue resulting from the short duration ofthe event is the lack of color measurements while thesource is magnified. According to Mr´oz et al. (2020),in microlensing events exhibiting strong finite-source ef-fects, the angular Einstein radius depends on the sur-face brightness of the source, which make color mea-surements critical for determining θ E . In the presentcase, we assumed that the source color is equal to thecolor of the baseline object, which is motivated by thelack of evidence for the blended light in the best-fittingmodels. This issue may become more important for the Roman telescope which is being designed to carry outobservations in
W146 filter with a 15 minute cadenceand in
Z087 filter with a 12 hr cadence. Johnson et al.(2020) estimate that only approximately 10% of short-timescale events due to 1 M ⊕ lenses would have a colormeasurement. We thus advocate that the frequency of Z087 filter observations should be increased.The discovery of OGLE-2016-BLG-1928 demonstratesthat current microlensing surveys are capable of findingextremely -short-timescale events. Although the mass ofthe lens cannot be unambiguously measured, propertiesof the event are consistent with the lens being a sub-Earth-mass object with no stellar companion up to theprojected distance of ∼ Alard, C., & Lupton, R. H. 1998, ApJ, 503, 325 Albrow, M. D., Horne, K., Bramich, D. M., et al. 2009,MNRAS, 397, 2099