Spitzer Microlensing of MOA-2016-BLG-231L : A Counter-Rotating Brown Dwarf Binary in the Galactic Disk

Brown dwarfs (BDs) are substellar objects that are not massive enough to burn hydrogen. BDs have a mass between gas giant planets and low-mass stars, and it is thought that the formation and evolution of BDs are different from those of planets and stars (Ranc et al., 2015). Thus, studying BDs is helpful to understand the formation and evolution of stars and planets.

Microlensing is an excellent method to detect faint low-mass objects, such as BDs and planets because it does not depend on the light from the objects, but the mass. Until now 32 BDs have been detected by microlensing. Only five of these are isolated BDs, while all the others belong to binary systems. Microlensing BDs are mostly binary companions to faint M dwarf stars (see Table 1), while most of the many BDs detected by the radial velocity, transit, and direct imaging methods are companions to solar-type stars (Ranc et al., 2015). Hence, microlensing BDs are important to constrain BD formation scenarios including turbulent fragmentation of molecular clouds (Boyd & Whitworth, 2005), fragmentation of unstable accretion disks (Stamatellos et al., 2007), ejection of protostars from prestellar cores (Reipurth & Clarke, 2001), and photo-erosion of prestellar cores by nearby very bright stars (Whitworth & Zinnecker, 2004).

However, with microlensing, it is generally difficult to measure the mass of a lens. This is because we usually obtain only the Einstein timescale,

where $\theta_{\rm E}$ is the angular Einstein radius corresponding to the total lens mass and $\mu_{\rm rel}$ is the relative lens-source proper motion. For the measurement of the lens mass, one needs to measure the angular Einstein radius and microlens parallax $\pi_{\rm E}$, which yields

where $\pi_{\rm rel}\equiv{\rm au}(D_{\rm L}^{-1}-D_{\rm S}^{-1})$ is the lens-source relative parallax, $D_{\rm L}$ and $D_{\rm S}$ are the distances to the lens and source, respectively, and $\kappa\equiv 4G/(c^{2}\rm au)\approx 8.14\,{\rm mas}/M_{\odot}$ (Gould, 2000). The angular Einstein radius can be measured from events with finite-source effects, while the microlens parallax can be measured from the detection of light-curve distortions induced by the orbital motion of Earth on a standard microlensing light curve (Gould, 1992, 2013). The large number of microlensing BDs detected to date would appear to indicate that M dwarf-BD binaries and BD binaries are very common. This is because their mass measurements (hence, unambiguous determination that they are BDs) require clear detection of a microlens parallax signal that can be detected from the ground despite a relatively short timescale. However, it is usually very difficult to measure the microlens parallax, especially for short timescale events, such as M dwarf-BD binaries. While the microlensing parallax (derived from Equation(2))(Gould, 2000)

is on average large for low-mass lenses, this is not by itself usually sufficient to render it measurable in short events. Rather, large $\pi_{\rm rel}$ is required as well. As a result, half (7/13) of microlensing binaries containing at least one BD and with mass measurements based on ground-based microlensing parallaxes have distances $D_{\rm L}\lesssim 2\,$kpc, which would be true of a tiny fraction of all microlensing events. Moreover, of the remainder, all lie in the Galactic disk $D_{\rm L}<5\,$kpc, and almost all have low, or very low proper motions (so long, or very long timescales), which again is rare.

Hence, it is important to check independently that these relatively frequent detections are not just due to systematics misinterpreted as $``$parallax signal”. The Spitzer satellite allows us to do that. Spitzer observations together with ground-based observations yield the microlensing parallax, which does not depend strongly on the event timescale and lens distance. Thus, Spitzer makes it possible to measure the masses of the lenses in short $t_{\rm E}$ binary BD events, which would be quite difficult using only ground-based observations. The microlens parallax is measured from the difference in the light curves as seen from the two observatories with wide projected separation $D_{\perp}$ (Refsdal S., 1966; Gould, 1994), which is represented by

where

and where the subscripts indicate the parameters as measured from the satellite and Earth. Here $t_{0}$ is the time of the closest source approach to the lens (peak time of the event) and $u_{0}$ is the separation between the lens and the source at time $t_{0}$.

In this paper, we report the discovery of the fifth binary BD from the analysis of the microlensing event MOA-2016-BLG-231, which was observed from the ground and from Spitzer. Although Spitzer can identify binary BDs events with short $t_{\rm E}$, observing such events is extremely challenging because of the short timescale and the $3-9$ day observation delay (see Figure 1 of Udalski et al. 2015). Thus, Spitzer does not usually observe caustic crossings of such events. However, we here show that even non caustic-crossing Spitzer light curves can resolve the nature of a binary BD lens.

As mentioned in Section 1, $\theta_{\rm E}$ and $\pi_{\rm E}$ should be measured for the measurements of the mass and distance of the lens. Thanks to caustic-crossing features, one measure $\rho$ from the modeling, while $\theta_{\rm E}=\theta_{\rm\star}/\rho$. Thus, we need to estimate $\theta_{\rm\star}$ for the measurement of $\theta_{\rm E}$. We estimate $\theta_{\rm\star}$ from the intrinsic color and brightness of the source, which can be derived by the offset between the source and the red clump in the instrumental color and magnitude diagram (CMD) (see Fig. 6) (Yoo et al., 2004). These are determined by

where $[\Delta(V-I),\Delta I]$ is the offsets of the color and brightness between the source and the clump. We determine the instrumental source color by regression of the $V$ on $I$ flux measurements and derive the source instrumental magnitude from the model. These have errors of 0.004 and $0.020$ mag, respectively. We then centroid the clump in color and magnitude with errors of 0.022 and 0.05 mag respectively. The measured color and magnitude offsets are $[\Delta(V-I),\Delta I]=[0.06\pm 0.02,-0.19\pm 0.05]$. Adopting $[(V-I),I]_{\rm clump,0}=(1.06,14.44))$ from Bensby et al. (2011) and Nataf et al. (2013), we find $[(V-I),I]_{\rm s,0}=[1.12,14.25]$. This indicates that the source is a K-type giant. We determine the angular radius of the source $\theta_{\rm\star}$ using $VIK$ color-color relation (Bessell & Brett, 1988) and the color/surface brightness relation (Kervella et al., 2004). As a result, we find that $\theta_{\rm\star}=7.23\pm 0.40\ \mu{\rm as}$. With the measured $\rho$ and $\theta_{\rm\star}$, we determine the angular radius of the Einstein ring corresponding to the total mass of the lens,

The relative lens-source proper motion is

Using the estimated $\theta_{\rm E}$ and $\pi_{\rm E}=0.99^{+0.31}_{-0.35}$, we measure the total mass of the lens system,

The lens is composed of low-mass BDs with masses $M_{1}=0.020^{+0.011}_{-0.005}\ M_{\odot}$ and $M_{2}=0.009^{+0.005}_{-0.002}\ M_{\odot}$, where $M=M_{1}+M_{2}$, and the projected separation of the two BDs is $a_{\perp}=0.88^{+0.27}_{-0.15}$ au. The relative parallax between the lens and the source is

Assuming that the source is located at 8.3 kpc (Nataf et al., 2013), we estimate the distance to the lens,

Hence, the lens is a BD binary located in the Galactic disk. This is the fourth BD binary discovered by microlensing. Because $\theta_{\rm E}$ is well measured due to a precise $\rho$ measurement, which comes from good coverage of caustic-crossing, the errors in the mass and distance of the lens primarily reflect the error in $\pi_{\rm E}$. As shown in Figure 4, the correlation between $\pi_{\rm E,N}$ and $\pi_{\rm E,E}$ components is not well approximated by a linear relation. Hence, we use the best-fit MCMC chains to determine the errors in physical lens parameters including the mass and distance of the lens. Then, one can determine the standard deviation of each physical parameter from the chains. Thus, physical lens parameters in Table 3 represent the median values of each physical parameter from the $(+,-)$ and $(-,-)$ MCMC chains, and their error bars represent the 16th and 84th percentile values of each parameter from the chains.

In order to check that the binary lens is a bound system, we compute $\beta$, i.e., the ratio of the projected kinetic to potential energy (An et al., 2002),

where $\gamma=[(ds/dt/s)^{2}+(d\alpha/dt)^{2}]^{1/2}$. Since $\beta=0.16^{+0.20}_{-0.10}$ (or $\beta=0.52^{+0.30}_{-0.21}$ for $(-,-)$ model) represents very typical values for a bound pair seen at random orientation, and in particular indicates that the lens system satisfies the condition of a bound system, $\beta<1$ (An et al., 2002), it is valid. We list estimated physical parameters of the lens system in Table 3. In Table 3, we also list physical lens parameters for four models with $\Delta\chi^{2}\sim 9$ from Figures 4 $\&$ 5. Three of the four models imply that the lens system is a low-mass BD binary in the disk, similar to the best-fit solutions, while for model 1 it is a binary composed of a low-mass star and a BD in the disk. This is because of large uncertainties of $\pi_{\rm E,N}$, as shown in Figure 4. We further discuss these models of the lens system in Section 6.

Table 2 shows that all $\pi_{\rm E,E}$ for two best models is negative. This means that the lens located in the Galactic disk is moving in the opposite direction to the disk orbital motion. It is unusual. In order to check that it is real, we conduct the modeling under the condition that high-order effect parameters set to zero and the initial values of the other standard model parameters including Spitzer fluxes set to the best-fit solutions of the two parallax+orbital models including $(+,-)$ and $(-,-)$. We also conduct the same test for the four models indicated in Figure 4. As a result, we find that for all of the six models the slope of the Spitzer fluxes is steeper than the model (see Figure 7). This means that Spitzer observed the event later than Earth, and the lens is moving toward Spitzer, i.e., toward the west. Therefore, it is real that the lens is moving in the opposite direction to the disk orbital motion. In addition, in order to demonstrate that the lens is moving west relative to the source, we measure the proper motion of the source (see Figure 8). We find that $\bm{\mu}_{\rm S}=(\mu_{\rm S,N},\mu_{\rm S,E})=(0.021\pm 0.699,0.231\pm 0.699)\ {\rm mas\ yr^{-1}}$. As shown in Figure 8, the source is certainly part of the bulge population and is not moving relative to the bulge stars. Hence, the measurement of the lens-source relative motion clearly means that the lens is not moving with the disk. With the measurement of the source proper motion, we can measure the heliocentric proper motion of the lens,

where $\bm{v}_{\oplus,\perp}$ is the velocity of Earth at the peak of the event and projected perpendicular to the directory of the event. For this case, $\bm{v}_{\oplus,\perp}=(v_{N},v_{E})=(1.09,26.89)\ \rm km\ s^{-1}$. From Equation (14), we find that $(\mu_{\rm L,hel,N},\mu_{\rm L,hel,E})=(-6.16^{+0.40}_{-0.38},-0.37^{+0.25}_{-0.39})\ {\rm mas\ yr^{-1}}$ for $(+,-)$ solution, while for $(-,-)$ solution, $(\mu_{\rm L,hel,N},\mu_{\rm L,hel,E})=(-6.09^{+0.43}_{-0.39},-0.87^{+0.34}_{-0.74})\ {\rm mas\ yr^{-1}}$. From the measurement of $\bm{\mu}_{\rm L,hel}$, we can also measure the transverse velocity of the lens, (Shvartzvald et al., 2018),

where $\bm{v}_{\odot}=\bm{v}_{\odot,\rm pec}+\bm{v}_{\odot,\rm cir}$. Here $v_{\odot,\rm pec}(l,b)=(12,7)\ \rm km\ s^{-1}$ are the transverse components of the Sun’s peculiar velocity relative to the Local Standard of Rest and $v_{\odot,\rm cir}(l,b)=(220,0)\ \rm km\ s^{-1}$ is the disk circular velocity. Thus, the peculiar velocity of the lens relative to the mean motion of the Galactic disk stars (Shvartzvald et al., 2018) is

where $\bm{v}_{\rm L,cir}=\bm{v}_{\odot,\rm cir}$ because both lens and Sun are disk stars. From Equation (16), we find that for two degenerate solutions, $(+,-)$ and $(-,-)$, $v_{\rm L,pec}(l,b)=(-144.12^{+37.69}_{-50.40},-27.78^{+5.95}_{-7.79})\ \rm km\ s^{-1}$ and $v_{\rm L,pec}(l,b)=(-172.29^{+38.72}_{-66.15},-24.98^{+7.17}_{-7.71})\ \rm km\ s^{-1}$, respectively. This means that the lens is counter-rotating relative to the motion of Galactic disk stars. In order to check that the four models with $\Delta\chi^{2}\sim 9$ in Figure 4 are compatible with the two solutions, we also estimate the proper motions and peculiar velocities of the lens for the four models, and they are as follows as

The lens proper motions of the four models are consistent with those of the two best-fit solutions. Therefore, the lens is a counter-rotating disk object. Until now two counter-rotating disk objects (OGLE-2016-BLG-1195L (Shvartzvald et al., 2017), OGLE-2017-BLG-0896L (Shvartzvald et al., 2018)) have been discovered by microlensing, the first being right at the hydrogen-burning limit and the second being a low-mass BD. MOA-2016-BLG-231L is the third such object, and it is the first counter-rotating BD binary discovered by Spitzer. The unusual kinematics of the BD binary suggest that the BD binary could be a halo object or a member of counter-rotating low-mass object population, as mentioned in Shvartzvald et al. (2018).

For the event MOA-2016-BLG-231, it is found that there is an extremely small offset between the source position and the baseline object. The offset is 0.05 pixels, and it corresponds to $0.05\times 0.26\ \rm arcsec=13$ mas. This means that the blend could be associated with the event. However, the blended flux is in fact consistent with zero. First, the formal estimate of the blended flux is $f_{b}=0.33\pm 0.17$, where one unit of flux corresponds to $I=18$. This in itself is consistent with zero at the $2\,\sigma$ level. Moreover, the estimate of $f_{b}$ is ultimately derived from $f_{b}=f_{\rm base}-f_{s}$, where $f_{s}$ is the source flux from the microlensing model and $f_{\rm base}$ comes from DoPhot (Schechter al., 1993) photometry of this field location. Because of the mottled background of unresolved turnoff stars in these croweded fields, $f_{\rm base}$ can easily have errors at this level. Therefore, there is no clear evidence for blended light. In addition, we measure $I_{\rm base}$ using OGLE deep stack images. From this, we find that $I_{\rm base}$ from the deep stack images is 0.06 mag brighter than the value obtained from the OGLE reference images (i.e. normal OGLE data) due to a nearby star at $0.6^{\prime\prime}$. Hence, $I_{\rm base}$ cannot be estimated from the reference image photometry to better than 0.06 mag due to the presence of the nearby star. Therefore blended light cannot be determined better than 0.5 flux units. This reinforces the naive conclusion that there is no evidence for light from the lens.

The result of modeling with chain variables $F_{\rm s,spitzer}$ and $F_{\rm b,spitzer}$ yields negative $F_{\rm b,spitzer}$ (see Table 2). $F_{\rm b,spitzer}\sim-4.5$ (see Table 2). This is somewhat unusual. As reported in Calchi Novati et al. (2015) and Shvartzvald et al. (2018), the negative Spitzer blending can be generated when the flux of unresolved faint stars is included in the global background flux. The event OGLE-2017-BLG-0896 (Shvartzvald et al., 2018) is also the event affected by the excess flux due to unresolved stars and has almost the same Spitzer blending as this event.

The result of this study indicates that the lens system of MOA-2016-BLG-231 consists of two low-mass BDs. In principle, there are two scenarios that would drive us to very different conclusions, but neither of them is likely to be true. First, a smaller parallax value would give a more massive and more distant lens system. For example, out of the four models indicated in Figure 4, model 1 has the smallest parallax, which produces the most massive and the most distant lens system and thus might in principle solve the puzzle on the counter-rotating motion. However, the model 1 gives a low-mass M dwarf-BD binary with $M_{\rm tot}=0.18\ M_{\odot}$ at a distance of $D_{\rm L}=6.3\ \rm kpc$, which still implies that it is in the disk. Moreover, with proper motion $\mu_{\rm L,hel}(N,E)=(-3.7,-5.0)$, it is still counter-rotating. Therefore, the smaller parallax scenario is unlikely to resolve the issue of a counter-rotating lens. Furthermore, for the other models, all the lens systems are low-mass BD binaries with $M_{\rm tot}=0.02\sim 0.03\ M_{\odot}$ that is located at the disk $D_{\rm L}<3.2\ \rm kpc$ and their proper motions are consistent with the two best-fit models (see Table 3), as mentioned before.

Another possibility could be that such an unusual solution arises from the unknown systematics in the Spitzer photometry. Poleski et al. (2016) and Zhu et al. (2017) have observed systematics on timescales of tens of days. In principle, the Spitzer light curve could be affected by such systematics, but they cannot be recognized because the total duration of the light curve is short. However, Zhu et al. (2017), based on the Spitzer event sample that was uniformly analyzed in that work, concluded that such long-term systematic trends in the Spitzer photometry appear in $<5\%$ of all cases. Therefore, there is a $<5\%$ probability that systematics in the Spitzer photometry led to our current solution. The proper motion is $6.6\ \rm mas\ yr^{-1}$ (independent of the parallax measurement). Thus, the solution can be checked with adaptive optics observations at first light of next-generation, thirty-meter class telescopes. For example, in 2028 the source and lens will be separated by $79\ \rm mas$, which is $\sim 6$ times the FWHM at 1.6 microns for a 30-m telescope, so easily resolved even though the source is a giant. If the lens really is a brown dwarf, no light should be detected. However, if the solution is wrong (e.g. due to unknown systematics), then the lens should be more massive, i.e. a star. In that case, light from the lens should be directly detected. Furthermore, if light is detected, this will also allow a check of the direction of the source-lens relative proper motion.

We present the analysis of the binary lensing event MOA-2016-BLG-231 that was observed from the ground and from Spitzer. Even though Spitzer did not cover caustic-crossing parts and it covered only partial wing parts of the light curve, we could determine the physical properties of the lens. It is found that the lens is a binary system composed of low-mass BDs with $M_{1}=21^{+12}_{-5}\ M_{J}$ and $M_{2}=9^{+5}_{-2}\ M_{J}$, and it is located in the Galactic disk $D_{\rm L}=2.85^{+0.88}_{-0.50}\ {\rm kpc}$. The BD binary is moving counter to the orbital motion of disk stars. This solution can be checked in the future with adaptive optics observations with thirty-meter class telescopes. This result shows how Spitzer plays a crucial role in short $t_{\rm E}$ BD events, despite the fact that it covered only the declining wing of the light curve in a relatively short event.