UKIRT-2017-BLG-001Lb: A Giant Planet Detected through the Dust

Gravitational microlensing is unique in its ability to probe relatively untapped reservoirs of exoplanet parameter space (Gaudi 2012). These include planets at all masses near the "snowline" where gas and ice giants are likely to form (Ida & Lin 2005), hence probing the frequency and mass function of snowline planets (Cassan et al. 2012; Shvartzvald et al. 2016; Suzuki et al. 2016); planetary systems around all types of stars throughout the Galaxy, thus allowing measurement of the Galactic distribution of planets (Calchi Novati et al. 2015a; Penny et al. 2016; Zhu et al. 2017); and free-floating planets (Sumi et al. 2011; Mróz et al. 2017). However, currently only 51 planetary systems hosting 53 planets¹⁰ have been discovered using microlensing. While new ground-based microlensing surveys with global networks of telescopes and high cadences (e.g., KMTNet; Kim et al. 2016) are expected to detect a few tens of planets per year (Henderson et al. 2014), it is clear that a space-based survey is required in order to significantly increase the number of detected systems and to detect planets with masses substantially less than that of the Earth, thus fully exploiting the microlensing potential (Bennett & Rhie 2002; Penny et al. 2013).

The proposed Wide Field InfraRed Survey Telescope (WFIRST) flagship mission (Spergel et al. 2015), which is planned to launch in mid-2020s, would dedicate ∼25% of its lifetime to a microlensing survey. This survey is predicted to discover thousands of exoplanets near or beyond the snowline via their microlensing light curve signatures (M. T. Penny et al. 2018, in preparation), enabling a Kepler-like statistical analysis of planets ∼1–10 au from their host stars and potentially revolutionizing our understanding of planet formation. In addition to the superb photometry, high cadence, and continuous observations, the survey will be conducted in the near-infrared (NIR). An NIR microlensing survey suffers from less extinction than traditional optical surveys, enabling observations closer to the Galactic plane and center, where the stellar surface density of sources and lenses, and thus the microlensing event rate, is highest. However, until recently no dedicated NIR microlensing survey has been conducted, and so the event rate in the NIR has not been measured, which is crucial for WFIRST field optimization (Yee et al. 2014).

During the 2015 and 2016 seasons, we conducted the first NIR microlensing survey with the United Kingdom Infrared Telescope (UKIRT) as support for the Spitzer and Kepler microlensing campaigns. From examination of 5% of the 2016 UKIRT fields (overlapping with optical survey fields), we discovered five highly extinguished, low-Galactic latitude microlensing events (Shvartzvald et al. 2017). These events were not detected by optical surveys, likely due to the high extinction. Combining these detections with additional events that were also detected by optical surveys in these fields, we found evidence that the event rate is indeed higher in this region, closer to the Galactic plane. In 2017, we initiated a full NIR microlensing survey with UKIRT, covering all potential WFIRST fields including the Galactic plane and center (see Figure 1), which are inaccessible to optical surveys due to the high extinction. The fields are observed with a daily cadence, allowing us to easily detect microlensing events with their typical timescale of approximately 20 days (Shvartzvald et al. 2016), and with a cadence of three epochs/night in the central fields, which are expected to have an excess of short-timescale bulge–bulge events (Gould 1995).

Zoom In Zoom Out Reset image size

Bound microlensing planets are discovered through their anomalous signature on the otherwise smooth single-lens light curve. These anomalies occur when the source passes near or over a caustic. There are three types of caustics (for a thorough review see Gaudi 2012): central caustics (associated with the host), planetary caustics (associated with the planet), and resonant caustics, which occur when the central and planetary caustics merge into a single caustic. The location and shape of the caustics are determined by the mass ratio q between the host and the planet, and the scaled (with respect to the Einstein radius θ_E) instantaneous projected separation s. The size of the caustics is mostly determined by the mass ratio q (as the separation s changes by only a factor 2 in the "lensing zone" (Griest & Safizadeh 1998)), and thus for events due to a star and a planet (and thus small q) the caustics are usually small. Anomalies due to planets typically last ≲1 day for planetary or central caustics, and thus require high cadence in order for them to be detected and well characterized. However, resonant caustics are always larger than either the planetary or central caustics (for a given mass ratio q) and thus have a larger cross section (Dominik 1999), as well as longer anomaly duration, and therefore can be detected even with a daily cadence. In fact, about 40% of the microlensing planetary systems to date have been detected through resonant caustic perturbations.¹¹ These include the first microlensing planet (Bond et al. 2004), the first planet simultaneously observed from ground and space (Udalski et al. 2015), and the two multiple-planet systems (Gaudi et al. 2008; Han et al. 2013). All of these events also had long timescales (60–150 days), thus the anomalies lasted several days.

In this Letter, we report the discovery of the planetary event UKIRT-2017-BLG-001. This is the first planet to be detected solely by the UKIRT survey. The planetary perturbation is caused by a resonant caustic of a giant planet (q ∼ 10⁻³) and the event has a long timescale of ∼100 days. Thus, the survey cadence was sufficient for the detection and characterization of the planet. The event lies close to the Galactic center (035) and in a field with significant total and spatially variable extinction, thus it could not be detected by optical microlensing surveys.

We describe the UKIRT observations and event detection in Section 2. In Section 3 we present the best-fit microlensing model of the event. In Section 4 we derive the source properties by analyzing the color–magnitude diagram (CMD) of the event, finding indications that suggest it is a far-disk source. In Section 5 we estimate the physical properties of the planetary system using a Bayesian analysis. An analysis of the multi-band extinction toward this field is presented in Section 6. Finally, we summarize and discuss our results in Section 7.

The UKIRT microlensing survey uses the wide-field NIR camera (WFCAM) at the UKIRT 3.8 m telescope on Mauna Kea, Hawaii. The 2017 fields cover the northern bulge (b > 0), the Galactic center, and the southern bulge (see Figure 1). Observations of the central fields are primarily done with the K_S-band, with a nominal cadence of three epochs/night. In addition, H-band observations are taken once every three nights. The northern and southern bulge fields are observed once per night with the H-band and once every five nights with the K_S-band. Each epoch is composed of 16 5-s co-added dithered exposures (two co-adds, two jitter points, and 2 ×2 microsteps). The UKIRT dithered images are reduced, astrometrically calibrated, and stacked by the Cambridge Astronomy Survey Unit (CASU; Irwin et al. 2004). The light curves of all sources are then extracted using two methods—(a) Two Micron All Sky Survey (2MASS)-calibrated soft-edge aperture photometry (standard CASU individual image catalogs; Hodgkin et al. 2009), and (b) 2MASS-calibrated point-spread function (PSF) photometry using SExtractor (Bertin & Arnouts 1996) and PSFEx (Bertin 2011). The latter is better for detecting and measuring faint objects in our crowded bulge fields.

We searched the full 2017 data set for microlensing events using the new event detection algorithm of Kim et al. (2018). Among the events found (the full analysis will be presented in a subsequent paper), we identified UKIRT-2017-BLG-001. The event lies inside one of our central fields, at equatorial coordinates (R.A., decl.)_J2000.0 = (17:46:36.98, −29:12:40.9), and Galactic coordinates (l, b) = (−0.12, −0.33).

After the discovery of the event and the anomaly over its peak, we extended the observations of the field covering the event for additional two weeks beyond the main 2017 campaign. However, these observations were only with one pointing (toward the specific field) unlike the standard observational sequence of a full tile (covering a continuous area with four pointings). This results in a different procedure for deriving the sky frames that are used to correct for additive artifacts such as scattered light, residual reset anomaly, and illumination-dependent detector artifacts (for more details see the UKIRT/WFCAM technical website¹² ). The sky estimation for single pointings leaves some residual object flux and thus introduces a small systematic bias that progressively increases at faint magnitudes. In order to incorporate this systematic effect in the PSF photometry light curve, we measured the distribution of flux offset using all of the sources near the event (<2') for each epoch in the additional two weeks compared to the mean flux during the main survey. We then estimate the systematic error per epoch using the flux offset dispersion after subtracting the mean flux offset dispersion of the main survey. Finally, we add (in quadrature) this systematic error to the reported PSF flux errors.

For the modeling of the event (Section 3) we use the PSF photometry of the full data set, while for the CMD that is used to derive the source properties (Section 4), we use the PSF catalog from the main survey only. Based on the rms distribution of the PSF photometry for all of the sources in our field, we add (in quadrature) a minimum error, e_min = 0.015, to the pipeline reported errors, in order to compensate for unrealistically small Poisson flux errors when the event is bright. We note that this is an empirical correction and is not driven by the fit for the model.

The light curve (Figure 2) has a clear and long anomaly (∼8 days) just after the peak of a moderate-magnification microlensing event. The anomaly starts with a small deviating rise above the expected single-lens model, followed by a sharp fall. Then the event rises again and continues to fall approximately as expected by a single-lens microlensing event. The combination of these features suggests a source crossing over a planetary resonant caustic, exiting at the "back" side of the caustic opposite to the planet (and thus the sharp fall below the single-lens model), followed by a cusp approach (see Figure 3).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We model the event using seven geometric parameters to calculate the magnification, A(t), of a binary-lens system. These are the three point-lens parameters $({t}_{0},{u}_{0},{t}_{{\rm{E}}})$ (Paczynski 1986), and three parameters for the companion: the mass ratio q, the instantaneous scaled projected separation s, and the angle α (measured counter-clockwise) between the source trajectory and the binary axis in the lens plane. The seventh parameter is the scaled angular source size $\rho ={\theta }_{* }/{\theta }_{{\rm{E}}}$. To calculate the model magnifications near and during the anomaly we employ contour integration (Gould & Gaucherel 1997) with 10 annuli to allow for limb darkening. We adopt linear coefficients u_H = 0.3895 and u_K = 0.3324 (Claret & Bloemen 2011), based on the source type derived in Section 4.¹³ Far from the anomaly we employ limb-darkened multipole approximations (Gould 2008; Pejcha & Heyrovský 2009). Finally, each data set has two flux parameters representing the source (${f}_{s,H/{K}_{S}}$) and any additional blend (${f}_{b,H/{K}_{S}}$): ${F}_{H/{K}_{S}}(t)={f}_{s,H/{K}_{S}}A(t)+{f}_{b,H/{K}_{S}}$.

To identify initial possible solutions we search over a grid of mass ratios ranging q = 0.0001–1.0 and separations s = 0.1–1.4, fully containing the relevant parameter space of resonant caustics. For each grid point we initiate a set of Markov-chain Monte-Carlo (MCMC) chains with all possible angles α. We find an isolated single minimum centered on (s, q) ∼ (1.02, 0.001) (see Figure 4) and initiate from this point a full MCMC analysis (with all parameters free) to find the best-fit model.

Zoom In Zoom Out Reset image size

The source of the event is faint, with ${K}_{S,s}\gtrsim 16$, which is very close to the survey limiting magnitude (see Figure 6). Moreover, our data do not completely cover the baseline of the event. Yee et al. (2012) showed that some of the standard microlensing parameters, in particular t_E and f_s, can be poorly constrained for such events with faint sources. However, they introduced the microlensing invariants $({t}_{0},{t}_{\mathrm{eff}}\equiv {u}_{0}{t}_{E},{t}_{* }\,\equiv \rho {t}_{E},{{qt}}_{E},{f}_{\mathrm{lim}}\equiv \,{f}_{s}/{u}_{0})$, which are well-constrained from the region near the peak and anomaly of the event. Table 1 gives the values and uncertainties of the best-fit parameters and the microlensing invariants.

Table 1.  Microlensing Model

Parameter
t0 [HJD']	7916.243 ± 0.022
u0	${0.0303}_{-0.0026}^{+0.0036}$
tE [day]	${101.0}_{-9.6}^{+8.2}$
ρ [10−3]	${6.64}_{-0.56}^{+0.75}$
α [rad]	${3.7036}_{-0.0076}^{+0.0084}$
s	${1.0318}_{-0.0045}^{+0.0032}$
q [10−3]	${1.50}_{-0.14}^{+0.17}$
${K}_{S,s}$	${16.07}_{-0.11}^{+0.09}$
${K}_{S,b}$	>17.7a
teff [day]	3.062 ± 0.050
t* [day]	0.671 ± 0.016
${{qt}}_{{\rm{E}}}$ [day]	0.1517 ± 0.0041
flim	198.2 ± 2.0

Note. HJD' = HJD-2450000.

^a5σ limit.

Download table as:  ASCIITypeset image

We further try to include in the model microlensing parallax or orbital motion of the binary-lens system. The inclusion of these higher-order effects does not significantly improve the fit (which requires two additional parameters for each effect) with Δχ² = 5 when including only microlens parallax, Δχ² = 9 when including only orbital motion, and Δχ² = 14 when including both effects. We do not consider either of these as a detection because they all require significant negative blending (which is unphysical) and systematics at that level are well known in microlensing. Nevertheless, we can set a conservative upper limit (${\rm{\Delta }}({\chi }^{2}-{\chi }_{\mathrm{best}}^{2})\lt 30$) on the microlensing parallax of ${\pi }_{{\rm{E}}}\lt 0.7$. In Section 5 we use this limit when deriving the physical properties of the system.

A standard way to derive the intrinsic source properties is by measuring the offset between the source position on a CMD and the centroid of the red clump (RC; Yoo et al. 2004). A fundamental assumption for this method is that the source and the RC are both behind the same dust column. The CMD is thus usually constructed using stars near (≲2') the target, as spatial extinction variations toward the bulge are usually on larger scales. However, in the case of our event there are several challenges with this method.

The first challenge is the observed source properties. The entire range of $(H-{K}_{S})$ colors for giants (which our source likely is, as we show below) is narrow, with only 0.24 mag difference between G0 to M7 giants. Even for dwarfs the range is only 0.38 mag between B8 and M6 dwarfs (Bessell & Brett 1988).¹⁴ This is because $(H-{K}_{S})$ basically probes the Rayleigh–Jeans tail for most stars. Therefore, a precise measurement of the source color is essential in order to derive the source properties. We determine the model-independent source ${(H-{K}_{S})}_{s}$ color from regression of H versus K_S flux as the source magnification changes (Gould et al. 2010), and find ${(H-{K}_{S})}_{s}=1.87\pm 0.01$. (This color is in excellent agreement with the color derived from the model of ${(H-{K}_{S})}_{s}=1.878\pm 0.004$). Each H epoch was taken in between two K_S observations on the same night, minimizing systematics and allowing us to achieve the required precision. Next, the source is faint and, as discussed in Section 3, we do not have coverage of the event's baseline. Therefore, the uncertainty on the source magnitude is relatively large. The K_S source magnitude as inferred from the microlensing model is ${K}_{S,s}={16.07}_{-0.11}^{+0.09}$. The model also indicates a fainter blend flux, with 5σ upper limit of ${K}_{S,b}\gt 17.7$. We use this measurement later as an upper limit on the lens flux when we estimate the physical properties of the planetary system (see Section 5). For completeness, we note that no centroid shift as a function of magnification was detected.

The second and more prominent challenge is the determination of the extinction and reddening using the RC centroid. There are clear dust clouds in the region around our event (see Figure 5), which cause large spatial reddening variations on scales smaller than the standard ≲2' region. We avoid obvious dust stripes and use stars in a 1.6 arcmin² box around the event, as marked in Figure 5, to construct the CMD (Figure 6). We measure the RC centroid following the procedure described in Nataf et al. (2013) and find ${(H-{K}_{S},{K}_{S})}_{\mathrm{cl}}=(1.63,14.63)$. The number of RC stars used, as derived from the fit, is ${N}_{\mathrm{RC}}=197\pm 14$, which is sufficient for a reliable measurement of the RC. By comparing the apparent clump centroid to the intrinsic centroid of ${(H-{K}_{S},{K}_{S})}_{\mathrm{cl},0}=(0.15,12.95)$ toward the event's location (assuming a distance of 8.17 kpc; Nataf et al. 2013, 2016; Hawkins et al. 2017) we find that the mean extinction and reddening toward the event are $\langle {A}_{{K}_{S}}\rangle =1.68$ and $\langle {E}_{H-{K}_{S}}\rangle =1.48$, respectively. We note that the reddening-to-extinction ratio that we find is non-standard, and we investigate this further in Section 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

However, both the color and magnitude dispersions of the RC stars are large, suggesting large extinction and reddening dispersions (and thus large uncertainties on the intrinsic source properties) as well as a possibly wide distance dispersion of clump stars. The observed clump color dispersion is ${\sigma }_{{(H-{K}_{S})}_{\mathrm{cl}}}=0.16$ (as derived from the clump centroid fit; see Nataf et al. 2013 for details), which combines the intrinsic clump color dispersion and reddening dispersion. In order to estimate the intrinsic dispersion we apply the RC centroid procedure to several of our low-Galactic-latitude UKIRT fields that have low and uniform extinction. From the clump color dispersion on these fields we find that the intrinsic clump color dispersion is only 0.04, and thus the uncertainty on the reddening (e.g., the reddening dispersion) in the event's field is ${\sigma }_{{E}_{H-{K}_{S}}}=0.15$. This large uncertainty, if taken as is, implies that the intrinsic color of the source cannot be well constrained, as the full range of colors for giant stars is within $2{\sigma }_{{E}_{H-{K}_{S}}}$ and the full dwarf color range is within $2.5{\sigma }_{{E}_{H-{K}_{S}}}$. The observed clump magnitude dispersion, ${\sigma }_{{K}_{S,\mathrm{cl}}}=0.36$, is even higher. Intrinsic dispersion, extinction dispersion, and distance modulus dispersion all contribute to the total magnitude dispersion (the dispersion due to photometric noise is negligible). The intrinsic magnitude dispersion is 0.17 (Hawkins et al. 2017), and the extinction dispersion can be estimated from the reddening dispersion and the mean extinction-to-reddening ratio to be ${\sigma }_{{A}_{{K}_{S}}}=0.17$. Thus, the distance modulus dispersion to the clump is 0.28. While this value is larger than typical toward Baade's window (Nataf et al. 2013), the event is located very close to the Galactic center (and thus the Galactic plane), implying a large geometrical dispersion due to the fact that there are significant contributions from both bulge and far disk stars (see more details in Section 4.1 below).

The large differential reddening implies that the basic assumption that the source is behind the same dust as the clump is not reliable, as the extinction itself is not uniform across the field. Moreover, assuming that the reddening and extinction to the source are the mean reddening and extinction in the field, $\langle {E}_{H-{K}_{S}}\rangle =1.48$ and $\langle {A}_{{K}_{S}}\rangle =1.68$ yield intrinsic source properties of [(H − K_S), K_S)]_s,"0" = (0.39, 14.39). While the magnitude suggests a giant source, the color is 0.05 redder than an M7 giant (which is already very rare, as suggested by the small intrinsic dispersion of clump star colors). This implies that the source is suffering from higher reddening than the mean clump. While we cannot use the usual method to derive the source properties with small uncertainties, we can use it to set possible limits on the source properties. We assume that the extinction and reddening toward the source are $\langle {A}_{{K}_{S}}\rangle +{\rm{\Delta }}{A}_{{K}_{S}}$ and $\langle {E}_{H-{K}_{S}}\rangle +{\rm{\Delta }}{E}_{H-{K}_{S}}$, respectively. In order to estimate the boundaries of the possible source angular size range, we take two extreme cases of giant sources. As the reddest limit, we assume an M7 giant with $(H-{K}_{S})=0.34$, thus suffering from additional reddening ${\rm{\Delta }}{E}_{H-{K}_{S}}=0.05$ and correspondingly additional extinction ${\rm{\Delta }}{A}_{{K}_{S}}=0.06$. As the bluest limit, we assume a G0 giant with $(H-{K}_{S})=0.10$, thus suffering from ${\rm{\Delta }}{E}_{H-{K}_{S}}=0.29$ and ${\rm{\Delta }}{A}_{{K}_{S}}=0.33$. For each case, we derive the source intrinsic magnitude by accounting for the total extinction. We then use the $(V-K)$ colors (Bessell & Brett 1988) corresponding to the source spectral type and the surface brightness to angular source size relation of Kervella et al. (2004) to infer that the source angular size is in the range $3.3\lt {\theta }_{* }/\mu \mathrm{as}\lt 8.5$.

4.1. Far Disk Source?

4.2. Spatial Differential Reddening

The angular Einstein radius and the relative geocentric proper motion between the source and the lens can be derived from the light curve model and CMD analysis, ${\theta }_{{\rm{E}}}={\theta }_{* }/\rho$ and ${\mu }_{\mathrm{geo}}={\theta }_{{\rm{E}}}/{t}_{{\rm{E}}}$. Applying the limits derived above on θ_* then gives

$$\begin{eqnarray}&&0.44\lt {\theta }_{{\rm{E}}}[\mathrm{mas}]\lt 1.39\qquad 1.5\lt {\mu }_{\mathrm{geo}}[\mathrm{mas}\,{\mathrm{yr}}^{-1}]\lt 5.5.\end{eqnarray} \tag{ 1 }$$

Unfortunately, even using these limits we cannot determine directly the physical properties of the planetary system, as the microlens parallax was not detected. We therefore estimate them using a Bayesian analysis that incorporates the limits on θ_E and μ_geo into a Galactic model. We follow the procedures described in Shvartzvald et al. (2014) and adopt the Galactic model of Han & Gould (1995, 2003), which reproduces well the observed statistical distribution of properties of microlensing events. For the source probability, we apply conservative color and magnitude limits of $15\lt {K}_{S,s}\lt 17$ and ${(H-{K}_{S})}_{s}\gt 1.3$. We furthermore set two limits on the mass–distance relation of the lensing system. The first is from the microlens parallax

$$\begin{eqnarray}&&{\pi }_{{\rm{E}}}\equiv \sqrt{\kappa M{\pi }_{\mathrm{rel}}}\lt 0.7\qquad \kappa \equiv \displaystyle \frac{4G}{{c}^{2}\mathrm{au}}\simeq 8.14\displaystyle \frac{\mathrm{mas}}{{M}_{\odot }}.\end{eqnarray} \tag{ 2 }$$

Here ${\pi }_{\mathrm{rel}}=\mathrm{au}({D}_{L}^{-1}-{D}_{S}^{-1})$ is the lens-source relative parallax. The second limit is an upper limit on the lens flux by using the 5σ upper limit on the blend flux that was derived from the model, ${K}_{S,l}\geqslant {K}_{S,b}\gt 17.7$. We use 5 Gyr Padova isochrones (Bressan et al. 2012; Marigo et al. 2017) and assume (as a conservative limit) that the lens is behind the overall dust column toward the bulge to convert the flux limit to a mass–distance limit. This limit alone already gives an upper limit on the planet mass (if all the blend flux is attributed to the lens), because at this age the maximal host mass is ≲1.25 M_⊙, and thus the planet mass will be $\lt 2.7{M}_{J}$. We note that if the host is a bulge star, with age of ≈10 Gyr, the limits on the host and planet masses are $\lesssim 1.05\,{M}_{\odot }$ and <2.2M_J, respectively. However, as the event is within one scale height of the thin disk throughout the Galaxy, we use the 5 Gyr isochrones to derive the limit on the lens flux for all distances, as there is a non-negligible probability that the lens might be a part of the disk population even at 6–10 kpc.

The 2D posterior distribution of the host mass and distance is shown in Figure 7, as well as the limits on θ_E, π_E, and the lens flux. We infer that the host is a ${0.81}_{-0.27}^{+0.21}\,{M}_{\odot }$ dwarf, likely in the Galactic bulge at ${6.3}_{-2.1}^{+1.6}$ kpc. Using the mass ratio and scaled instantaneous projected separation from the model, we find that the planet mass is ${1.28}_{-0.44}^{+0.37}{M}_{J}$, and it orbits its host beyond the snowline at a projected separation of ${4.18}_{-0.88}^{+0.96}$ au. The estimated physical properties are summarized in Table 2.

Zoom In Zoom Out Reset image size

Table 2.  Physical Properties

M1 [M⊙]	${0.81}_{-0.27}^{+0.21}$
m2 [MJ]	${1.28}_{-0.44}^{+0.37}$
${r}_{\perp }$ $[\mathrm{au}$]	${4.18}_{-0.88}^{+0.96}$
DL [kpc]	${6.3}_{-2.1}^{+1.6}$
DS [kpc]	${11.2}_{-2.6}^{+3.6}$
${\mu }_{\mathrm{geo}}$ [mas yr−1]	${2.34}_{-0.42}^{+0.84}$

Download table as:  ASCIITypeset image

One of the main uncertainties in our Galactic model is the source distance. However, we note the the posterior probability for the lens mass (and thus the planet mass) is weakly dependent on the source distance. For a source at 8 kpc the mass is only 10% lower than for a source at 15 kpc, well within the range of our uncertainty. The lens distance, and consequently the projected separation between the planet and its host, is more sensitive to the source distance, with a difference of 45% between a bulge source and a far disk source. Future observations can resolve this and give a better estimation of the source distance, as we discuss below in Section 7.

Recent studies of dust properties in the inner bulge suggest deviations from the standard extinction law (Nataf et al. 2016; Alonso-García et al. 2017). The extinction coefficient measured in our field, $\langle {A}_{{K}_{S}}\rangle /\langle {E}_{H-{Ks}}\rangle =1.14$, is lower than the value of 1.37 from Nishiyama et al. (2009). This offset is a disconcertingly large ∼0.4 mag in K_S. The offset relative to the value of ${A}_{{Ks}}/{E}_{H-{Ks}}\sim 2.07$ predicted by Fitzpatrick (1999) is even larger—about 1.4 mag in K_s.

The magnitude of the RC can also be predicted in the YJ bands by assuming the color–color relations from Nataf et al. (2016) and the extinction coefficients from Fitzpatrick (1999). These are Y_RC ≈ 14.63 + 1.63 + (7.35–3.12) ≈ 20.49 and J_RC ≈ 14.63 + 1.63 + (5.44–3.12) ≈ 18.58, where we add the corresponding intrinsic color offsets and reddening offsets, respectively, to the observed clump H-band magnitude.

Given that the limiting magnitudes for the DECam and VVV data sets are, respectively, Y ∼ 21.2 and J ∼ 19.4 (see Section 4.2 above), the RC should be clearly detected in each of these bandpasses by a wide margin of 0.7–0.8 mag, respectively. However, the photometry in both cases barely covers the RC. Figure 8 shows the HK_s CMD of our field (gray points), with the sources detected in Y shown as yellow circles in the left panel, and the points detected in J shown as purple in the right panel. We conclude that the extinction toward this field is steeper than standard (as defined by Fitzpatrick 1999).

Zoom In Zoom Out Reset image size

We have presented the discovery of a roughly Jupiter/Sun mass ratio planet. The system likely resembles a version of the Jupiter/Sun system, with the host being a G/K dwarf slightly less massive than the Sun, located in the Galactic bulge. The event was detected as part of our UKIRT microlensing survey, which has the primary goal of deriving the NIR event rate toward the WFIRST target field region. As such, the survey is designed to have a ∼daily cadence, which in principle is not optimal for the detection of short planetary anomalies. However, the perturbation was due to a resonant planetary caustic, and the event had a long timescale, so the cadence was sufficient to detect the planetary anomaly. This event contributes to the set of interesting planetary microlensing events that were discovered through resonant caustic perturbations. The resonant caustic parameter space is relatively wide for giant planets. For q = 10⁻³, resonant caustics exist for instantaneous projected separations of 0.93 < s < 1.15, corresponding to 2.6–3.3 au for a typical Einstein radius. However, the range in projected separations for which resonant caustics exist scales as q^1/3, and thus resonant caustics are less important for lower mass ratios q.

The field around the event suffers from high and differential extinction that creates a challenge for deriving the intrinsic source properties, and thus the physical properties of the planetary system. Follow-up observations of UKIRT-2017-BLG-001Lb can reduce these uncertainties in several ways. First, deep high-angular resolution imaging of the field in the infrared would enable an accurate measurement of the extinction toward each star (including the source) using the Rayleigh–Jeans color excess method (Majewski et al. 2011). This can be done from the ground with Keck in the NIR or, preferably, with JWST in mid- and near-infrared bands allowing for a wider spectral range. Second, JWST could also measure a medium-resolution spectrum of the source star in the NIR to estimate its temperature and improve the estimate of the angular Einstein radius (the upgraded Near Infrared Spectrograph (NIRSPEC) on Keck will potentially have the sensitivity for a low-resolution spectrum of the source in H-band). Third, the high-resolution image will resolve out possible unrelated sub-arcsecond blend stars around the target (see, e.g., Beaulieu et al. 2018), allowing the measurement of any excess flux from the target above the measured source flux. However, the blend flux from the microlensing model (${K}_{S,b}\gt 17.7$) already sets an upper limit on the excess flux. For such faint excess the probability for a significant contribution from stars that are not the lens (i.e., companion to lens, companion to source, or ambient star) is high (see, e.g., Koshimoto et al. 2017). Finally, a second high-resolution epoch of UKIRT-2017-BLG-001Lb could resolve the lens and the source and directly measure the relative lens-source proper motion and the lens flux (e.g., Batista et al. 2015). Given that the relative proper motion is in the range 1.5 < μ[mas yr⁻¹] < 5.5 (with Bayesian estimate of μ_geo ∼2.5 mas yr⁻¹) the source and the lens will be sufficiently separated in 10–50 years. However, centroid shifts due to the relative lens-source proper motion (e.g., Bennett et al. 2015; Bhattacharya et al. 2017) could be detected earlier.

Future NIR microlensing surveys, such as the planned WFIRST microlensing program or the one proposed with Euclid (Penny et al. 2013), may consider avoiding such regions with large spatial extinction variations. On the other hand, our current results (Shvartzvald et al. 2017) suggest that the NIR event rate is highest close to the Galactic center, and thus the field selection should balance between the high differential extinction and the high event rate. We note that the currently planned WFIRST fields (M. T. Penny et al. 2018, in preparation) lie at lower Galactic latitudes (b ≈ −2.05 to −0.45), in regions of lower extinction and lower differential extinction than UKIRT-2017-BLG-001Lb.¹⁷

The multi-band analysis of our field suggests a non-standard (steeper) extinction law. This supports previous suggestions of interstellar extinction law variations toward the inner bulge (Nataf et al. 2016; Alonso-García et al. 2017). Our ongoing UKIRT survey will enable the creation of extinction and reddening maps of the Galactic center and bulge. These, combined with deep optical surveys, will allow us to confirm and fully constrain these variations. These results can also have important implications for observational cosmology, as they suggest that, if non-standard extinction laws occur in external galaxies, they may lead to systematic errors in the distances derived from studies of SNe Ia and Cepheids, if those studies adopt standard extinction laws.

We would like to dedicate this paper to the memory of Neil Gehrels, whose support made the 2017 UKIRT survey possible. We thank M. Irwin for useful discussions, D. Imel for assistance with developing cloud processing for the light curves, and the anonymous referee for important comments that helped to improve the manuscript. Work by Y.S. was supported by an appointment to the NASA Postdoctoral Program at the Jet Propulsion Laboratory, administered by Universities Space Research Association through a contract with NASA. D.M.N. was supported by the Allan C. and Dorothy H. Davis Fellowship. This work was partially supported by NASA grants NNX17AD73G and NNG16PJ32C. UKIRT is owned by the University of Hawaii (UH) and operated by the UH Institute for Astronomy; when some of the data reported here were acquired, UKIRT was supported by NASA and operated under an agreement among the University of Hawaii, the University of Arizona, and Lockheed Martin Advanced Technology Center; operations were enabled through the cooperation of the East Asian Observatory. Based on data products from observations made with ESO Telescopes at the La Silla Paranal Observatory under programme ID 179.B-2002.