No Clear, Direct Evidence for Multiple Protoplanets Orbiting LkCa 15: LkCa 15 bcd are Likely Inner Disk Signals

Young, 1–10 Myr old Jovian protoplanets embedded in disks around newly born stars provide a crucial link between the first stages of planet formation and the properties of directly imaged, fully formed planets orbiting 10–100 Myr old stars (e.g., Marois et al. 2008b, 2010b). LkCa 15, a solar-mass T Tauri star and member of the 1–3 Myr old Taurus–Auriga star-forming region (Kenyon et al. 2008), is a superb laboratory for studying planet formation and searching for protoplanets. The star is surrounded by an accreting, gas-rich protoplanetary disk with multiple dust components: hot (T_eff = 1400 K), sub-au scale dust producing broadband near-infrared (NIR) excess and cooler massive outer dust, which are separated by a solar system-scale cavity plausibly created by Jovian protoplanets (Espaillat et al. 2007; Thalmann et al. 2010; Andrews et al. 2011; Dodson-Robinson & Salyk 2011; Dong & Fung 2017; Alencar et al. 2018).

Using sparse aperture masking interferometry (SAM; Tuthill et al. 2006) of LkCa 15, Kraus & Ireland (2012) reported the detection of one protoplanet located within an ostensibly cleared gap in dust emission. Also using SAM, Sallum et al. (2015b) then identified three protoplanets within ρ ∼ 015 (≈25 au) (LkCa 15 bcd), one of which was recovered in H_α (LkCa 15 b). Thus, LkCa 15 appeared to show evidence for multiple Jovian protoplanets, the first such system ever reported.

However, the closure phase signals of disks in SAM data can mimic those of protoplanets (Cieza et al. 2013; Kraus et al. 2013). LkCa 15's circumstellar environment as seen in scattered light is complex, including a bright outer dust wall (Thalmann et al. 2010, 2014). Additionally, inner dust disk material is now resolved at optical wavelengths and NIR polarimetry out to LkCa 15 bcd-like separations (Oh et al. 2016b; Thalmann et al. 2016). Depending on this dust disk's brightness and spatial extent in (a) total intensity at (b) the longer wavelengths where LkCa 15 bcd were identified (2.2–3.8 μm), it could instead be the signal masquerading as these protoplanets. However, previous 2.2–3.8 μm total intensity data lack the image quality/sensitivity to probe these regions (Thalmann et al. 2014).

In this Letter, we use multi-epoch direct imaging observations of LkCa 15 obtained from the Subaru Coronagraphic Extreme Adaptive Optics (SCExAO) project coupled with the Coronagraphic High Angular Resolution Imaging Spectrograph (CHARIS) in the NIR (JHK/1.1–2.4 μm; Groff et al. 2015, 2017; Jovanovic et al. 2015) and Keck/NIRC2 in the thermal infrared (L_p/3.78 μm). These data provide the first direct imaging look at the same wavelengths and in the same locations where previous studies identified the LkCa 15 protoplanets (K, L_p), and thus offer the first decisive test of their existence.

2.1. SCExAO/CHARIS JHK Direct Imaging/Spectroscopy

2.2. Keck/NIRC2 L_p Direct Imaging

2.3. PSF Subtraction

Figure 1 shows the SCExAO/CHARIS NIR images in broadband (a median-combination of all channels) and in K band (top panels) and Keck/NIRC2 L_p images (bottom panels). All data easily resolve the forward-scattering side of the crescent-shaped outer dust disk wall (e.g., Thalmann et al. 2010, 2014). However, no data set reveals direct evidence for LkCa 15 bcd. Instead, all data resolve another crescent-shaped extended structure interior to the outer disk wall, consistent with the wall of an inner dust disk previously only seen in polarized light (Thalmann et al. 2015; Oh et al. 2016b).

Zoom In Zoom Out Reset image size

Inspection of individual CHARIS data cubes and NIRC2 images shows that this extended inner disk emission cannot be explained by residual speckle noise that is preserved when images are derotated and combined (for CHARIS and NIRC2) or wavelength-collapsed (for CHARIS). RDI-reduced images obtained using a range of principal components (for KLIP) or a range of SVD cutoffs (for A-LOCI) all recover the same structure. For CHARIS, the inner disk is visible in most individual channels, especially those covering the H and K passbands. Furthermore, ADI and ASDI-reduced images (2018 January CHARIS data and two of the three Keck/NIRC2 data sets) also show negative self-subtraction footprints of this inner disk.³⁴

We further confirmed that we could have detected LkCa 15 bcd-like planets in absence of disk emission. To empirically assess our sensitivity to point sources, we injected and attempted to recover model planets with an early L dwarf-like spectrum into our raw LkCa 15 data reduced with RDI (2017 September CHARIS data and 2017 December NIRC2 data). We considered the half-field of view opposite the peak brightness of the inner disk and at a range of angular separations.³⁵ We varied the brightnesses of these planets with respect to the star to be equal to or fainter than that for LkCa 15 bc at K and L_p as reported by Sallum et al. (2015b): ΔK ∼ 5.5–6 and ΔL_p ∼ 5–5.9.

Figure 2 shows example RDI-reduced SCExAO/CHARIS and Keck/NIRC2 data sets with injected planets. The planets' throughputs are high, ranging between 75% and 100%. In spite of some contamination from residual inner disk emission, planets with separations comparable to LkCa 15 bcd (ρ ∼ 009–01) are detected at LkCa 15 bcd-like contrasts (ΔK ∼ 5.75, 6.15; ΔL_p ∼ 5, 5.9) and visible as point sources. The contrasts of these recovered planets are similar to limits achieved for diskless stars with SAM in Kraus et al. (2011) and Lacour et al. (2011). Planets even fainter than proposed for LkCa 15 bcd—ΔK ∼ 6.5, ΔL_p ∼ 6.3—are detected and identified as point sources at ρ ∼ 009–01 in regions of the lowest disk emission (not shown). At wider separations (ρ ≳ 04, our contrast limits are equal to or deeper than achievable with SAM (ΔK, ΔL_p ∼ 10, 7.5).³⁶

Zoom In Zoom Out Reset image size

Comparisons between our images and SAM results strongly suggest that this inner disk emission is the same astrophysical source previously interpreted as the LkCa 15 bcd protoplanets. For both CHARIS and NIRC2 data, the inner disk emission extends from ρ ∼ 007 to ρ ∼ 025 (r_proj ≈ 10–40 au) with an apparent semimajor and semiminor axis for the emission's peak at ρ ∼ 02 and 01, respectively (r_proj ≈ 17–32 au). In the RDI-reduced data sets, the emission subtends an angle of ∼100°, which is roughly the same position angle range for LkCa 15 bcd reported in Sallum et al. (2015b). Planet positions reported in Sallum et al. (2015b; circles in the 2018 January CHARIS data) trace this emission. The aggregate flux density for LkCa 15 bcd from Sallum et al. (2015b) is ≈3.7 ± 1.2 mJy and 5.4 ± 1.5 in K and L_p, respectively. Over the same range of position angles/separations reported for LkCa 15 bcd, the summed inner disk flux densities in the CHARIS K band and NIRC2 L_p data reduced using RDI are the same, within uncertainties³⁷ : ≈2.8 and 3.9 mJy.

LkCa 15 images obtained at different wavelengths reveal some evidence for color differences between the spatially resolved inner and outer disk components. In the SCExAO/CHARIS broadband image (JHK, λ_o = 1.63 μm), the peak brightness of the inner component is about 30% higher than the peak of the outer component. At K-band (λ_o = 2.18 μm), the peak brightness of the inner disk is about 1.75 times than the outer disk, while at L_p the inner disk is more than twice as bright as the outer disk. The physical origin of these differences will be addressed in Section 4.2.

We now compare the LkCa 15 images to forward-models (Marois et al. 2010a) for LkCa 15 bcd and an inner disk. Our analysis adopts the approaches in Pueyo (2016) and Currie et al. (2018b) for KLIP and A-LOCI, using the eigenvalues/eigenvectors in KLIP or coefficients in A-LOCI drawn from the real data and applying them to synthetic planet/disk signals injected into empty data cubes/images. Our goal is to (1) confirm that the emission we interpret as an inner disk cannot be reproduced by properties previously attributed to LkCa 15 bcd, and (2) then explore the general properties of this inner disk.

We focused on the highest-quality data easily amenable to forward-modeling at wavelengths where LkCa 15 bcd were identified (K, L_p). Thus, we considered the K-band portion of the 2017 September SCExAO/CHARIS data processed with RDI/KLIP, the 2009 November NIRC2 L_p data processed with ADI/A-LOCI, and the 2017 December NIRC2 L_p data processed with RDI/A-LOCI.

4.1. Planet Forward-modeling

We produced forward-models of (a) all three planets (LkCa 15 bcd) and (b) just the two identified in Sallum et al. (2015b) from multiple epochs (LkCa 15 bc), (1) at the planets' last reported positions in Sallum et al. (2015b) in 2014 November–2015 February, and (2) at the planets' estimated positions in 2009 November, 2017 September, and 2017 December. To predict the planets' positions in multiple epochs, we adopted the Sallum et al. astrometry and the Gaia second data release (DR2) distance to LkCa 15 (158.9 pc), assuming that the planets are on circular orbits in the same plane as the outer disk (i ∼ 50°, PA_minor ∼ 60°; Thalmann et al. 2014, 2015; Oh et al. 2016b). Their deprojected orbital separations in 2014 November–2015 February are ∼16–18 au; their position angles change by ≈5° yr⁻¹ in the orbital plane.

We adopted the Sallum et al. (2015b) L_p photometry for LkCa 15 bcd. In K, we also adopted their LkCa 15 bc photometry. LkCa 15 d has no claimed detection in K from Sallum et al. (2015b). We assumed that LkCa 15 d's K-L_p colors are similar to LkCa 15 bc's and thus adopted ΔK = 7.

Figure 3 shows forward-models of the LkCa 15 planets for CHARIS K-band (top panels) and NIRC2 L_p (bottom panels). The emission's apparent brightness in the CHARIS data is comparable to the combined brightness proposed for LkCa 15 bcd. However, LkCa 15 bcd would be clearly distinguishable as separate point sources in K, whereas the CHARIS data instead show a continuous structure. Thus, the SCExAO/CHARIS data are inconsistent with planets being responsible for this emission.

Zoom In Zoom Out Reset image size

At L_p, LkCa 15 bcd's PSFs are partially blended.³⁸ However, due to orbital motion, the aggregate emission from LkCa 15 bc(d) should rotate clockwise by ∼35°–40° between 2009 and 2017: the emission centroid, measured in the forward-modeled planet images from regions within 50% of the peak intensity, changes by ≈1 λ/D. In contrast, the measured center of mass for this emission in the real 2009 and 2017 data is constant to within 0.05–0.1 λ/D, implying a static morphology over 8 yr. Thus, the Keck/NIRC2 data are inconsistent with planetary orbital motion.

4.2. Disk Forward-modeling

To explore the general properties of inner disk emission previously attributed to LkCa 15 bcd, we produced and then forward-modeled synthetic scattered-light disk images with SCExAO/CHARIS using the MCMax3D radiative transfer code (Min et al. 2009), adopting the formalism from Mulders et al. (2010, 2013). Our approach considered three spatially extended components: (1) an optically thick (sub-)au scale hot component responsible for the NIR broadband excess and 10 μm silicate feature, (2) a warm component responsible for the inner disk resolved with SCExAO/CHARIS and Keck/NIRC2, and (3) the optically thick outer disk, which has been resolved in optical/NIR scattered light (e.g., Thalmann et al. 2014, 2016) and with (sub-)millimeter data (Andrews et al. 2011; Isella et al. 2014). Following Thalmann et al. (2014, 2016), we envisioned that components 1 and 2 shadow and may be slightly misaligned with the outer disk (component 3). We explored a small range of component parameters, settling on a fiducial model with properties listed in Table 1. Except for a few Spitzer/IRS channels probing the unresolved sub-au component, the model fits LkCa 15's entire SED from the optical to millimeter to within ∼20%–30% (Andrews et al. 2011; Isella et al. 2014; Ribas et al. 2017).

Table 1.  Disk Model Parameters

Parameter	Value
Global Parameters
Distance	158.9 pc
Teff	4730 K
L⋆	1.2 L⊙
R⋆	1.65 R⊙
M⋆	1.01 M⊙
AV	1.7
Disk position angle (θ)	60°
Dust size power law, pa	3.5
Dust carbon fraction	0.1
Component Parameters	Component 1	Component 2	Component 3
Disk inclination (i)	50°	515	50°
Inner radius, Rin (au)	0.12	20	55
Outer radius, Rout (au)	3	40	160
Disk wall radius, Rw (au)	0.12	25	82.5
Wall shape (w)	flat/vertical	rounded/0.3	rounded/0.25
Mdust (M⊙)	5 × 10−8	7.25 × 10−6	1.4 × 10−3
Radial surface density power law ()	1	0.5	1
Minimum dust size (amin, μm)	0.1	0.6	0.1
Maximum dust size (amax, μm)	0.25	1000	1000
Scale height at inner radius, Ho,in	0.05	0.08	0.05
Scale height power law, pgas	1.15	1.25	1.15

Note. The disk component surface density follows Σ (R < R_w) ∝ ${R}^{-\epsilon }$  ×  exp$\left(-{\left(\tfrac{1-R/{R}_{\exp }}{w}\right)}^{3}\right)$ and Σ (R ≥ R_w) ∝ ${R}^{-\epsilon }$. The wall shape parameter defines the spatial scale over which the disk surface density increases from R_in to R_w. See Mulders et al. (2010, 2013) and Thalmann et al. (2014) for detailed explanations of MCMax3D terminology.

Download table as:  ASCIITypeset image

Figure 4 compares the SCExAO/CHARIS broadband image (top) and Keck/NIRC2 L_p image with our fiducial model. The PSF-subtracted model reproduces the brightness and morphology of the inner/outer disk components: the subtraction residuals do not reveal any emission consistent with LkCa 15 bcd. The peak pixel intensity at positions covering LkCa 15 bcd (circled) is always less than 1/3 (1/4) that predicted for LkCa 15 b(c). Residuals at ρ ≲ 02 that do remain are attributable to slight mismatches with extended disk emission.³⁹

Zoom In Zoom Out Reset image size

While a wide range of models match either the SCExAO/CHARIS or Keck/NIRC2 data, the combined data point toward different grain properties for the three disk components. A larger minimum dust grain size for the resolved inner disk versus resolved outer disk (∼0.6 μm versus 0.1 μm) better reproduces the inner disk's redder color and more pronounced forward-scattering peak. While unresolved, the sub-au disk component requires submicron-sized grains to reproduce the 10 μm silicate feature (see also Espaillat et al. 2007). A future paper will thoroughly analyze LkCa 15's disk structures and derive best-fit parameters.

Our modeling also (a) implies that LkCa 15's disk structures should be detectable in optical total intensity imaging and (b) is consistent with the millimeter detection of the outer disk and non-detection of the inner disk. At 0.65 μm, the inner disk's continuum signal compared to the star (convolved with a Gaussian and integrated within 1.5–2 FWHM) near the reported LkCa 15 b position in H_α is just slightly lower than LkCa 15 b's reported H_α contrast (ΔF ∼ (2.5–5) × 10⁻³), as is the forward-scattering peak of the outer disk (ρ ∼ 02). The predicted signal of the forward-scattering peak of the inner disk (ρ ∼ 008) is comparable in contrast to LkCa 15 b ((5–8.5) × 10⁻³). At 7 mm, the model reproduces the outer disk edge's typical intensity, with a characteristic brightness of ∼24 μJy beam⁻¹ for a beam size of 015; for a 007 beam, it accurately predicts that the inner disk (0.5 μJy beam⁻¹) would be undetected given a 1σ noise floor of 3.6 μJy beam⁻¹.

Instead of protoplanets, our direct images of LkCa 15 obtained with SCExAO/CHARIS show extended, unresolved inner disk emission. Forward-modeling shows that the SCExAO data were capable of distinguishing between disk emission point sources with K band photometry and astrometry reported for LkCa 15's planets by Sallum et al. (2015b). While Kraus & Ireland (2012) identify concentrated emission sources in SAM data, they use a binary (LkCa 15 A+ companions) light distribution model for image reconstruction, which is valid only if the brightness distribution resembles point sources. Our data show that it does not.

On the other hand, the inner disk signal is comparable to the total flux density reported for LkCa 15 bcd from Sallum et al. (2015b) at K and L_p. Thus, we emphasize that the Sallum et al. (2015b) SAM data likely detected the inner disk at multiple wavelengths. Furthermore, the gaps and misalignments between LkCa 15's resolved disk structures, as well as a warp inferred from the sub-au component (Alencar et al. 2018), may be evidence for unseen Jovian planets (Dong & Fung 2017), which could be detected with future facilities (e.g., the Thirty Meter Telescope; Skidmore et al. 2015).

Our Keck/NIRC2 L_p data obtained between 2009 and 2017 reveal this emission to be static. Based on SAM data taken over a shorter timescale, Sallum et al. (2015b, 2016) argued that LkCa 15 bcd astrometry reveals evidence for orbital motion, although different components are detected in different epochs and the combined astrometry appears consistent with stationary sources given large error bars. While the evaluation of our data is straightforward, several factors may complicate this aspect of SAM data interpretation for LkCa 15. For example, variable u − v coverage between epochs can induce apparent astrometric offsets when a binary model is assumed in the image reconstruction process (C. Caceres 2019, in preparation). Instead of bare stellar photospheres, the calibrators used for LkCa 15 in Sallum et al. (2015b) and especially Kraus & Ireland (2012) include multiple stars with bright resolved disk emission on the same spatial scale as LkCa 15's disk: some are also highly variable (e.g., GM Aur, UX Tau; Tanii et al. 2012; Oh et al. 2016a).

Another common argument is that LkCa 15 bcd are too red to be consistent with scattered-light disk emission (Kraus & Ireland 2012; Ireland & Kraus 2014). However, for a system with a pre-transitional disk structure like LkCa 15, (a) scattering can be extremely red because (b) the sub-au dust component contributes significantly to the NIR broadband flux and intercepts (and then re-emits) a significant fraction of the starlight (Mulders et al. 2013; Currie et al. 2017b). The light that LkCa 15's 20 au scale disk "sees" is then far redder than the star. Indeed, our fiducial disk model successfully reproduces the brightness of the inner dust disk at K and L_p. While Sallum et al. (2015b) argued that a disk cannot explain LkCa 15 bc(d) in current SAM data, they use a very simple inclined disk model, not a radiative transfer model. Additionally, from inspection of their Figure 8, the inner component of this model appears to have semimajor and semiminor axes of ∼008 and ∼005, which are inconsistent with the larger, spatially resolved, and extended disk as resolved at K and L_p in this study (02 and 01).

Our analyses do not directly refute the claimed single-epoch MagAO H_α detection for LkCa 15 b, which technically remains a candidate companion. However, they help strengthen arguments voicing strong skepticism. As LkCa 15 A itself is bright in H_α due to accretion its disk structures should have an elevated H_α luminosity. Mendigutia et al. (2018) recently found that LkCa 15's spectroastrometric signature at H_α is inconsistent with that of a planet but consistent with a disk. They rule out H_α emission from a LkCa 15 b unless the candidate has an H_α contrast fainter than 5.5 mag or a continuum contrast brighter than 6 mag: the H_α photometry and continuum upper limits from Sallum et al. (2015b) are just barely consistent with these spectroastrometric limits. Their predicted emitting region for H_α is ρ ∼ 007–016, consistent with our resolved images of LkCa 15's inner disk.

Furthermore, SPHERE/ZIMPOL data (Thalmann et al. 2015) and our modeling show that both the inner disk and outer disks are bright, modest-contrast structures and should be detectable at optical wavelengths covering the MagAO H_α observations. Yet Sallum et al. (2015b) did not report a disk detection with MagAO, implying that their H_α planet detection may instead be spurious or a misidentified, partially subtracted piece of the H_α-bright disk. Their quoted position for LkCa 15 b in H_α is conspicuously close to the inner disk's major axis. Given the MagAO observations' poor field rotation (1.5 λ/D at 01) and negligibly small rotation gap (5° or ∼0.12 λ/D at 01), any inclined disk at a comparable separation will suffer severe self-subtraction: its residual emission near the major axis would be preferentially preserved and appear point-like.

Forward-modeling of both  a planet and a disk through the MagAO data—as performed to assess HD 100546c (Currie et al. 2015)—could determine which signal better reproduces the images. However, this test is absent from the Sallum et al. (2015b) analysis. The MagAO H_α data are proprietary, not public, preventing any independent verification that the planet hypothesis is preferred. The public availability of archival Keck/NIRC2 data presented here was crucial in assessing evidence for planets orbiting LkCa 15 from aperture masking.

In summary, we rule out the proposed LkCa 15 bcd protoplanets as being primarily responsible for emission seen at small angles in SAM data because the emission (a) would be resolved as separate point sources in the SCExAO data (when it is not) and (b) would rotate between 2009 and 2017 Keck/NIRC2 data due to the planets' orbital motion (which it does not). Our results also strengthen the argument from Mendigutia et al. (2018) that H_α data also likely identifies a disk, not LkCa 15 b.

Thus, there is currently no clear, direct evidence for multiple protoplanets orbiting LkCa 15. While the system shows indirect evidence for at least one unseen Jovian planet, the bright inner dust disk impedes the detection of this companion(s). To confirm Jovian companions around LkCa 15 from future observations, the inner disk should be resolved and its effect modeled, removed, and shown to be distinguishable from planets. Protoplanet candidates identified from similar systems should likewise be clearly distinguished from disk emission through multi-wavelength and/or multi-epoch modeling (e.g., Keppler et al. 2018).

Distinguishing between disk emission and bona fide protoplanets will continue to be a key challenge for the field of direct imaging (e.g., Cieza et al. 2013; Kraus et al. 2013; Sallum et al. 2015a; Ligi et al. 2018; Christiaens et al. 2019; Rich et al. 2019; this work).

We thank Michiel Min for graciously sharing the MCMax3D code and the Subaru and NASA/Keck Time Allocation Committees for their generous support. We thank the anonymous referee for a careful, thoughtful review. Laurent Pueyo, Jun Hashimoto, Christian Thalmann, Catherine Espaillat, Nienke van der Marel, Hannah Jang-Condell, Geoff Bower, and Scott Kenyon provided helpful comments and/or additional, independent assessments of this manuscript. We emphasize the pivotal cultural role and reverence that the summit of Maunakea has always had within the Hawaiian community. We are most fortunate to conduct scientific observations from this mountain. T.C. was supported by a NASA Senior Postdoctoral Fellowship and NASA/Keck grant LK-2663-948181; L.C. was supported by CONICYT-FONDECYT grant No. 1171246. C.C. acknowledges support from project CONICYT PAI/Concurso Nacional Insercion en la Academia, convocatoria 2015, folio 79150049. M.T. is supported by JSPS KAKENHI grant Nos. 18H05442 and 15H02063. This work utilized the Keck Observatory Archive (KOA), which is operated by the W. M. Keck Observatory and the NASA Exoplanet Science Institute (NExScI), under contract with the National Aeronautics and Space Administration.