Surprise non-detection of Upsilon Andromedae b with MIRC-X and MYSTIC at the CHARA Array

TL;DR

This study investigates direct detection of the non-transiting hot Jupiter Ups And b using CHARA Array interferometry with MIRC-X and MYSTIC. By self-calibrating closure-phase data and performing a grid search over orbital parameters, the authors report a very tentative H-band detection in 2019 that is not corroborated by subsequent K-band or joint datasets. Injection tests show that contrasts at the 1–2×10^-4 level can be recovered, implying the non-detection is surprising given model predictions and highlighting remaining instrument systematics, particularly polarization and dispersion effects. The work demonstrates the promise of Project PRIME for atmosphere characterization of close-in companions and outlines future improvements needed to realize robust detections with long-baseline interferometry.

Abstract

Ground-based long baseline interferometry is a powerful tool for characterizing exoplanets which are too close to their host star to be imaged with single-dish telescopes. The CHARA Array can resolve companions down to 0.5 milli-arcseconds, allowing us in principle to directly measure the near-infrared spectra of non-transiting "Hot Jupiter" exoplanets. We present data taken with the MIRC-X and MYSTIC instruments at the CHARA Array on the Hot Jupiter Upsilon Andromedae b. By resolving the star-planet system, we attempt to directly detect the flux from the planet. We describe our self-calibration methods for modeling systematics in the closure phase data, which allows us to reach sub-degree precision. Through combining multiple nights of data across two MIRC-X runs in 2019 and 2021, we achieved a very tentative detection of Ups And b in the H-band at a planet/star contrast of 2-3 x 10^-4. Unfortunately, we cannot confirm this detection with 2021 MYSTIC data in the K-band, or in a 2023 joint MIRC-X and MYSTIC dataset. We run updated global circulation models and create post-processed spectra for this planet, and report the resulting model spectra in H- and K-bands as a function of orbital phase. We then run planetary injection tests to explore H/K-band contrast limits, and find that we can confidently recover planets down to a planet/star contrast of 1-2 x 10^-4. We show that we are probing contrasts fainter than predicted by the model, making our non-detection surprising. We discuss prospects for the future in using this method to characterize companions with interferometry.

Figures (9)

• Figure 1: We plot the MIRC-X (left) and MYSTIC (right) closure phase for one spectral channel across hour angle during the 2023Oct run. Each box represents one of the 20 closing triangles of the CHARA Array, with the designated telescope names labeled. Though we are already looking at sub-degree features, the drifts and dips need to be calibrated out in order to reach the precision needed to detect the flux from a planet. The black data points represent the closure phase data, while the red points show our model of systematic trends. The light blue crosses show the expected planet signal at a planet-to-star contrast ratio of 2e-4.
• Figure 2: We show our grid searches for Ups And b to MIRC-X data across the unknown orbital elements from RV: $\Omega$, inclination, and semi-major axis. In 2019 September we find a $\chi^2$ minimum at $\Omega = 49^{\circ}$, $i=25^{\circ}$ (top left). We show the inclination/semi-major axis grid search at the best value of $\Omega$, and find $a=4.5$ mas (bottom left). The value for inclination agrees remarkably well with the results of piskorz2017, and the semi-major axis is also consistent with previous measurements rosenthal2021 (Top center) In the 2021 October MIRC-X dataset our best solution is an inclined orbit near 90$^{\circ}$, which we know is not real due to the lack of transits. However, we also see a solution near $\Omega = 53^{\circ}$, though we show that this solution prefers an inclination of $i=130^{\circ}$ (bottom center). If we consider only prograde orbits between 0--80$^{\circ}$, we find the same solution as the 2019 September dataset (right panels). As can be seen from the multiple solutions across different runs, this detection is extremely tentative and requires follow-up data to confirm.
• Figure 3: Although we had a very tentative detection of Ups And b in independent 2019Sep and 2021Oct MIRC-X datasets, we do not see this same detection in MYSTIC data from the same 2021Oct run (left). (Center and right) we also do not detect the planet at this location in a follow-up 2023Oct run with either MIRC-X or MYSTIC.
• Figure 4: In order to test MYSTIC contrast limits, we inject planet signals of varying star/planet contrasts. Here we show a planet at an inclination of 156$^{\circ}$ and the same $\Omega$ as the MIRC-X detection. The planet signal is recovered down to a contrast of 0.7 times the value of the model planet flux. At fainter values than this, other peaks in the grid search become more prominent. This would suggest that, if the 2019/2021 MIRC-X detections were real, we should have recovered the planet again.
• Figure 5: Same as Figure \ref{['mystic_injection']}, but for 2023Oct MIRC-X data. Just as for MYSTIC, we recover the planet down to a factor of 0.7 times the model flux from the planet. Once we get to 0.5 times the model flux and lower, other solutions become more prominent in the grid searches.