HD 219134 Revisited: Planet d Transit Upper Limit and Planet f Transit Nondetection with ASTERIA and TESS

HD 219134 is a bright, nearby K3V star reported to host six radial-velocity discovered exoplanets (Motalebi et al. 2015; Vogt et al. 2015; Johnson et al. 2016; Gillon et al. 2017). The two innermost planets HD 219134 b and c, with periods of 3.1 and 6.8 days respectively, were found to transit with Spitzer (Motalebi et al. 2015; Gillon et al. 2017). For planetary system parameters see Table 1 and Figure 1. HD 219134 is the nearest and brightest star presently known to host transiting planets (6.53 pc, Gaia Collaboration et al. 2018; Stassun et al. 2018; and V = 5.57, Oja 1993), making HD 219134 b and c eminently well suited for follow-up atmospheric characterization. HD 219134, also known as TOI 1469, has a stellar radius of $0.778\pm 0.005{R}_{\odot }$ (Boyajian et al. 2012) and stellar mass of $0.81\pm 0.03{M}_{\odot }$ (Gillon et al. 2017), with R.A. = 23:13:16.780 and decl. = +57:10:10.05 (J2015.5). For other stellar parameters see the Appendix.

Zoom In Zoom Out Reset image size

Table 1.  Parameters for HD 219134 Planets b, c, f, and d

Planet	P (d)	a (au)	M (${M}_{\oplus }$)	R (${R}_{\oplus }$)	e
b	3.092926 ± 0.000010	0.03876 ± 0.00047	4.74 ± 0.19	1.602 ± 0.055	0 (fixed)
c	6.76458 ± 0.00033	0.06530 ± 0.00080	4.36 ± 0.22	1.511 ± 0.047	0.062 ± 0.039
f	22.717 ± 0.015	0.1463 ± 0.0018	>7.30 ± 0.40	⋯	0.148 ± 0.047
d	46.859 ± 0.028	0.2370 ± 0.0030	>16.17 ± 0.64	⋯	0.138 ± 0.025

Note. P is period in days, a is semimajor axis in astronomical units, M_⊕ is planet mass in Earth masses, R_⊕ is planet radius in Earth radii, and e is orbital eccentricity. Planet masses for f and d are minimum masses, hence the > sign. Planets f and d (and g, not shown) are less firm than planets b and c due to their orbital periods' proximity to the stellar rotation period (45.6 days) or its harmonics (Johnson et al. 2016; Folsom et al. 2018). Data from Gillon et al. (2017). For updated P for planets b and c, see Table 3.

Download table as:  ASCIITypeset image

The HD 219143 planetary system has promise to be the nearest, brightest example of a compact multiple-planet transiting system—over 100 times closer and more than 200 times brighter than the canonical Kepler 11 system (Lissauer et al. 2011). Indeed, even with periods as long as 22.717 and 46.859 days for planets f and d respectively, the planets still have a fairly high transit probability owing to the transiting geometry of planets b and c. Assuming coplanarity of planets f and d with planets b and c, the a posterior transit probabilities of planets f and d are 13.1% and 8.1%, respectively (Gillon et al. 2017).

The detections of planets HD 219134 f, d, and g are less firm than planets b and c, owing to the proximity of their orbital periods to the stellar rotation period (45.6 days) or its harmonics (Johnson et al. 2016; Folsom et al. 2018). While it is not unexpected to see a 2:1 period ratio between planets (e.g., Winn & Fabrycky 2015), it is concerning that the periods of planets f, d, and g happen to be at 1/2, 1, and 2 times the stellar rotation period.

A ground-based search for transits of planets HD 219143 f and d is impractical because of the low frequency and long duration of the transits and the ephemeris uncertainty (see parameters in Table 2). The long periods (approximately 23 and 47 days) mean that the number of transit events per observing season is small, making observable transit events from a single ground-based location rare. The combination of the relatively long transit duration (3.8 hr for planet f and 5.6 hr for planet d) with the uncertainty on the predicted transit time (on the order of half a day, longer than the anticipated transit duration itself) means that it is both difficult and unlikely to catch an entire transit from a single ground-based location. Catching only the ingress or egress makes it difficult to be confident in the detection of the flux decrease or increase. The issues are exacerbated by the shallow transits expected to be roughly 0.1% deep. This small transit depth is at or beyond the typical capabilities of 1 m class ground-based telescopes, the telescopes that are commonly being used for this purpose, even under ideal conditions such that the HD 219134 f and d transits are expected to be shallow with amplitude similar to systematic noise signals that are known to plague ground-based light curves. We note that despite a shallow transit depth of 0.13% ± 0.02%, NGTS-4b was discovered with a single 20 cm aperture telescope, enabled by a combined 28 transits for the 1.34 day orbit with a 272 night baseline (West et al. 2019).

To search for transits of HD 219134 d and f we used two complementary space-based photometric data sets. The data set from the Arcsecond Space Telescope Enabling Research in Astrophysics (ASTERIA; Smith et al. 2018) has targeted observations with extensive coverage of the transit windows, while the data set from the Transiting Exoplanet Survey Satellite (TESS; Ricker et al. 2015) has better photometric precision.

ASTERIA was a small space telescope (10 cm × 20 cm ×30 cm, 12 kg, a "6U CubeSat") deployed from the International Space Station in 2017 November into a low-Earth orbit. The spacecraft was operational until 2019 December and re-entered Earth's atmosphere in late 2020 April. ASTERIA's refractive camera had a 6 cm aperture, 85 mm focal length, a bandpass from 500 nm to 900 nm, a plate scale of 158 per pixel, and a field of view of 112 x 96.

ASTERIA was a technology demonstration mission to prove precise pointing and detector thermal control are possible in a small package, with an eye toward enabling high photometric precision. ASTERIA successfully met its requirements by achieving the following performance goals: reaching line-of-sight pointing stability of approximately 05 rms over 20-minute observations, pointing repeatability of 1 mas rms from one observation to the next, and focal plane temperature stability better than ±0.01 K over 20-minute observations (Pong 2018; Smith et al. 2018). Developed at the Jet Propulsion Laboratory (JPL) in collaboration with the Massachusetts Institute of Technology (MIT), ASTERIA traces its roots back to the ExoplanetSat mission concept developed by MIT and Draper Laboratory (e.g., Smith et al. 2010; Pong et al. 2010). For details of the ASTERIA development, flight system, ground system, on-orbit operation, and on-orbit performance see Smith et al. (2018) and Knapp et al. (2020).

ASTERIA used a complementary metal-oxide semiconductor (CMOS) array detector instead of the traditional charge-coupled device (CCD) detector. We chose a CMOS detector primarily for its fast read out (20 Hz; necessary for the fine pointing control algorithm), its ability to operate at room temperature, and its relative insensitivity to radiation damage. CMOS detectors can be operated at room temperature, as they generate ultra-low dark current (20 e-/pixel/s) compared to 250–500 e-/pixel/s generated by CCDs at the same temperature. Hence, CMOS detectors do not require passive or active cooling systems that would otherwise add significantly to the mass and size of the spacecraft. Despite these distinct advantages, CCDs have traditionally dominated space-based imaging and spectroscopy due to their sensitivity via high quantum efficiency.

ASTERIA was not designed to deliver precise photometry, even though its underlying precise pointing and thermal control are enabling technologies for highly precise photometry. Nonetheless, ASTERIA was able to detect the transit of 55 Cnc e, a planet nearly twice the radius of Earth transiting a Sun-sized star with a transit depth of approximately 400 ppm (Knapp et al. 2020). After completion of its 90 day prime mission, ASTERIA was in part used for opportunistic science, including observations of the bright nearby stars Algol, Alpha Centauri A and B, 55 Cancri, and HD 219134.

TESS is an MIT-led, NASA Medium-class Explorer Mission that launched in 2018 April (Ricker et al. 2015). TESS carries four identical, specialized wide-field CCD cameras, each covering 24° x 24° on the sky with a 10 cm aperture, a red bandpass of 600–1000 nm, and plate scale of 21'' per pixel. TESS is in an innovative high-Earth lunar resonance orbit (Gangestad et al. 2013) and a highly inclined, highly elliptical orbit in resonance with the moon, giving TESS a relatively unobstructed, continuous orbit night observations in a thermally stable environment.

During its two-year prime mission, TESS surveyed  70% of the sky, by monitoring each 24° x 96° strip of the sky (called a Sector) for two orbits, or 27.4 days. TESS surveyed the southern ecliptic hemisphere in the first year of its prime mission and will complete the northern ecliptic hemisphere during its second and final year of its prime mission in 2020 June.

TESS is anticipated to find thousands of exoplanets during its prime mission (Sullivan et al. 2015; Barclay et al. 2018; Huang et al. 2018), most with orbital periods less than one month. Although TESS will not be able to discover true Earth analogs (that is, Earth-sized exoplanets in 365 day period orbits about Sun-like stars), TESS will be capable of finding Earth-sized and super-Earth-sized exoplanets (where super-Earths are loosely defined as planets with radii up to 1.7 times Earth's radius) transiting M-dwarf stars, stars that are significantly smaller and cooler than our Sun. One of the TESS Level One mission requirements is to deliver 50 small (${R}_{p}\lt 4{R}_{\oplus }$) planets with measured masses to the community.

During TESS' survey of the sky, TESS naturally monitors many known transiting planet-hosting stars, including HD 219134.

In this paper we proceed with a description of our observations and photometric methods in Section 2, present our transit nondetection results and discussion in Section 3, and conclude with a summary in Section 4.

ASTERIA orbited Earth approximately once every 90 minutes, with orbit night typically lasting between 20 and 35 minutes. ASTERIA only observed during orbit night, therefore observations of HD 219134 were not continuous. Furthermore, because of its small aperture, data from different observing sequences must be phase-folded and binned in order to reach the photometric precision needed to detect shallow transits. ASTERIA's observational constraints (e.g., Sun, Moon, and Earth limb exclusion angles) and operational constraints (e.g., spacecraft related factors) are described in detail in Knapp et al. (2020).

We scheduled ASTERIA observations of HD 219134 (see the Appendix) to target the predicted transit windows of planets f and d (Table 2 with values from Gillon et al. 2017). The transit window predictions take into account uncertainty in the midtransit time T₀ and other parameters fit in Gillon et al. (2017). The orbital periods of planets f and d are close to 2:1 (see Table 1), so their predicted transits occur close to one another. The first targeted campaign (so08 through so12, where "so" means science observation) spanned the predicted transit of HD 219134 f (midpoint 2018 March 15) and HD 219134 d (midpoint 2018 March 17). The second campaign (ex11, so18, where "ex" means experimental observation) targeted planet f only (midpoint 2018 April 7). Observations during this campaign were affected by unfavorable geometry between the spacecraft and the star, resulting in short duration observations. The third campaign (so21, so22) targeted the f transit primarily, with some overlap with the beginning of the d transit window. The fourth campaign (so33 through so35) targeted both planets' transit windows (2018 June). Campaign five (2018 July, so37, so38) targeted the transit window of HD 219134 f. The final two campaigns in 2018 August and September (so39 through so48) were focused on covering the full 3σ transit window for planet d, with incidental coverage of planet f transit windows. Other opportunistic observations were carried out when science time was available and spacecraft viewing geometry with HD 219134 was favorable.

For a visual representation of the HD 219134 observing sequences of ASTERIA see Figure 2. For transit coverage of HD 219134 in the form of data density per planet phase see the Appendix.

Zoom In Zoom Out Reset image size

Here we provide a summary of our data reduction pipeline, with full details described in Knapp et al. (2020).

2.2.1. ASTERIA CMOS Detector and Calibration Frames

ASTERIA was the first space telescope to attempt science using CMOS detectors. The main difference between CMOS and CCD detectors for precise photometry is that the CMOS readout is parallelized column-wise whereas the pixels of a CCD detector are read out at a single point. For the CMOS detector, each pixel has a different amplifier and each column has a different analog-to-digital converter (ADC). Gains and offsets are therefore column-dependent. The systematics for the ASTERIA CMOS are not as spatially uniform compared to a CCD, because the readout electronics vary column to column.

Due to ASTERIA's status as a technology demonstration mission, there was no requirement or opportunity to characterize the detector in-depth, take calibration data, or validate the photometric performance prior to launch. This affected our data reduction pipeline because we have no ground-based dark frames or flat frames from pre-flight laboratory testing. ASTERIA's camera had no shutter so we could not obtain space-based darks, although dark frames are ultimately not needed for ASTERIA data reduction due to the short 50 ms integration time. ASTERIA did acquire approximated flat frames in space by purposely pointing the telescope in such a way that stray sunlight illuminated the detector. See Knapp et al. (2020) for more details.

Bias was removed from light images column-wise through the use of calibration windows collected contemporaneously with light windows. The calibration window was placed at the bottom edge of the detector where there are both electrically dark pixels (electrically tied to the ground so that photoelectric charge cannot accumulate) and separate optically dark pixels (blocked from light exposure) on the same columns as the light window. We used the electrically dark pixels for the bias correction. See Figure 3, where HD 219134 is centered in window 1 and window 8 is the calibration window.

Zoom In Zoom Out Reset image size

2.2.2. ASTERIA Data Reduction

Windowed images from ASTERIA observations were downlinked to the operations center at JPL and then processed into FITS files using custom software. The image data included contemporaneous calibration windows chosen from the edge of the detector (Figure 3). Metadata, including spacecraft housekeeping telemetry, were downlinked separately from the images and then included in the FITS header. Each FITS image contains a 1-minute integration that resulted from the onboard coadding of 1200 individual very short exposure (50 ms) frames. Individual frames were limited to 50 ms in order to support ASTERIA's 20 Hz fine pointing control loop (Pong 2018).

We began the data reduction process by reviewing every frame from the science observations. We manually flagged and discarded frames that were obtained during orbital sunrise or sunset. These frames appeared either at the beginning of an orbit or toward the end of the orbit, and displayed significant background contamination.

After discarding the sunrise/sunset frames, we proceeded to calibrate. We started with bias correction by subtracting the median of the bias column (from the calibration window) from all pixels in the respective column in the science frame. We then selected a few pixels (rows 0–15) in the neighborhood of the target that were devoid of any stellar flux to remove background and/or sky noise. We took the median of the pixels for each column, and subtracted the median value from every pixel in that column. Next, we subtracted the median of the bias rows column by column from every flat frame, and took the median of the reduced flat frame. Lastly, we performed a pixel-by-pixel division of the target star pixels with the reduced flat-frame pixel values. Note that the flat frame used to calibrate HD 219134 data was generated in space using stray light from the Sun to uniformly illuminate the detector as explained earlier (and see Knapp et al. 2020). A representative raw and calibrated image set is shown in Figure 4. The image of the target star is much cleaner than the raw image, although there is still some residual column-dependent noise. For more details on data reduction see Knapp et al. (2020) and Krishnamurthy (2020).

Zoom In Zoom Out Reset image size

2.2.3. ASTERIA Aperture Photometry and Detrending

2.2.4. ASTERIA Transit Injection and Recovery Tests

TESS observed HD 219134 in camera 3 during Sector 17, from 2019 October 7 to 2019 November 2 and in camera 4 during Sector 24. HD 2190134 has the TESS Input Catalog number TIC 283722336 (Stassun et al. 2018).

The 3σ transit time window of HD 219134 f falls within the TESS observations, based on the ephemeris from Gillon et al. (2017). In contrast, 3σ transit window of HD 219134 d falls outside of the TESS observations.

TESS data calibration, reduction, detrending, and light-curve fitting is covered in a large number of papers (e.g., Huang et al. 2018; Vanderburg et al. 2019; Dragomir et al. 2019; Huang et al. 2020). Here we provide a summary of our process.

2.4.1. TESS Data Reduction and Aperture Photometry

We used the calibrated TESS two-minute cadence data produced by the Science Processing Operations Center (SPOC) pipeline based at NASA Ames Research Center (Jenkins 2015; Jenkins et al. 2016).

Because of its considerable brightness (T_mag ∼ 4.6), HD 219134 saturated on the TESS CCDs. We chose to optimize the photometry based on our past experience working with TESS bright stars (Huang et al. 2018; Vanderburg et al. 2019). We re-extracted the photometry using a series of irregular shaped apertures placed on the SPOC-calibrated target pixel stamp. We chose the aperture that minimized the scatter in the out-of-transit portion of the light curve (Figure 5). We next used the SPOC quality flags to eliminate low-quality data points.

Zoom In Zoom Out Reset image size

For bright stars, the TESS systematics are mainly due to the pointing jitter of the spacecraft that occur on timescales shorter than the exposure time. We tried to remove such pointing systematics using the standard deviation of quaternion measurements within an exposure (Vanderburg et al. 2019), the first seven SPOC Presearch Data Conditioning (PDC) cotrending basis vectors¹⁸ , and also a trigonometric polynomial for longer-term trends. We used a least-squares method to optimize the detrending coefficients by minimizing the out-of-transit scatter of the detrended light curve (for details on the the SPOC PDC see Smith et al. 2012 and Stumpe et al. 2012, 2014).

2.4.2. TESS Transit Injection and Recovery Tests

We evaluate the range of transit times of planets f and d, considering TTVs of mutual gravitational interactions of the four planets (b, c, f, and d). We used analytic formulae that are accurate to the first order in mass ratios and eccentricity, from Agol & Deck (2016). We verified the validity of the transit times from our analytic approach by using an N-body integrator REBOUND (Rein & Liu 2012), but opted to use the faster analytic approach for our results for efficiency.

We ran a Monte Carlo analysis of the expected TTV amplitudes using a distribution of planetary parameters of the planets in the HD 219134 system to analyze the maximum expected shift in observed time of transit due to TTVs. For this purpose, we used the median values and credible intervals from the global MCMC fit of results from Gillon et al. (2017) to reconstruct Gaussian posteriors for the orbital parameters of planets HD 219134 b, c, f, and d. This combined fit used HARPS-N data and three Spitzer transits (two of planet b and one of planet c) as input. We performed 10,000 draws from these distributions, preserving covariances, and simulated the expected TTV amplitudes for each draw (Figure 6). We used the resulting distributions of TTVs as inputs to the injection and recovery tests in the ASTERIA data, using different T₀ and period offsets at each trial. The conclusions from our TTV analysis are discussed in Section 3.3.

Zoom In Zoom Out Reset image size

The ASTERIA data are sensitive to planets as small as 3.7 R_⊕ for HD 219134 f and 3.6 R_⊕ for HD 219134 d at the 5σ threshhold. These results are based on our injection and recovery completeness tests (see Section 2.2.4). For comparison, known exoplanets with masses within 10% that of HD 219134 d's radial-velocity measured minimum mass of 16.2 M_⊕, empirically fall into a radius range¹⁹ from 1.7 to 7.0 R_⊕. Uranus at 14.54 M_⊕ has a radius of 4.007 R_⊕ and Neptune at 17.15 M_⊕ has a radius of 3.883 R_⊕. For HD 219134 f and d, ASTERIA therefore can rule out planets larger than the established limits, and those limits are somewhat smaller than Uranus and Neptune.

We are unable to detect transits b and c in our ASTERIA data. HD 219134 planets b and c are small, 1.6 and 1.5 R_⊕ respectively (Table 1), with shallow transits of 358 ± 24 and 315 ± 19 ppm respectively, from Spitzer data (Gillon et al. 2017). We ran the injection and recovery completeness test on the ASTERIA data for the radial-velocity ephemeris of HD 219134 b and found that the injected planet was retrieved systematically at a radius >2.7 R_⊕, which is larger than the measured planet radius of 1.6 R_⊕. For planet HD 219134 c the injected planet was recovered at a radius >4.0 R_⊕, which is larger than the actual radius of 1.5 R_⊕.

The limiting factor in our transit nondetections is ASTERIA's residual correlated noise. Indeed ASTERIA reaches a photometric precision typically within a factor of 2 to 3 of the photon noise due to the residual correlated noise which we are neither able to attribute nor remove. The photon noise for HD 219134 is 350 ppm per minute, whereas our photometric precision for different observing nights ranges from 502 ppm per minute (1.4 times the photon noise) to 1038 ppm per minute (3.0 times the photon noise; Appendix). The overall ASTERIA HD 219134 data set has a median of 673 ppm per minute photometric precision. We show the ASTERIA HD 219134 detrended light curves as a time series and also phase-folded for each planet b, c, f, and d in Figure 7.

Zoom In Zoom Out Reset image size

The SPOC pipeline automatically detected the transits of HD 219134b and HD 219134c. The TESS Science Office designated the planets TOI 1469.01 and 1469.02 respectively. The known transits of HD 219134 b and c are clearly apparent in our detrended TESS two-minute cadence data (Figure 8).

Zoom In Zoom Out Reset image size

We derived updated ephemerides for planets HD 219134 b and c using the MCMC method as implemented by the allesfitter code (Günther & Daylan 2020). In particular, we ran our MCMC routine with 100 walkers, 20000 steps, and a burn-in of 2000. For each free parameter, we ensured convergence by requiring that the chain lengths are at least 50 times the samplers autocorrelation time (Foreman-Mackey et al. 2013). The median and 1σ uncertainties of the posterior distributions, including the updated orbital periods and midtransit times are presented in detail in Table A3, with a summary of the midtransit times and periods in Table 3. We present the corresponding transit b and c fits in the Appendix.

Table 3.  HD 219134 b and c Updated Midtransit Times (T₀) and Planet Orbital Periods (P) from Our Detrended TESS SPOC Two-minute Cadence Data

Planet Parameter	Planet b	Planet c
T0 (BJD)	2458765.95501 ± 0.00047	${2458766.16927}_{-0.00066}^{+0.00069}$
P (days)	3.092920 ± 0.000011	${6.765149}_{-0.000036}^{+0.000033}$

Note. The values are median values with 1σ uncertainties.

Download table as:  ASCIITypeset image

Beyond the known transits of HD 219134 b and c, no additional transits are detected in the SPOC light curves by the pipeline. We also find with our own detrended SPOC data that there are no obvious signs of any additional transit-like signals in the TESS Sectors 17 or 24 HD 219134 light curves.

Our TESS-data-derived radius lower limit for HD 219134 f is 1 R_⊕, based on our injection and recovery tests (Section 2.4.2; Figure 9). This value is stringent enough to rule out transits.

Zoom In Zoom Out Reset image size

Our derived lower limit of 1 R_⊕ is smaller than the 1.3 R_⊕ for a pure iron planet for HD 219134 f's radial-velocity minimum mass of $7{M}_{\oplus }$ (e.g., Fortney et al. 2007; Seager et al. 2007; Zeng et al. 2016; Gillon et al. 2017). More specifically, we derive a lower limit of 0.925 R_⊕ and 1.25 R_⊕ for a 50% and 80% completeness in the recovery rate, respectively. With our derived lower radius limit smaller than any plausible planet, we thus rule out HD 219134 f as a transiting planet.

With a robust nondetection, we can state that, if real, that inclination of HD 219134 f is less than 886 (assuming we cannot detect a grazing transit).

TESS provides no constraint on HD 219134 d which did not pass through inferior conjunction during the TESS observations.

From our TTV analysis (Section 2.5) we find no evidence that TTVs would have prevented either planet f or d from being detectable with the current recovery algorithms.

The TTV amplitude of HD 219134 f (Figure 6) could have been larger than the expected transit duration. However, since planet f should have transited once during the TESS data with a depth of about 238 ppm, it should have been detectable regardless of TTVs. Specifically we found that given the samples of orbital parameters, the distribution of TTVs ranges from 50 minutes to 350 minutes, larger than the expected transit duration of ≲230 minutes (Table 1).

The potential TTV amplitude of HD 219134 d (Figure 6) is smaller than the expected transit duration. More specifically we find that given the samples of orbital parameters, the distribution of TTVs ranges from 50 minutes to 250 minutes, less than the expected transit duration of ≲340 minutes (Table 1). Since the vast majority of TTV amplitudes are within the expected transit duration, TTVs most likely did not have an effect on the planet being observed.

A limitation to our TTV approach is that we used results from a Keplerian fit to infer TTV amplitudes that result from non-Keplerian dynamics. Another limitation is that both injection and recovery algorithms used linear ephemeris parameterization. One could push the TTV analysis to further accuracy by performing a full N-body photodynamical fit of the radial-velocity data and the photometry, including a non-Keplerian fit. Such an approach, however, is not warranted in the present paper because the expected transit depths of planets f and d are below the sensitivity threshold of ASTERIA.