1. Introduction
Giant planets orbiting close to their host stars are an uncommon product of the planet formation process. Despite their inherent rarity, the strong observational biases of the transit and radial velocity methods have allowed for the discovery and characterisation of hundreds of such close-in, giant planets (Winn & Fabrycky, Reference Winn and Fabrycky2015). This sample has revealed multiple trends, including the correlations of giant planet occurrence with stellar mass and metallicity (e.g. Johnson et al., Reference Johnson, Aller, Howard and Crepp2010), a lack of close companions to hot giants (e.g. Huang et al., Reference Huang, Wu and Triaud2016) and radius inflation from stellar irradiation (e.g. Weiss et al., Reference Weiss, Marcy and Rowe2013). These facts have provided tantalising hints into the formation and evolution mechanisms of hot, giant planets (Dawson & Johnson, Reference Dawson and Johnson2018).
Another major discovery from Kepler (Borucki et al., Reference Borucki, Koch and Basri2010) was a dearth of sub-Jovian planets below periods of roughly 3 days, known as the ‘sub-Jovian’ or ‘Neptunian’ desert (Szabó & Kiss, Reference Szabó and Kiss2011; Mazeh et al., Reference Mazeh, Holczer and Faigler2016). This was soon connected with evolutionary processes, such as photoevaporation, tidal disruption and possibly migration scenarios (e.g. Owen & Lai, Reference Owen and Lai2018). In recent years, planets have begun to be discovered in the desert, converting the region into a so-called ‘savanna’ (for examples, see Kálmán et al., Reference Kálmán, Szabó and Borsato2023). These are excellent targets for not only understanding the processes shaping the desert/savanna, but will also provide insights into giant planet formation more broadly. Furthermore, they are accessible for studies with small telescopes due to their large transit depths.
Since the launch of the Transiting Exoplanet Survey Satellite (TESS, Ricker et al., Reference Ricker, Winn, Vanderspek and Oschmann2014), thousands of new exoplanet candidates have been reported. A major advantage of TESS over previous surveys is its nearly all-sky coverage, enabling studies of the exoplanet population across the galaxy. Modern programs are now using TESS data to investigate specific sub-populations of giant planets. For example, the ‘Searching for Giant Exoplanets around M-dwarf Stars’ survey (GEMS, Kanodia et al., Reference Kanodia, Cañas and Mahadevan2024) and ‘M dwarfs Accompanied by close-iN Giant Orbiters with SPECULOOS’ project (MANGOS, Triaud et al. Reference Triaud, Dransfield and Kagetani2023) are aiming to study the occurrence rates of giant planets around low mass stars to test classical planet formation scenarios, such as core accretion (e.g. Lissauer, Reference Lissauer1993) and gravitational instability (Boss, Reference Boss1997). This was a difficult task in the Kepler era, due to the limited sky coverage resulting in a small sample of M-dwarf hosts and, therefore, incomplete statistics. Another example is the ‘TESS Hunt for Young and Maturing Exoplanets’ (THYME, Newton et al., Reference Newton, Mann and Tofflemire2019). This program is searching for young transiting planets in stellar associations in order to understand evolutionary processes, including for gas giants.
With this context, we introduce the ‘TASmanian Search for Inclined Exoplanets’ (TASSIE) follow-up program. Based at the University of Tasmania Greenhill Observatory (UTGO), we aim to study giant transiting planets in the southern sky using photometric and spectroscopic measurements. In particular, we wish to contribute to the understanding of the sub-Jovian desert/savanna through the search for more elusive planets in this region. By leveraging our own telescopic infrastructure, we hope to ensure a consistent approach to photometric data collection and analysis. A side benefit of our program will be the addition of another facility for follow-up of southern planet candidates from TESS and future space-based transit surveys. This will be useful for ruling out false positives scenarios, such as eclipsing binaries (EBs), along with refining orbital ephemerides to allow for successful observations with time-limited facilities.
In this paper, we present the first results of the TASSIE follow-up program, along with the photometric characterisation of the new Harlingten 50 cm telescope at UTGO. The paper is organised as follows. In Section 2, we introduce the instrumentation, data reduction process and target selection strategy for TASSIE. In Section 3, we present our first results with observations of five short-period TESS TOIs. In Section 4, we begin our analysis by first deriving the stellar parameters using photometric and spectroscopic measurements. We then combine astrometry, archival imaging and statistical tests with TRICERATOPS (Giacalone et al., Reference Giacalone, Dressing and Jensen2021) to rule out false positive scenarios. We then move to joint transit modelling using data from TESS and the Harlingten 50 cm. In Section 5, we discuss searching for transit timing variations and the photometric performance of our telescope. Finally, we present our conclusions in Section 6.
2. TASSIE follow-up program
2.1 Instrumentation – The Harlingten 50 cm telescope
The Harlingten 50 cm telescope (H50, see Figure 1) is a Planewave 20-inch (508 mm) of corrected Dall-Kirkham design, with a 191 mm secondary mirror. It is equipped with two corrector lenses, producing a 3454 mm focal length. The telescope is mounted onto an Alcor Systems Nova 120 direct-drive mount. For guiding, we utilise a 648
$\times$
486 SBIG ST-i autoguiding CCD with the PhD2 Guiding softwareFootnote
a
. For the imaging camera, we are using an Apogee Alta U42 CCD. With 2048
$\times$
2048 13
$\mu$
m pixels and a pixel scale of 0.8
$^{\prime\prime}$
/pixel, the system achieves a 27.2 sq. arcminute field-of-view (FOV). Accompanying this camera is an Optec TCF-S focuser (with a step resolution of 2.2
$\mu$
m) and an Apogee FW50-10S filter wheel. We have Bessell B & V and SDSS g
$^{\prime}$
, r
$^{\prime}$
, i
$^{\prime}$
broadband filters, along with narrowband H
$\alpha$
, H
$\beta$
, OIII & SII.
The Harlingten 50 cm from inside the dome, displaying the camera and filter wheel attached.
2.1.1 Noise characteristics
To determine the noise characteristics of the current CCD camera, we employ the methods described in Howell (Reference Howell2006). We measure the gain and read noise using sets of short exposure and high signal-to-noise ratio (SNR) twilight flat frames and a collection of bias frames. To measure the typical dark current, we use multiple sets of dark frames at different exposure times. After creating a master bias frame and subtracting this from each set of dark frames, we calculate mean and standard deviation in each set of exposures. We then convert this to electrons using the previously calculated gain. The dark current is estimated by performing a linear regression of the mean electron count against exposure time. The measured values with the CCD cooled to the typical operational temperature (-30
$^\circ$
C) are shown in Table 1. Empirically, we have found a magnitude limit in 1-h integration of
$r' = 20.08 \pm 0.19$
for a
$SNR \gt 10$
in dark time.
Telescope and Imaging Camera Information for the H50
2.2 Data reduction and photometry
We have developed a predominately Python-based reduction pipeline, utilising the Prose package (Garcia et al., Reference Garcia, Timmermans and Pozuelos2022). An auto-reduction process performs daily checks for new data. After organisation and separation of science and calibration data, master calibration frames are produced by either a sigma-clipped mean (for bias and dark frames) or using a simple median (for flats). The science images are then reduced using standard bias, dark and flat-field corrections with the closest available calibration frames by date. Bad pixel maps are also built using the darks to help remove hot and dead pixels. The reduced science images are then passed to a locally installed copy of Astrometry.net (Lang et al., Reference Lang, Hogg, Mierle, Blanton and Roweis2010) to determine the World Coordinate System (WCS) information.
For transit data, the most common tool for photometric analysis is differential aperture photometry. We implement this once again with Prose, which chooses optimal comparison stars using the algorithm outlined in Broeg et al. (Reference Broeg, Fernández and Neuhäuser2005). In summary, differential light curves are extracted for all detected sources and assigned a weighting depending upon the amount of variance. Then comparison stars are recursively rejected until only the least varying stars remain. It should be noted that this algorithm does not take into account the colour of the target, which may cause systematics in the data due to differential refraction effects. Thus, we have also implemented star selection based on closest colour and magnitude through queries to the Gaia DR2 catolog (Brown et al., Reference Brown, Vallenari and Prusti2018).
2.3 Target selection strategy
The main goal of TASSIE is to validate giant planets, with a particular emphasis upon those residing within the sub-Jovian desert/savanna. We therefore focus on candidates with periods below
$\approx$
3 days. As a small telescope program, we are limited to relatively bright targets. Combining results from our Exposure Time Calculator (ETC)Footnote
b
and observations of confirmed planets, we have found that transit depths greater than 5 parts per thousand (ppt) are required for a reasonable transit SNR to be obtained with our instrument. This value is typical of other sub-metre ground-based telescopes (e.g. Gillon et al., Reference Gillon, Jehin and Magain2011; Addison et al., Reference Addison, Wright and Wittenmyer2019). This limit is adequate for detection of planets ranging from sub-Saturns to Hot Jupiters orbiting FGK stars, with the possibility to reach sub-Neptunes around M dwarf stars.
We impose a condition of exposure times below 200 s (with less than 120 s being preferred) to improve upon the TESS cadence. This allows us to search for transit timing and transit depth variations (TTVs and TDVs, see Agol & Fabrycky, Reference Agol, Fabrycky, Deeg and Belmonte2018) with greater precision. We choose to use the SDSS r’ filter primarily, as it has peak quantum efficiency (QE = 0.92) for our camera. It also reduces atmospheric effects on the photometry (such as scintillation/dispersion) compared to bluer filters. This leads to a magnitude limit of r < 15 mag (for 1% photometry), with brighter targets amenable to radial velocity follow-up being prioritised. We discuss the experimental photometric precision achieved in Section 5.3, based upon the results reported in the following sections.
3. Observations
Observations for TASSIE began in the summer of 2023/2024. The targets were alerted through a custom Python script that downloads the list of TOIs each week from the Exoplanet ArchiveFootnote
c
. The Astroplan package (Morris et al., Reference Morris, Tollerud and Sipöcz2018) is then used to calculate the expected transit times for the next 10 epochs for any system observable from Tasmania. If any transits occur during local nighttime hours and pass our observing criteria (T < 15,
$\delta$
> 5 ppt), then they are added to a target list. We show details of each observing session and target, including the coordinates, filters, exposure time, median FWHM and number of images in Table 2. Note that the H50 telescope was not fully collimated during this period, leading to below-average FWHMs. We download all TESS target pixel files (TPF) and Science Processing Operations Center (SPOC) light curves from Mikulski Archive for Space Telescopes using the Lightkurve package (Lightkurve Collaboration et al., Reference Collaboration, Cardoso and Hedges2018).
Observation logs for TASSIE targets using the H50 telescope.
3.1 Photometry
3.1.1 TIC 452049244 (TOI3070)
TIC 452049244 is a V
$\approx$
14.3 mag star in the constellation of Centaurus. TIC 452049244 was observed by TESS in Sectors 10 (26/03/19–22/04/19), 37 (02/04/21–28/04/21) and 64 (10/03/2023–06/04/2023). TOI3070 was originally alerted as a TOI at 04/06/2021 01:36:30 UTC using the Faint Star Search pipeline (see Kunimoto et al., Reference Kunimoto, Daylan and Guerrero2022). A clear transit-like signal is visible with an approximate period of 0.66 days, determined using a simple Box-Least Squares search (Kovács et al., Reference Kovács, Zucker and Mazeh2002).
TOI3070 was observed with the H50 telescope on two nights. Beginning on the evening of the 16th of January 2024, 41 images were taken with an exposure time of 200 s in the SDSS r’ filter. Whilst the sky was clear and the moon illumination low (roughly 25 %), conditions were non-optimal with relatively strong winds resulting in a poor median FWHM of
$4.9 ^{\prime\prime}$
. The observations covered ingress and egress, but unfortunately captured limited out-of-transit baseline. On the 12th of February 2024, TOI3070 was re-observed with the same filter and camera settings. In this epoch the conditions were marginally better, with lower winds and clear, moonless skies. This lead to an improved median FWHM of
$3.5 ^{\prime\prime}$
. 24 images were obtained, covering the baseline and ingress. However, observations were interrupted due to high humidity forcing the closure of the dome early.
3.1.2 TIC 163262555 (TOI3097)
TIC 163262555 is a V
$\approx$
13.8 mag star residing in the constellation of Centaurus. It was observed by TESS in Sectors 10 (26/03/19–22/04/19), 37 (02/04/21–28/04/21), 63 (10/03/2023–06/04/2023) and 64 (06/04/2023–04/05/2023). TOI3097 was originally alerted as a TOI on 04/06/2021 01:36:34 UTC from the Faint Star Search. This target is known to have a binary companion within
$1.5^{\prime\prime}$
of the primary target (Mugrauer et al., Reference Mugrauer, Rück and Michel2023), corresponding to a separation of 576 au. A possible transit signal is visible in the SPOC light curves, with an approximate period of 1.368 days. The periodogram and TPF are shown in the Appendix B.
The TOI3097 system was observed with the H50 on the 15th of March 2024 using the SDSS r’ filter. A total of 91 images were obtained, each with exposure time of 90 s. This covered the baseline prior to ingress through to end of egress. Sky conditions were clear and moonless. Low winds allowed for the best FWHM obtained during the summer season, with the median for the night at
$2.7^{\prime\prime}$
.
3.1.3 TIC 454918388 (TOI3124)
TIC 454918388 is a V
$\approx$
13.3 mag star in the constellation of Chamaeleon. TIC 454918388 was observed by TESS on multiple occasions, with the most recent in Sectors 64 (06/04/2023–04/05/2023), 65 (04/05/2023–02/06/2023) and 66 (02/06/2023–01/07/2023). TOI3124 was alerted as a TOI on 04/06/2021 13:54:19 UTC via the Faint Star Search. Inspection of the SPOC light curve shows moderate flux variations, indicating that this is a variable/active star. To determine if a planetary signal is hidden within this variability, we apply a Savitzky-Golay filter to the data (Savitzky & Golay, Reference Savitzky and Golay1964). Performing the usual BLS search on this detrended light curve gives the periodogram shown in the bottom right of Figure 2 (a). The planetary periodogram indicates a period of
$\approx$
2.255 days, which aligns with the stellar rotation period of 2.32 days (discussed in Section 4.1.4).
Summary plots for observations of TOI3124 (a) and TOI3163 (b). The TESS Sector 64 TPF, normalised light curves and periodogram plots are shown for each target. Cutouts of the stacked H50 images are also shown, matching the approximate orientation and scale of each TPF. In (a), the orange light curve shows the raw, normalised flux for the star. The blue light curve is the detrended light curve (offset by a constant). The periodogram refers to the planetary period, based upon the detrended data.
The H50 first observed TOI3124 on 17/03/2024 with the SDSS r’ filter. The exposure times were 120 s, with a total of 67 images obtained. The conditions were clear and moonless with low winds to begin with, before increasing later in the night. This resulted in a median FWHM of
$3.2^{\prime\prime}$
for the observing session. We re-observed TOI3124 on 01/07/2024, this time in the Bessell V filter with exposures of 120 s. The sky conditions were clear and moonless, but the humidity was high. The median FWHM for this session was
$3.1^{\prime\prime}$
.
3.1.4 TIC 425316308 (TOI3163)
TIC 425316308 is V
$\approx$
12.1 mag star in the constellation of Centaurus. The star was observed by TESS in Sectors 11 (23/04/19–20/05/19), 37 (02/04/21–28/04/21), 38 (29/04/21–26/05/21) and 64 (06/04/2023–04/05/2023). TOI3097 was originally alerted as a TOI on 04/06/2021 13:54:19 UTC by the Faint Star Search. Strong transit-like dips at roughly 1.5 % are visible in the non-folded light curve. The period with the highest BLS power was 3.072 days, as can be seen in Figure 2 (b).
TOI3163 was observed on 10/02/2024 with the H50 using the SDSS r’ filter with exposures of 120 s. 90 images were obtained over an airmass range of 1.8–1.1. The sky was initially clear and moonless, but relatively strong winds lead to a poor median FWHM at 3.3". The baseline, ingress and midpoint were captured, but the egress was missed due to cloud.
3.1.5 TIC 442050698 (TOI4266)
TIC 442050698 is a V
$\approx$
14.2 mag star in the constellation of Carina. The star was observed by TESS in Sectors 9 (28/02/19–25/03/19), 10 (25/03/19–22/04/19), 37 (02/04/21–28/04/21), 63 (10/03/23–06/04/23) and 64 (06/04/23–04/05/23). It was alerted as a TOI on 12/07/2021 14:48:48 UTC via the Faint Star Search. A possible transit signal is visible in the SPOC light curves, with an approximate period of 0.811 days.
TOI4266 was observed by the H50 on 15/01/2024 with the SDSS r’ filter with an exposure time of 150 s per image. The sky was clear and the moon illumination was low (at roughly 15 %). Conditions were sub-optimal with very strong winds ranging from 15 to 20 km/h, resulting in a poor median FWHM of
$3.6^{\prime\prime}$
. A total of 56 images were obtained over an airmass range of 1.1–1.3, covering baseline, ingress and egress.
3.2 Spectroscopy
To aid in the characterisation of the target stars and attempt to rule out double-lined binaries, we obtained mid-resolution reconnaissance spectra from the Wide Field Spectrograph (WiFeS, Dopita et al., Reference Dopita, Rhee and Farage2010) on the Australian National University (ANU) 2.3 m telescope. We used the R = 7000 red and blue gratings, which span the wavelength range of
$\lambda$
= (420,700) nm. The velocity resolution of the instrument is
$\delta$
v = c/R
$\approx$
45 kms
$^{-1}$
. We took standard calibration data including bias, dome flats, twilight flats, Ne-Ar arc lamp and wire frames. Standards stars were also observed after each target for telluric corrections and flux calibration. Data were reduced using the PyWiFeS pipeline (Childress et al., Reference Childress, Vogt, Nielsen and Sharp2014). The observation logs for the spectroscopic measurements are shown in Table 3.
Observation logs for TASSIE targets using the WiFeS spectrograph on the ANU 2.3 m telescope at Siding Spring Observatory.
4. Data analysis and results
4.1 Stellar parameters
4.1.1 Spectral energy distribution fitting
As a first step in determining the stellar parameters, we used the spectral energy distribution (SED) fitting code ARIADNE (Vines & Jenkins, Reference Vines and Jenkins2022). Four model grids were utilised in our analysis, including Phoenix v2 (Husser et al., Reference Husser, Wende-von Berg and Dreizler2013), Kurucz 1993, ATLAS9 (Castelli & Kurucz, Reference Castelli, Kurucz, Piskunov, Weiss and Gray2003) and BT-Settl (Allard et al., Reference Allard, Homeier, Freytag, Johns-Krull, Browning and West2011). We queried Vizier for all publicly available photometry from various surveys including Gaia DR3, 2MASS (Skrutskie et al., Reference Skrutskie, Cutri and Stiening2006) and SkyMapper (Wolf et al., Reference Wolf, Onken and Luvaul2018). We set broad normal priors on the effective temperature (
$T_{eff}$
), surface gravity (
$\log{(g)}$
), metallicity ([Fe/H]) and extinction (
$A_{v}$
, corrected from
$A_{0}$
) centred on the reported Gaia DR3 values from XP spectra (Gaia Collaboration et al., 2023). As the uncertainties are often underestimated, we used empirically measured precisions of
$\sigma_{log(g)}$
= 0.1 dex and
$\sigma_{[Fe/H]}$
= 0.2 dex from the DR3 documentationFootnote
d
and set 10
$\%$
errors on
$T_{eff}$
and
$A_{v}$
as prior widths. The Gaia DR3 parallaxes with uncertainties were provided to constrain the distances.
The radius and distance of each star is determined by scaling the stellar model grid surface fluxes to the observed fluxes (see Equation 1 of Vines & Jenkins, Reference Vines and Jenkins2022). The final stellar parameters come from Bayesian model averaging of individual model posteriors. MESA Isochrones and Stellar Tracks (MIST, Dotter, Reference Dotter2016) models were then used to estimate the mass by interpolation. We ran nested sampling with 1000 live points and 100000 samples, with evidence tolerance of
$\Delta\log(Z)$
= 0.05. An example of the results of this fitting process are shown in Figure 3. It is worth noting some ambiguity in the results due to uncertain distances and possible multiplicity for TOI3070 and TOI4266, as discussed in Section 4.2.
The results of SED fitting for TOI3124 with ARIADNE. (a) The spectral energy distribution of TOI3124 with Phoenix V2 model overlaid. (b) The corner plot for stellar parameters of TOI3124.
4.1.2 Semi-empirical relations
We excluded TOI3097 and TOI3163 from the SED fitting, due to strong contamination of ground-based photometry. For TOI3163, there is a moderately bright (G
$\approx$
15) non-companion source at roughly 3" away. TOI3097 has already been noted to have a bound wide companion. We instead estimate stellar parameters using the semi-empirical relations of Torres et al. (Reference Torres, Andersen and Giménez2010) for early-to-mid type dwarfs and Mann et al. (Reference Mann, Feiden, Gaidos, Boyajian and von Braun2015) for late dwarfs in the range of K7-M7. For TOI3097A and TOI3163, we adopt estimates for effective temperature, surface gravity and metallicity from Gaia DR3 (Gaia Collaboration et al., 2023) as listed in Table 4. We find mass and radii of (0.63
$\pm$
0.04
$M_{\odot}$
, 0.87
$\pm$
0.03
$R_{\odot}$
) and (1.29
$\pm$
0.08
$M_{\odot}$
, 1.56
$\pm$
0.05
$R_{\odot}$
) for TOI3097A and TOI3163, respectively.
Stellar parameters for TASSIE TOI host stars. Measurements with (*) indicate suspected blended photometry. Derived properties with (
$\dagger$
) are from SED-fitting, with the remainder combining Gaia DR3 and semi-emperical relations.
For TOI3097B, we used the BPRP-
$T_{eff}$
relation of Mann et al. (Reference Mann, Feiden, Gaidos, Boyajian and von Braun2015) to estimate the effective temperature of
$T_{eff}$
= 3940
$\pm$
223 K. We then used the radius-
$T_{eff}$
relation to estimate a radius of 0.30
$\pm$
0.02
$R_{\odot}$
. We note that Gaia reports a low metallicity of [Fe/H]
$\approx$
-1.0 for the TOI3097 system, implying these are subdwarf stars. As the Mann et al. (Reference Mann, Feiden, Gaidos, Boyajian and von Braun2015) sample does not extend below [Fe/H] = -0.6 dex, we also checked the radii using data from Kesseli et al. (Reference Kesseli, Kirkpatrick and Fajardo-Acosta2019). We refit a colour-radii relation for subdwarfs (as discussed in the Appendix) and find a matching radius of 0.28
$R_{\odot}$
. As subdwarf masses have been scarcely measured, we can only roughly estimate mass and surface gravity using the dwarf sequence of Pecaut & Mamajek (Reference Pecaut and Mamajek2013). Interpolating with radius, we find M = 0.27
$\pm$
0.03
$M_{\odot}$
and log(g) = 4.91
$\pm$
0.08.
4.1.3 Template matching
As an independent check of the properties of the TOIs, we now investigate the WiFeS spectra. To assign stellar classifications, we utilise PyHammer v2 (see Kesseli et al., Reference Kesseli, West and Veyette2017; Roulston et al., Reference Roulston, Green and Kesseli2020). By comparing multiple spectral indices and colour regions measured from Sloan Digital Sky Survey (York et al., Reference York, Adelman and Anderson2000) spectral templates with those measured from the input spectra, spectral types and metallicity are estimated through the minimisation of a
$\chi^2$
difference. An advantage of this method is robustness against flux calibration errors and low SNR spectra. The code can also identify some double-lined binaries. Due to the low radial velocity precision and limited wavelength coverage, we shifted the WiFeS spectra to the rest wavelength by assuming the Gaia DR3 radial velocities. The estimated values were consistent with those found from the previous methods. However, the low SNR of data prevented further refinement of the stellar parameters.
4.1.4 Kinematics
The kinematics of stars can provide information upon the system age and possible associations of the star within the galaxy (Bensby et al., Reference Bensby, Feltzing and Lundström2003). We calculate the galactic velocities of the five TOIs using the equations defined by Johnson & Soderblom (Reference Johnson and Soderblom1987), with updated transformation matrix for the J2000 epoch. We adopt the parallax, proper motions and radial velocities in Table 4 from Gaia DR3 (Gaia Collaboration et al., 2023). We take the solar velocities of (
$U_{\odot}, V_{\odot}, W_{\odot}$
) = (8.5,13.38,6.49) km/s from Coşkuno
$\unicode{x01E7}$
lu et al. (2011). Here, U is defined as positive towards the centre of the galaxy, V is positive with the direction of galactic rotation and W is positive in the direction of the North galactic pole. The results of this analysis are listed in Table 4.
Using the membership probabilities defined by Bensby et al. (Reference Bensby, Feltzing and Lundström2003), we place four of the TOIs in the thin disk. However, the kinematics of TOI3097 (specifically the large negative value of W and positive value of V) potentially indicate a thick disk star, with a relative thick disk to thin disk probability of
$\approx$
1.6. This is consistent with the low metallicity measured. We also utilise the BANYAN
$\Sigma$
tool (Gagné et al., Reference Gagné, Mamajek and Malo2018) to check if any of the targets are members of known young stellar associations. We obtain a 99.9 % probability of the targets being field stars.
4.1.5 Stellar activity
Stellar activity can add significant complexity to analysis of transit signals. However, it can be used to infer the stellar rotation period which is often used as a proxy for age (e.g. Mamajek & Hillenbrand, Reference Mamajek and Hillenbrand2008). As TESS observes in cycles of roughly 27 days per sector, rotational periods longer than this cannot be not well-measured. We therefore also investigate public archival photometry from the ASAS-SN (Kochanek et al., Reference Kochanek, Shappee and Stanek2017) and ATLAS (Tonry et al., Reference Tonry, Denneau and Heinze2018) surveys. We search for periodicity using a Lomb-Scargle periodogram search, calculating the False Alarm Probability (FAP) for the maximum power period. Aside from TOI3124, we do not find any consistent periods in the three data sets of significance.
It is apparent from the TESS and ASAS-SN light curves that TOI3124 is an active star at the 5-10 % level. We attribute at least part of this variability to stellar spots. This conclusion comes from the quasi-periodic and morphologically variable modulation, along with the low effective temperature. The latter rules out the possibility of most pulsators, such as
$\delta$
Scuti and
$\gamma$
Doradus, which are usually F-type or earlier (Reinhold et al., Reference Reinhold, Reiners and Basri2013; Balona & Abedigamba, Reference Balona and Abedigamba2016). We also find evidence of at least one flaring event in the Sector 64 light curve (at BJTD
$\approx$
3044.38), further indicating this is an active star. Inspecting the periodogram from each survey in Figure 4, we find a matching period of 2.31
$\pm$
0.01 days. A marginally weaker signal is seen in the TESS periodogram at 2.43 days. This is unusually quick for a mid G-dwarf and may be related to multiplicity in the system.
The ASAS-SN (in blue) and TESS (in black) folded light curves and periodogram for TOI3124. The inset shows the peak power periods, closely spaced at 2.31 and 2.43 days.
4.2 Planet validation
In order to validate these planet candidates, we will need to rule out other astrophysical scenarios that could explain the transit-like features. As we have detected these signals with the H50, we can dismiss the possibility of TESS systematics creating the signal. We can also exclude many potential host stars, due to our much higher resolution of 0.8
$^{\prime\prime}$
/pixel. We now explore other methods to exclude false positive scenarios.
4.2.1 Astrometry
Eclipsing binaries can mimic the signatures of transiting planets. One method for identifying the presence of bound companions that could introduce the transit-like signal is through checking the quality of the Gaia astrometric solution. To do this we look at the ‘Renormalised Unit Weight Error’ (RUWE), which is defined as:
$\begin{equation} RUWE = \frac{1}{f} \sqrt{\frac{\chi^2}{(n - m)}}\end{equation}$
here, n is the number of good observations, m is the number of fit parameters and f is a reference value which is a function of magnitude and colour. Generally, sources with RUWE > 1.4 are considered to have questionable astrometry, potentially indicating the presence of multiplicity in the system (Belokurov et al., Reference Belokurov, Penoyre and Oh2020). TOI3070 and TOI4266 are notable as they have RUWE values of 3.31 and 5.03 respectively, indicating very poor astrometric fits. TOI3163 is also interesting as the brightest star has a RUWE = 0.8, but a close background star has a value of 2.8. We must therefore be cautious with the quoted values (such as parallax and proper motions) for these stars. The other stars have RUWE values below the threshold.
4.2.2 Archival imaging
To help rule out background EBs, we retrieve archival images from the 2MASS survey (Skrutskie et al., Reference Skrutskie, Cutri and Stiening2006). These range from 20 to 25 years prior to imaging with the H50. We extract catalogues from our stacked images using SExtractor (Bertin & Arnouts, Reference Bertin and Arnouts1996) and plot the detected sources over the 2MASS images. Due to the generally low proper motions, the positions of the targets do not change drastically across the two epochs. No contaminating stars were identified which were not already flagged from our own imaging.
4.2.3 Statistical testing with TRICERATOPS
We perform statistical tests for false positive scenarios using TRICERATOPS (Giacalone et al., Reference Giacalone, Dressing and Jensen2021). We utilise the TESS light curves and transit depth estimates, along with the stellar parameters from Section 4.1 in this analysis. We remove many surrounding stars from the calculations due to the transit-like signals detected with the H50, further tightening the constraints. The output of this process are relative probabilities of various scenarios, along with estimates of the false positive probability (FPP) and nearby false positive probability (NFPP). We repeated the fitting process 10 times on each TOI to ensure the FPPs and NFPPs were consistent. We list these estimated values and uncertainties in Table 5 for each candidate.
TRICERATOPS Results for TASSIE TOIs
We can see that TOI3070 has a very high FPP of
$\approx$
0.79, with the most likely scenario being an EB (either around the primary or a secondary star). The high RUWE value also supports this hypothesis. Interestingly, the SED fitting process produced odd radius values for the estimated spectral type (F5–F7 V), ranging from 1.8 to 3
$R_{\odot}$
instead of the expected 1.3–1.4
$R_{\odot}$
(Pecaut & Mamajek, Reference Pecaut and Mamajek2013). Similarly for TOI4266, we have a significant FPP of
$\approx$
0.84, with the most probable scenario being an EB (nearby or with double period). Once again, this is consistent with the high RUWE value. We shall exclude these candidates from further modelling, as we are interested in planets only. Finally, we note that whilst the mean FPP for TOI3124 is very high at
$\approx 0.99$
(making the signal a likely false positive), we wanted to check that this result was unbiased by the method of activity detrending. Hence, we decided to include it in the modelling stage. The H50 light curves for these likely false positives are shown in the Appendix B.
4.3 Transit modelling
We model the transit light curves using the Juliet package (Espinoza et al., Reference Espinoza, Kossakowski and Brahm2019), which combines nested sampling algorithms with the light curve calculations of BATMAN (Kreidberg, Reference Kreidberg2015). For each of the remaining candidates, we perform joint modelling with high cadence (i.e exposures of 200 s or less) TESS data and light curves from the H50. We choose to incorporate Gaussian Processes (GPs) into our modelling to account for any stellar activity or remaining systematics. To estimate the quadratic limb darkening coefficients for each star, we use the ExoCTK limb darkening calculatorFootnote e with Phoenix V2 (Husser et al., Reference Husser, Wende-von Berg and Dreizler2013) models. We then convert to the limb darkening parameterisation suggested by Kipping (Reference Kipping2013), and set a normal prior with a conservative error of 10 %. We use the flux ratio calculated by TRICERATOPS to estimate the dilution factor for the H50 light curves. The SPOC light curves are already corrected for contamination, so we set the dilution factor to 1 for TESS data. We list descriptions of parameters and priors for the modelling in the Appendix in Table 7 (for further details, see Espinoza, Reference Espinoza2018).
4.3.1 TOI3097
Due to this being a binary system, we started with a coarse search of the parameter space with no constraint on the stellar density. As we can use the light curve to recover the stellar density from Kepler’s law, we use this to check which star the planet candidate is orbiting. It was found that the brighter, primary star is the likely host. We then remodelled the light curves, this time using the stellar density constraint of
$\rho = 1347.1 \pm 490.6$
kg/
$m^3$
(3
$\sigma$
error) and incorporating an (approximate) Matern kernel GP of the form:
$\begin{equation}k(\tau) = \sigma_{GP}^{2} \Big[\Big(1+\frac{1}{\epsilon}\Big) \exp^{-(1-\epsilon) s_{i}} + \Big(1-\frac{1}{\epsilon}\Big) \exp^{-(1+\epsilon) s_{i}}\Big]\end{equation}$
where,
$s_{i}$
=
$3\tau/\rho_{GP}$
is a scaled timescale of the GP and
$\sigma_{GP}$
relates to the amplitude of the GP. The variable
$\epsilon$
is set to 0.01 (see Foreman-Mackey et al. (Reference Foreman-Mackey, Agol, Ambikasaran and Angus2017) and Espinoza et al. (Reference Espinoza, Kossakowski and Brahm2019) for more discussion of the interpretation and use of this kernel). We ran nested sampling with 1000 live points, initially assuming a circular orbit. We also performed a separate run adding eccentricity (e) and argument of periastron (
$\omega$
), using the parameterisation of
$\sqrt{e} sin(\omega)$
and
$\sqrt{e} cos(\omega)$
. However, we found this produced a worse fit with a lower log-likelihood (
$\Delta lnZ = -2.276$
). The measured and derived planetary parameters are shown in Table 6, being consistent with a sub-Jovian to Jovian size planet on a circular orbit. The light curves with best fit model are shown in Figure 5.
The modelling results from Juliet for the two planets.
The light curve and joint transit model for TOI3097Ab. The top plot shows the full Sector 63 & 64 data from TESS, with the best-fit transit + GP model overlaid. In the bottom left plot, we show the phase-folded TESS data, model and residuals. On the bottom right plot, we show data and residuals from the H50 telescope.
4.3.2 TOI3124
The stellar variability required experimentation with the form of the GP used for this target. Like TOI3097, we started with a default Matern 3/2 kernel. We also tried a quasi-periodic kernel introduced by Foreman-Mackey et al. (Reference Foreman-Mackey, Agol, Ambikasaran and Angus2017), with four hyperparameters relating to the amplitude, shifting, length-scale and period of the GP. The advantage of this kernel is it is physically motivated by the quasi-periodic signals seen in stellar activity. We ran nested sampling with 1000 live points, again assuming a circular orbit and then relaxing this in a latter modelling run. Both runs resulted in unphysical values for the impact parameter (b > 1) and radius ratio (p > 0.2), with strong degeneracy between the two. Combining this with the proposed ‘planetary’ period matching the activity period of the star, we conclude that this signal is non-planetary in nature. We either attribute this to small-scale, long-lived star spots or a grazing eclipsing binary. This interpretation is supported by the absence of a clear transit signal in more recent V band observations from the H50 (shown in the Appendix). Roughly achromatic transit depths are expected for true planetary signals.
4.3.3 TOI3163
Finally, we modelled the data for TOI3163 using a prior on the stellar density of
$\rho$
= 478.4
$\pm$
164.2 kg/
$m^3$
(3
$\sigma$
error). The light curve is shown in Figure 6, with the results listed again in Table 6. The posterior distributions are shown in the Appendix B. We find a Jovian size planet with
$R_{3163b} = 1.42 \pm 0.05 \, R_{J}$
and
$P_{3163b} = 3.074966 \pm 0.000022$
days. A circular orbit model is preferred, with a log-likelihood difference for the eccentric model of
$\Delta lnZ = -7.373$
.
The light curve and joint transit model for TOI3163b. The top plot shows the full Sector 64 data from TESS, with the best-fit transit + GP model overlaid. In the bottom left plot, we show the phase-folded TESS data, model and residuals. On the bottom right plot, we show data and residuals from the H50 telescope.
5. Discussion
5.1 The short-period planet population
To put these two candidates into context, we show the short-period exoplanet population in Figure 7. We see that TOI3163 is likely a classical Hot Jupiter, as it falls in the middle of the distribution of similar planets. More interestingly, TOI3097Ab leans towards being a Saturn-size planet. Inspecting the position of TOI3097Ab in the radius-period plane, we note that it is on the upper edge of the sub-Jovian desert. This boundary is expected to be sculpted by photoevaporation, which will be testable through the detection of atmospheric loss from desert planets with the upcoming Ariel mission (Tinetti et al., Reference Tinetti, Drossart and Eccleston2018; Kálmán et al., Reference Kálmán, Szabó and Borsato2023). TOI3097Ab could also conform to the high-eccentricity tidal migration scenario with the presence of the wide M sub-dwarf companion acting to perturb the planet into an eccentric orbit. The current low eccentricity would then be consistent with tidal circularisation (Dawson & Johnson, Reference Dawson and Johnson2018). Combining this information with the potentially low metallicity suggests that the TOI3097 system will be an interesting target for further characterisation with mass measurements and transmission spectroscopy.
The radius-period plane for short period exoplanets, coloured by stellar effective temperature. TOI3097Ab and TOI3163b are shown as points with error bars (corresponding to 3
$\sigma$
).
5.2 Search for transit timing variations
The presence of undetected transiting or non-transiting planets in a system can act to perturb the expected Keplerian orbits of known transiting planets, resulting in TTVs and TDVs (Agol & Fabrycky, Reference Agol, Fabrycky, Deeg and Belmonte2018). We performed a search for TTVs for TOI3097Ab and TOI3163b using Juliet. We retained the best-fit parameters from the normal modelling run, this time passing priors for each orbital epoch. These were centred upon the values expected assuming a linear transit ephemeris. We then plot the residual between the observed and expected (calculated) transit mid times, as shown in Figure 8. For TOI3097Ab, we see minor deviations from expectations. However, these are not statistically significant, due to the large uncertainties (
$\approx$
4 min). This is validated by a significantly lower log-likelihood for the model (
$\Delta ln(Z) = -454.21$
). For TOI3163, we also find no evidence for TTVs above 2 min.
Residuals between the observed and expected transit times for TOI3097Ab (top) and TOI3163b (bottom). The error bars correspond to the 1
$\sigma$
uncertainty.
5.3 Photometric performance
Using the transit light curves, we can assess the quality of our data and photometry pipeline. We calculate the standard deviation of the raw residuals (without subtraction of the GP models) and measure the median photometric error for both targets. For TOI3097 (at r’ = 13.48
$\pm$
0.02), we find
$\sigma_{3097}$
= 2.6 ppt and <
$\delta F_{3097}$
> = 3.2 ppt with exposure times of 90 s. For TOI3163 (at r’ = 12.19
$\pm$
0.03), we obtain
$\sigma_{3163}$
= 1.8 ppt and <
$\delta F_{3163}$
> = 1.9 ppt with exposures of 120 s. Using our aforementioned ETC, we calculate the theoretical photon noise to be
$\approx$
1.5 ppt and
$\approx$
0.8 ppt for TOI3097 and TOI3163, respectively. This indicates that other sources of error are limiting the photometry, most likely being related to flat-field errors, atmospheric transparency changes and scintillation. Despite this, these values are consistent with other small telescope programs (e.g. Gillon et al., Reference Gillon, Jehin and Magain2011; Addison et al., Reference Addison, Wright and Wittenmyer2019) and are sufficient for TASSIE’s goals.
To determine if the assumption of white noise is valid, we also test for correlations within the data. We perform a Pearson-R test on the residuals with various variables, such as time/phase, airmass, position on the CCD, sky background and FWHM. For TOI3097, all R-values are below
$\pm$
0.1 with the largest at R = 0.06 for correlation with FWHM. Furthermore, testing the null hypothesis of uncorrelated and normally distributed samples returns p-values all greater than 0.5. This indicates that the data is consistent with white noise. However, the same is not true for TOI3163 with weak correlations between almost all variables and the residuals (R = 0.2–0.4). This could be explained by two factors. Firstly, the TESS data shows more intrinsic scatter for TOI3163, implying stellar activity is adding red (correlated) noise. Secondly, the worse conditions during the TOI3163 observing session (reported in Section 3.1) may have also contributed to these correlations. This highlights that whilst our telescope and photometry pipeline is performing well generally, data quality will be situational and this may restrict our effectiveness in the follow-up of small planets. It also reiterates that careful detrending of light curves is important, validating the use of GPs in our transit modelling.
6. Conclusion
The TESS mission has provided a large sample of new giant planet candidates to explore. We have presented the observations and analysis of two likely exoplanetary systems residing in the southern sky. The TOI3097 binary system hosts a sub-Jovian to Jovian size planet (
$R_{3097Ab}$
= 0.89
$\pm$
0.04
$R_{J}$
) orbiting the primary K-dwarf star with a period of
$P_{3097Ab}$
= 1.368386
$\pm$
0.000006 days. The potentially low metallicity ([Fe/H]
$\approx$
-1) of the system and proximity of the planet to the sub-Jovian desert makes it an exciting target for further follow-up studies. The TOI3163 system was found to contain a hot, Jupiter-size companion (
$R_{3163b}$
= 1.42
$\pm$
0.05
$R_{J}$
and
$P_{3163b}$
= 3.074966
$\pm$
0.000022 days) around a late F-type dwarf star in the thin disk. Radial velocity observations will be required to confirm these candidates are in the planetary mass regime. We also report three targets (TOI3070, TOI3124 and TOI4266) that we determine to be false positives, based upon their high RUWEs, statistical testing with TRICERATOPS and chromaticity. In future, we hope to compliment our photometric capabilities with high-resolution spectroscopic measurements. By searching for more elusive planets in the sub-Jovian desert, we will contribute to understanding the processes shaping the formation and evolution of hot, giant planets.
Acknowledgement
The authors wish to thank Caisey Harlingten for his generous support of the Greenhill Observatory. Without his donations, we would not have the H50 and 1.3 m telescopes. We thank Dr. Kym Hill, Mr. Keith Bolton, Dr. Tony Sprent and Dr. David Warren for their assistance in setting up the H50. We also thank the anonymous referee for their comments that helped to improve the manuscript. We acknowledge J. B. Marquette for his preliminary ideas for the data reduction pipeline.
This work uses data acquired at the Siding Spring Observatory with the Australian National University 2.3m Telescope. We acknowledge the traditional custodians of the land on which the telescope stands, the Gamilaraay people, and pay our respects to elders past and present.
This paper includes data collected by the TESS mission, which are publicly available from the Mikulski Archive for Space Telescopes (MAST). Funding for the TESS mission is provided by NASA’s Science Mission directorate. We acknowledge the use of public TESS Alert data from pipelines at the TESS Science Office and at the TESS SPOC. Resources supporting this work were provided by the NASA High-End Computing (HEC) Programme through the NASA Advanced Supercomputing (NAS) Division at Ames Research Centre for the production of the SPOC data products.
This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC, https://www.cosmos. esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.
This research has made use of the SIMBAD data base and VizieR catalogue access tool, operated at CDS, Strasbourg, France.
This research has made use of NASA’s Astrophysics Data System. This research has made use of the NASA Exoplanet Archive, which is operated by the California Institute of Technology, under contract with the National Aeronautics and Space Administration under the Exoplanet Exploration Programme.
Funding statement
This research is supported by an Australian Government Research Training Program (RTP) Scholarship.
Competing interests
The authors are not aware of any competing interests.
Data availability statement
The code/notebooks used for analysis can be found on Github here: https://github.com/tjplunkett/TASSIE-I. Light curves and spectra will be uploaded to the TESS ExoFOP website after publishing.
Appendix A. Subdwarf radius-colour relation
Whilst trying to fit the radius for the potential sub-dwarf TOI3097B using the work of Kesseli et al. (Reference Kesseli, Kirkpatrick and Fajardo-Acosta2019), we noticed that their quoted relation (Eq. 5 in the paper) did not match the graphed results. The authors argued for a decreasing exponential form, requiring that no radius should be below
$\approx$
0.1
$R_{\odot}$
(due to degeneracy pressure). However, the form and coefficients listed did allow for such values. We therefore decided to re-fit the relations, slightly adjusting the form to:
$\begin{equation} \frac{R}{R_{\odot}} = 0.1 + A \times e^{-(b(BP-RP) \, + \, c[Fe/H]),}\end{equation}$
where, BP and RP are the Gaia blue and red bands magnitudes, respectively. We queried Vizier for their catalogue of 88 subdwarfs with photometry and derived stellar parameters. We cut any star with [Fe/H] < −0.5 dex to ensure the relations captured only subdwarf behaviour. Fitting the above relation, we find best fit coefficients of A = 84.9138
$\pm$
36.1258, b = 2.7186
$\pm$
0.1940 and c = −1.1537
$\pm$
0.1341. The data and best-fit models for three metallicity bins are shown in Figure 9.
Appendix B. Auxillary plots and tables
Stellar radius versus Gaia BP-RP colour for subdwarf stars from the Kesseli et al. (Reference Kesseli, Kirkpatrick and Fajardo-Acosta2019) sample. Data is coloured based upon the metallicity. The dashed lines show the best fit model at [Fe/H] = −0.5, −1.0 and −2.0.
The priors used in modelling of TASSIE TOIs with Juliet. Note: N, U and logU correspond to normal, uniform and log-uniform distributions.
Summary plots for observations of TOI3097. The TESS Sector 64 TPF, normalised light curve and periodogram plots are shown. A cutout of the stacked H50 image is also displayed on the top right, matching the approximate orientation and scale of the TPF.
Light curves for non-planetary TOIs from the H50 telescope. Points with errorbars represent the data binned to 600 s.
Corner plots of the posterior distributions from the modelling of TOI3097Ab (top) and TOI3163 (bottom) with Juliet.
Open access
