eleanor Light curves

eleanor: An open-source tool for extracting light curves from the Tess Full-Frame Images

[ Department of Astronomy and Astrophysics, University of Chicago, 5640 S. Ellis Ave, Chicago, IL 60637, USA [ Department of Astronomy and Astrophysics, University of Chicago, 5640 S. Ellis Ave, Chicago, IL 60637, USA [ Center for Computational Astrophysics, Flatiron Institute, 162 Fifth Ave, New York, NY 10010, USA [ Center for Computational Astrophysics, Flatiron Institute, 162 Fifth Ave, New York, NY 10010, USA Nicholas Saunders NASA Ames Research Center, Moffett Field, CA 94035, USA Bay Area Environmental Research Institute, 625 2nd St Ste. 209, Petaluma, CA 94952 [ Department of Astronomy and Astrophysics, University of Chicago, 5640 S. Ellis Ave, Chicago, IL 60637, USA [ Caltech/IPAC-NASA Exoplanet Science Institute, M/S 100-22, 770 S. Wilson Ave, Pasadena, CA 91106 USA [ NASA Ames Research Center, Moffett Field, CA 94035, USA Bay Area Environmental Research Institute, 625 2nd St Ste. 209, Petaluma, CA 94952 [ Center for Computational Astrophysics, Flatiron Institute, 162 Fifth Ave, New York, NY 10010, USA [ Department of Physics, Duke University, 120 Science Drive, Durham, NC, 27708, USA José Vinícius de Miranda Cardoso Universidade Federal de Campina Grande, R. Aprígio Veloso, 882 Universitário, Campina Grande, 58429, Brazil Adina D. Feinstein; \faTwitter@afeinstein20 afeinstein@uchicago.edu
Abstract

During its two year prime mission the Transiting Exoplanet Survey Satellite (TESS) will perform a time-series photometric survey covering over 80% of the sky. This survey comprises observations of 26 24 sectors that are each monitored continuously for approximately 27 days. The main goal of TESS is to find transiting planets around 200,000 pre-selected stars for which fixed aperture photometry is recorded every two minutes. However, TESS is also recording and delivering Full-Frame Images (FFIs) of each detector at a 30 minute cadence. We have created an open-source tool, eleanor, to produce light curves for objects in the TESS FFIs. Here, we describe the methods used in eleanor to produce light curves that are optimized for planet searches. The tool performs background subtraction, aperture and PSF photometry, decorrelation of instrument systematics, and cotrending using principal component analysis. We recover known transiting exoplanets in the FFIs to validate the pipeline and perform a limited search for new planet candidates in Sector 1. Our tests indicate that eleanor produces light curves with significantly less scatter than other tools that have been used in the literature. Cadence-stacked images, and raw and detrended eleanor light curves for each analyzed star will be hosted on MAST, with planet candidates on ExoFOP-TESS as Community TESS Objects of Interest (CTOIs). This work confirms the promise that the TESS FFIs will enable the detection of thousands of new exoplanets and a broad range of time domain astrophysics.

binaries: eclipsing, methods: data analysis, planets and satellites: detection, techniques: image processing, techniques: photometric

0000-0002-9464-8101]Adina D. Feinstein

0000-0001-7516-8308]Benjamin T. Montet \altaffiliationSagan Fellow

0000-0002-9328-5652]Daniel Foreman-Mackey

0000-0001-9907-7742]Megan E. Bedell

0000-0003-4733-6532]Jacob L. Bean

0000-0002-8035-4778]Jessie L. Christiansen

0000-0002-3385-8391]Christina Hedges

0000-0002-0296-3826]Rodrigo Luger

0000-0002-4934-5849 ]Daniel Scolnic

1 Introduction

The recently retired Kepler and K2 missions (Borucki et al., 2010; Howell et al., 2014) revealed tremendous new insight into the frequency and architectures of exoplanetary systems. It is therefore timely that TESS, the Transiting Exoplanet Survey Satellite (Ricker et al., 2015), is already detecting new transiting planets (Cloutier, 2018; Dragomir et al., 2019; Esposito et al., 2018; Gandolfi et al., 2018; Günther et al., 2019; Huang et al., 2018; Huber et al., 2019; Jones et al., 2018; Kostov et al., 2019; Quinn et al., 2019; Rodriguez et al., 2019; Trifonov et al., 2019; Vanderspek et al., 2019; Wang et al., 2019).

The TESS prime mission is a two year survey observing roughly 80% of the sky for exoplanet transits. TESS’s four cameras are aligned along a 96 degree sector of the sky and are observed for approximately 27 days. There are 20,000 stars pre-selected by TESS mission operators and through Guest Investigator (GI) proposals observed every sector in a short 2-minute cadence mode. These targets have been selected not only for new exoplanet candidate searches, but also for asteroseismic studies, Solar System research, and additional galactic and extragalactic astrophysics.111See https://heasarc.gsfc.nasa.gov/docs/tess/approved-programs.html for examples

In addition to the short cadence photometry of the main targets, TESS obtains images, known as Full-Frame Images (FFIs), of each sector at 30-minute cadence. There are roughly one million stars in the FFIs brighter than I=16 mag in each sector of observations. As such, the FFIs provide a huge data mining archive for many different areas of astronomy, including the search for new transiting exoplanet candidates. Simulations from Barclay et al. (2018) predict that within the FFIs there will be an additional 3,100 detectable exoplanets orbiting bright (Tmag 11.0) stars, and a further 10,000 detectable exoplanets orbiting fainter stars. Within the FFIs, the authors project 1,500 large planets (R 4 R) and 400 small planets (R 4 R) will be identified, with 67% of the planets orbiting F and G type stars (Barclay et al., 2018).

Asteroseismology, the study of stellar oscillations to probe the internal structure of stars, will additionally benefit from the TESS FFIs. The ability to measure stellar oscillations with long cadence data has been previously demonstrated with the Kepler 30-minute cadence data (Campante et al., 2016). There is also the possibility to study Solar System objects, and galactic and extragalactic sources using the FFI data. Using a difference imaging approach and K2 long-cadence data, Dimitriadis et al. (2018) were able to obtain a light curve for a supernova, SN2018oh, roughly 52.7 Mpc away. The light curve begins 18 days before peak brightness, which is a feat that cannot be achieved by even the most advanced surveys that are triggered by supernova events. In addition to supernovae, teams such as Molnar et al. (2015) were able to detect extragalactic RR Lyrae stars in Leo IV, a dwarf galaxy at a distance of 154 kpc. Using K2 observations, they observed the farthest measurement of the Blazhko effect, or long-period modulations in the period and amplitude of the light curve. This was the first discovery of the effect outside of the Milky Way and the Magellanic Clouds.

Despite their substantial scientific potential, there is significant processing that needs to be completed before extracting usable light curves from the FFIs. A background correction can be approximated over the entire FFI, however, it would not properly account for regions with more localized issues (see Figure 1). On the FFI scale, systematic effects can overwhelm astrophysical signals, especially when the telescope is near perigee. Additionally, the FFIs are not in a user-friendly format. Each FFI is 35 MB. In order to complete photometry for a single target in a single sector, the user must have access to 45 GB of storage for any given sector and 1 TB for the entire Southern Hemisphere. This makes it challenging for users without vast computational resources to fully exploit the FFIs.

Figure 1: An example FFI from Sector 1, Camera 4. There is noticeable structured background in the corner of the CCDs 3 and 4 as well as in the center of CCDs 2 and 4. The Large Magellanic Cloud is seen in CCDs 1 and 2.

The recently ended K2 mission motivated the creation of several community driven pipelines for data reduction. There is significant benefit to having multiple pipelines with different methods for data reduction for the same data sets. For example, one could find a new planet candidate in the EVEREST light curves (Luger et al., 2018) and compare to the K2SFF light curves (Vanderburg & Johnson, 2014) to determine if the signal is real. Additionally, Shaya et al. (2015) created the Kepler Extra-Galactic Survey (KEGS), with the goal of producing light curves for extra-galactic sources. These pipelines were especially useful due to the large systematics found within the K2 data. Other pipelines, including K2VARCAT (Armstrong et al., 2015), K2SC (Aigrain et al., 2015), and POLAR (Barros et al., 2015) created light curves for the public to use as well, each with their own methods for removing systematics, and therefore their own strengths and weaknesses.

In this article, we present the eleanor pipeline222https://github.com/afeinstein20/eleanor for light curve extraction from the TESS FFIs and publicly available eleanor light curve data products. In Section 2, we describe the methods used to create our light curves. In Section 3, we demonstrate the photometric capabilities of eleanor by presenting early science results, including recovering known transiting planets, new planet candidates, and other stellar variability. In Section 4, we discuss the light curve data products and their availability, and how to download the open-source software package.

2 Creating Light Curves

In this section, we describe how eleanor extracts light curves from the FFIs. We first create a pointing model and assign quality flags on the FFI level. Then, we cut out intermediate “postcards” (148 104 pixels) that are time-stacked and background-subtracted. The Target Pixel Files (TPFs; 13 13 pixels) are extracted from the postcards. eleanor tests multiple apertures to find the best light curve for transiting exoplanet searches for each target. The TPF pixel time-stacked cut-out and light curves are stored in the eleanor data products, which are described in 2.5.

2.1 Pointing Model

We first download all of the FFIs for a given sector. Within the FFIs, we build a pointing model to ensure a true position of the star on the detector. Due to spacecraft motion, the World Coordinate System (WCS) written in the headers of the FFIs may not return an accurate transformation from pixel space to sky position. For each CCD, we complete a search of targets in the TESS Input Catalog (TIC, version 7.0; Stassun et al., 2018) with 7.5 mag 12.5 and 0 contamination 5 10-3, leaving us with bright, but not saturated, uncrowded stars to calibrate our pointing model.

We then determine the affine transformation that minimizes the square of the differences between the predicted and observed positions for all the stars that met our search criteria on the detector at each cadence. The affine transformation accounts for any difference between the predicted and observed stellar positions, such as a rotation or translation of the position of the spacecraft, or changes in apparent position of the stars due to differential velocity aberration. The predicted position of the stars are determined using the RA and Dec from the TIC and the WCS solution given in the FFIs. Corrected pixel positions, using the pointing model, of the stars are saved in the eleanor data product.

2.2 From FFIs to Postcards

Before extracting light curves, we create intermediate data products called “postcards” that represent a more efficient format for analyzing single targets with FFI data. Postcards are 148 104 pixel cut-out regions of the FFIs, and are created with a 50 pixel overlap between each postcard to avoid edge effects for individual stars. Unlike the FFIs, the postcards are time-stacked, including all cadences for which observations are available, and are background-subtracted.

As the background of the FFIs is highly structured and varies greatly across the detector (Figure 1), the more localized scaling of the postcards provides a sufficient region for initial backgroud subtraction. We use a constant background on the postcard level with the photutils function MMMBackground, which calculates a background of

(1)

for each cadence in the postcard. TPFs and light curves are extracted from the background-subtracted postcards.

Within each postcard, the WCS from each FFI is conserved. The postcards also contain quality flags to highlight potentially corrupted cadences. We follow a two-step process for assigning the quality flags.

2.3 Quality Flags

The TESS mission assigns twelve different quality flags, eight of which are applicable to the FFIs. The quality issues are: attitude tweaks; the spacecraft is in coarse point; the spacecraft is in Earth point; an argabrightening event occurs; reaction wheel desaturation event occurs; a cosmic ray is detected on a collateral pixel row or column; there is stray light from the Earth or Moon in the camera field-of-view; or a “manual exlude” set in the processing of short cadence data (for more information see Table 28 in Tenenbaum & Jenkins, 2018). We copy these quality flags into our postcards by identifying short-cadence targets that fall on each camera-CCD pairing for a given sector. There are roughly 15 short-cadence observations for every FFI observation, and we follow the most conservative procedure: any applicable quality flag that falls during an individual FFI exposure is recorded in the postcards, using the same numeric identifiers for each individual quality flag as applied in the short-cadence data.

Furthermore, we introduce our own quality flag based on the pointing model. We fit a line to the measured and pixel coordinates with the applied pointing model and complete an iterative sigma clipping at , see Figure 2. The stars in Figure 2 represent the bad pointing model cadences. We fit each orbit independently and assign a quality flag value of 4096. We apply the bad pointing model quality flag value to cadences that have quality flags from the short cadence data as well.

Figure 2: An example of quality flags marked by significant shifts in , pixel positions given by the pointing model (PM) for Sector 1. A line is fit to each orbit (corresponding to the different colors). Stars represent flagged data points offset from the pointing model from each orbit of the appropriate color. Black points represent quality flags taken from the short cadence data that apply to the entire FFI. There is some overlap between our marked quality flags and those from the short cadence data.

2.4 From TPFs to Light curves

A Target Pixel File (TPF) is cut out for each target from the postcard. The default size for eleanor TPFs is 13 13 pixels. The target is at the center of the TPF when possible. Photometry is completed on the TPF level. Due to the 50-pixel overlap between each postcard, targets may fall on multiple postcards; we extract the TPF and light curve from the postcard in which the target is closest to the center.

2.4.1 Aperture Selection

Once the TPF has been extracted, eleanor uses a pre-defined library of apertures of various shapes and sizes (see Figure 3) to measure photometry. We test apertures that were shown to work well for Kepler photometric extraction, including 2 1 rectangles and an ‘L’ shape, both in four different orientations about the center. We also test standard circular and square apertures defined using the photutils package. We use both binary (pixel values of either 0 or 1) and weighted (pixel values ranging from 0 to 1) masks when extracting photometry. The weights for the non-binary apertures are determined by the exact fractional overlap of the aperture and each pixel, dictated by photutils.aperture_photometry(). Circular apertures have radii of 1.25, 2.5, 3.5, and 4 pixels. Squares have lengths and widths of 3, 4.1, and 5 pixels, as well as a 3 pixel length and width that is rotated 45. We chose these orientations to maximize diversity in the aperture testing. All apertures and associated extracted light curves are saved in the eleanor data product.

Figure 3: The default library of apertures eleanor uses to create light curves. We try a range of shapes and sizes as well as assigning varying weights to each pixel. The binary apertures have pixel values of 0 and 1; the weighted apertures have pixel values from 0 to 1. All apertures are applied and tested unless crowded_field = True, in which case only apertures with aperture pixel sums less than 9 are used to extract light curves.

A user of the eleanor software also has the option to define their own aperture masks as well, in the eleanor.TargetData.get_lightcurve() function. Masks are required to be 2D arrays of the same length and height as the TPF. There is an additional option for users to customize a photutils aperture using eleanor.TargetData.custom_aperture(). This is a function that allows the user to choose the radius or length and width and angle of a circular or rectangular aperture. Users can also define a position in TPF pixel space to offset the aperture from the center, where it is placed by default. The custom apertures can additionally be defined as binary or weighted.

The photometry is completed by multiplying the aperture and the TPF and summing the pixel product in each cadence. We define this as “RAW_FLUX”. Note that the “RAW_FLUX” is background-subtracted as the pixels were extracted from a background-subtracted postcard. After the raw photometry for each aperture is extracted, we correct for possible systematic effects on an orbit-by-orbit basis, creating a flux time series called “CORR_FLUX”. We regress the raw flux time series against a linear model of the pixel position, pixel position (both taken from our pointing model), measured background at the location of the TPF, and time, effectively removing any signals correlated with these parameters. We note that long timescale astrophysical signals, such as starspot-induced stellar variability for slowly rotating stars, or transients like supernovae could produce an approximately linear signal over a single orbit. This signal would then be removed by “CORR_FLUX” in a similar way as it was removed by the Kepler pipeline in the generation of their pre-search data conditioning (PDCSAP) flux.

We test if additional background subtraction on the TPF level will increase the precision of the extracted light curve. The amount of background subtraction is marked in the header by the flag “BKG_LVL”, which reads as either “PC_LEVEL” indicating the background subtraction on the postcard level was used or “TPF_LEVEL” indicating the TPF-level background subtraction was applied. The background is again defined using the photutils function, MMMBackground, but only on the 13 13 pixel scale of the TPF, or other-sized region set by the end user.

After a light curve has been extracted using every aperture shape and size and the light curve has received background treatment on both the postcard and postcard and TPF level, an ideal aperture is chosen for that target. Here, we are primarily interested in creating the best light curve possible for detecting planets, preserving the sharp features on short timescales induced by planet transits. In order to achieve this, we minimize the combined differential photometric precision (CDPP) of the light curve on one-hour timescales. The CDPP is a metric in units of parts-per-million (ppm) that was originally defined for Kepler to assess the ability to detect a weak terrestrial planet transit signal in a light curve (Christiansen et al., 2012). By definition, the CDPP is the root mean square of the photometric noise on transit timescales (Jenkins et al., 2010). By choosing to minimize the CDPP in our light curves, we are optimizing them for planet searches. The CDPP is calculated for the “CORR_FLUX” light curve and is flattened using lightkurve.flatten(), which applies a 2nd order Savitzky-Golay filter to the light curve. The data are binned on one hour timescales and the CDPP is evaluated using the built-in function lightkurve.flatten().calculate_cdpp(). The filter is only applied to calculate the CDPP, and not used in the creation of the “CORR_FLUX” light curves saved in the eleanor data product. The associated aperture and level of background subtraction for the minimum CDPP light curve are identified in the header of the eleanor data product.

2.4.2 Principal Component Analysis

The eleanor package also performs principal component analysis (PCA) to remove any additional systematics that are still potentially present and shared between nearby stars on the detector. PCA is a machine learning technique that calculates orthogonal eigenvectors between a set of input vectors, such as light curves. We use the python package sklearn (Pedregosa et al., 2011) to perform PCA in eleanor. PCA is the basis of the development of PDCSAP light curves in the Kepler mission (Smith et al., 2012; Stumpe et al., 2012).

We develop our eigenvectors by collecting 12,000 stars across each camera. We use the “CORR_FLUX” flux from each star as inputs to identify the first 16 principal components. The components are visually inspected as a quality check. The user has the ability to check the components themselves as well. The eleanor light curve products include a “PCA_FLUX”, which is created by subtracting the first 8 principal components for the appropriate camera to minimize the possibility of astrophysical variability being imprinted in an eigenvector. All 16 components are available for more artisinal analyses of light curves; the option to use more or fewer components is built into eleanor.TargetData().pca(). The PCA components are not stored in the eleanor data products, but are stored on a forward-facing server that users have access to.

2.4.3 Point-Spread Function Modeling

We also include modeling of the point spread function of TESS as an option in eleanor, stored as “PSF_FLUX.” At present, this system models a uniform background level across the TPF, as well as an arbitrary number of Gaussians representing each star in the FOV. The relative positions of each Gaussian are set by the user, as the position of stars in the FOV is generally known to a very small fraction of a TESS pixel. The absolute position of the network of Gaussians is allowed to vary in each frame. We also fit a single width parameter in and as well as a single rotation angle. Each Gaussian has its own amplitude. We then maximize the likelihood value of each parameter conditioned on the data in each frame using the tensorflow interface to the scipy implementation of the Truncated Newton minimizer, with options to maximize either a Gaussian or Poissonian likelihood function,

A single Gaussian does not, across most of the detector, accurately represent the shape of the actual PSF, which can have extended, asymmetric wings especially on the corners of the detector. However, sums of Gaussians centered on each star could be used to more accurately model the shape of the PSF. Regardless, even this simple but fast PSF model provides, in many cases, superior precision to aperture photometry methods, as can be seen for the case of WASP-100 in Figure 4. More sophisticated PSF models will be useful to accurately model stars in relatively crowded fields, and will be a focus point for future development of eleanor.

2.5 The eleanor Data Product

The final data product is stored as a Flexible Image Transport System (FITS) file. Each FITS file contains the cadence-stacked background-subtracted flux pixels for a 13 13 region, centered on the source, and the cadence-stacked flux error pixels for the same region. The files contain all 21 aperture masks tested in the light curve extraction process as well as the raw and corrected fluxes for these apertures. For the automatically selected best aperture, three light curves are available: “raw” flux, “corrected” flux regressed against instrumental effects, and “PCA” flux with common modes subtracted. Users also have the option to create a PSF modeled light curve with the eleanor software package. However, the PSF flux is not a default light curve in the data product due to the relatively large processing time required. An example of each type of light curve can be seen in Figure 4, which additionally shows the recovery of the known planet WASP-100b (Hellier et al., 2014; Stassun et al., 2017).

Figure 4: An example of the four types of light curves that can be extracted and modeled using eleanor. Raw flux is the sum of pixels in the aperture. Corrected flux is raw flux with a linear regression as a function of pixel location, background, and time. PCA flux is a Principal Component Analysis subtracted flux, to remove common systematics between targets on the same camera. PSF modeled flux is the 2D Gaussian point-spread function modeled flux. The 10 transits for WASP-100 are clearly seen in all four light curves. The gaps in the light curves result from two different sources: The first gap is the result of data down-link and re-pointing to the same position on the sky. The second gap is the result of the telescope losing fine-pointing. The second gap has been masked by our quality flags.

In addition to photometric information, the produced files contain the and centroid positions, as inferred by our pointing model, and quality flags, based on the two-minute cadence targets and our own quality flag for cadences, as discussed previously. We create light curve files for sources in the TIC with I 16. This includes potentially saturated stars, which should be handled cautiously by the user. See Section 3.4 for more information on eleanor limitations. Members of the community can use the eleanor package to create light curves for fainter or extragalactic objects, or for a more detailed or optimized analysis of individual objects.

3 Results

We calculated the CDPP for 32,000 light curves in each of TESS’s four cameras in Sector 1. The CDPP as a function of TESS magnitude is shown in Figure 5. We subdivide into CCDs to demonstrate how light curves extracted from regions with more significant background effects (see Camera 4, CCD 1 in Figure 1) are affected.

The CDPP remains fairly consistent for all CCDs in a given camera, with the exception of Camera 4. CCD 4 experiences noticeably more systematics than the other CCDs, leading to an overall increase in CDPP values. Note that the presence of the Large Magellanic Cloud in CCDs 1 and 2 does not lead to a significant difference in CDPP values when compared to other cameras; while diffuse light from the LMC is obvious, it is stable.

Figure 5: The CDPP as a function of TESS Magnitude for 32,000 stars on each camera. Stars are colored by which CCD they fall on. Camera 4 (see Figure 1) experiences significantly more background features than the other three cameras, leading to an overall increase in CDPP. The colors correspond to the CCD. We provide guiding lines at Tmag = 12 and CDPP = 103 ppm to allow for easier comparison between CCDs. CCDs with less background have lower CDPP values than those with more prominant background features.

3.1 Comparison to Other Pipelines

We compare the eleanor light curve CDPP to that of Oelkers & Stassun (2019) and the MIT Quick-Look Pipeline (QLP) to assess its performance in remove instrumental and astrophysics systematics. The QLP light curves are not available to the public except for a small number of published, confirmed planets; for comparison we use the published light curve of TOI-172 b (Rodriguez et al., 2019). We apply the eleanor quality flags across all light curves for a uniform comparison. We apply additional masks for times when the observatory went out of its fine-pointing mode (1338 Time 1339 and 1346.8 Time 1348.6) and for the 3 transits of TOI-172 b, marked in Figure 6 by the vertical orange lines, to quantify the level of scatter in the light curve out of transit. We show the transits in Figure 6 to demonstrate eleanor’s transit detection capabilities. Each light curve has been flattened using identical processes.

For TOI-172, a host star, the Oelkers & Stassun (2019) light curve has CDPP = 579 ppm; the QLP light curve has CDPP = 376 ppm; and the eleanor light curve has CDPP = 325 ppm. The smaller scatter in the eleanor light curves will enhance the community’s ability to detect small planet transit signals in the FFIs.

Figure 6: A comparison of results from using three separate tools to extract light curves from the TESS FFIs: Oelkers & Stassun (2019, OS19; green), MIT’s Quick-Look Pipeline (QLP; blue), and eleanor (purple). Each light curve received the same quality flags and flattening treatment when calculating the CDPP. The transits were removed from the calculation and are highlighted in the figure to demonstrate that all three pipelines recover it. TIC29857954 is a T = 10.71 host star. The eleanor light curve has the lowest CDPP of the three pipelines.

3.2 Recovery of Known Planets

In order to further demonstrate the quality of the eleanor light curves, we recover a few known transiting planets that were observed in Sector 1. We show the light curves for four of these planets in Figure 7. We use the batman (Kreidberg, 2015) transit fitting software package, following the methods from Mandel & Agol (2002), to derive planet parameters from the eleanor light curves and compare our results with the known parameters. The planet parameters we derived are consistent with values quoted in Maxted et al. (2016) and Stassun et al. (2017) for the WASP planets, and with Huang et al. (2018) and Gandolfi et al. (2018) for Mensae c and are quoted in Table 1.

We complete a more extensive Markov Chain Monte Carlo (MCMC) analysis for WASP-126b (Maxted et al., 2016) using emcee (Foreman-Mackey et al., 2013), an implementation of the affine-invariant ensemble sampler of Goodman & Weare (2010). We initialized our MCMC run with the best-fit parameters from a single batman fit and ran it for 200 steps, to complete a burn-in. The parameters from the 200th run were then used as the starting point for the next run of 1500 steps. The parameters and uncertainties are quoted in Table 1. All system parameters derived with the eleanor light curve fall within 1 of the accepted parameters from Maxted et al. (2016), with the exception of , which agrees within 1.5. Overall, the derived parameters for WASP-126 correspond well to parameters in the literature.

3.3 New Science with eleanor Light Curves

We demonstrate the potential for new science to come out of the TESS FFIs using eleanor by performing a limited search for periodic signals in Sector 1 data. Using the 12,000 stars generated for the PCA components, we performed a search using the box-least squares (BLS; bls.py333https://github.com/dfm/bls.py) module in astropy to phase fold light curves and identify new periodic candidates, ranging from new planet candidates to RR Lyrae stars. We searched a period range of 0.5 to 8 days and recorded the candidates where the maximum peak in the periodogram (period vs. log likelihood) was more than 9 above the mean.

The identified candidates with periodic signals were then vetted by-eye. The candidates that passed this check can be found in Table 2. We present the properties from the BLS fits and uncertainties for new planet candidates and eclipsing binaries. This is an incomplete list of new exoplanet candidates identified with eleanor. A full list of candidate signals will be included in future work. An example of the light curves with candidate periodic signals can be seen in Figure 8.

We note that Sullivan et al. (2015) concluded that there will be roughly 1000 false positives within the 2-minute targets. Barclay et al. (2018) conducted an additional analysis of the false positive rate for the FFIs and concluded the false positive rate should increase from 1 false positive per 180 stars in the 2-minute targets to nearly 5 times that for the FFIs, but could increase to nearly 11 times that, depending on the parameter space probed for planet transits. Therefore, further vetting is required for the candidates identified in this work. All planet candidates identified with eleanor will be hosted on ExoFOP-TESS as Community TESS Objects of Interest (CTOIs).

In addition, we demonstrate the use of eleanor to explore extragalactic astrophysics by recovering known supernovae that occurred in Sectors 1 and 2 (Figure 9). SN2018fhw and SN2018exc are classified as Type Ia supernovae; SN2018eph is classified as a Type II supernova; MOA 2018-LMC-002 is an unknown astrophysical event; MOA 2018-LMC-003 is a known microlensing event towards the Large Magellanic Cloud. By having light curve information before the triggering of the supernova event, the photometric information from TESS can be used to infer information about the supernova progenitor. Furthermore, uniform high-cadence light curves will be useful to search microlensing events for the signatures of planetary companions to the lensing stars.

Figure 7: An example of the recovery of four known planets in the FFIs. We plot the normalized eleanor PSF-modeled flux with our quality flags. Transit fits modeled with batman are shown by the solid black lines in the phase-folded subplots. Retrieved system parameters are quoted in Table 1. The data gaps correspond to those described in Figure 4
Figure 8: Six examples of new periodic astrophysical signals identified in the TESS FFIs using eleanor. The first four rows represent previously unidentified planet candidates; the fifth row represents a new contact binary; the sixth row represents a previously known RR Lyrae. We completed a simple BLS fit to each light curve to obtain the time from transit center, period, and depth, which are presented in Table 2. This population represents a small subset of sources, which will all be published in later work and available for the community to use.
Figure 9: Example of five recovered transient events in TESS Sectors 1 and 2. SN2018fhw and SN2018eph are known Type Ia supernovae; SN2018exc is a known Type II supernova; MOA 2018-LMC-003 is a previously identified microlensing event within the LMC; the event causing the light curve for MOA 2018-LMC-002 is unknown. Even though the eleanor light curves are tailored towards finding new transiting exoplanets, the software can be used for other astrophysical detections.

3.4 Software Limitations

There are several cases in which eleanor light curves may produce non-optimal results. The first is the presence of moving solar system objects. We do not perform any corrections for when Mars saturated a significant fraction of one detector in Sector 1 or when asteroids pass through an aperture for a given target (see Figure 10 as an example). An asteroid induces a clear spike in the raw light curve and its location can be traced as it passes in front of different stars on this postcard as a function of time. For stars lying in or near the ecliptic plane, we recommend spikes in any light curve be checked to ensure they are not a foreground Solar System object.

The saturated stars on the detector are beyond the intended scope of eleanor. Although we create eleanor data products for these targets, we recommend the user look to see if the target has been observed at 2-minute cadence (which is likely), create a custom aperture larger than the apertures in the eleanor default library (Fig. 3), or use a halo aperture approach (White et al., 2017). Additionally, PSF modeling of saturated stars will produce poor results as the PSF model is not an accurate representation of the behavior of the star on the detector in these cases.

In the case of crowded fields, we limit the default apertures to only the smallest possibilities to mitigate the possibility of a nearby star contaminating the aperture too significantly. In these cases, it is recommended to investigate the aperture and the locations of other nearby stars to ensure the reliability of the light curve. For additional information on other spacecraft issues and warnings that may affect an extracted eleanor light curve, see the TESS Science Data Products Description (Tenenbaum & Jenkins, 2018).

Figure 10: An example of an asteroid crossing through a light curve for three stars on the same postcard. There is a clear increase in flux as the asteroid passes through the aperture. The TPF and associated light curves are identified by the same color. The crossing occurs before and after the data downlink gap between the two orbits during Sector 1.

4 Data Availability and Software Tools

For each sector in the Southern Hemisphere, we will create and release eleanor data products (See 2.5) to the community. After TESS completes all observations in the southern hemisphere, we will reprocess all light curves for a uniform library. This also allows us to potentially improve earlier light curves as we continue to learn about new methods to remove the background, determine the pointing of the spacecraft, and manage crowded regions of the detector. The data will be available as a high level science product (HLSP) at the Mikulski Archive for Space Telescopes (MAST).

In addition to our light curve products, eleanor is an open-source package that can be downloaded through GitHub or via the Python Package Index. eleanor was designed with the user in mind. Users have the ability to create TPFs and light curves for any source or position on the detector by passing in a TIC ID, Gaia ID, or RA and Dec coordinates. When creating a new TPF, eleanor will go through all of the steps described above for the user’s request.

Beyond the standard steps described in Section 2, there are several features the user can customize if they make a light curve with the eleanor software. Users can set the size (height width) of the TPF eleanor extracts. The height and width are forced to be odd numbers to allow the target to be at or very near the center of the cutout. Users can call do_psf = True when initializing the light curve object, which allows for a PSF modeled light curve in the output. Additionally, the user has the ability to set the region around a source that they want to base the background subtraction on. The default background size is the size of the TPF. For the publicly available light curve products, this will be pixels (see Appendix A for syntax examples in Python using eleanor).

After each sector is observed, we will search for new planet candidates within the FFIs. Light curves that have been identified as planet candidates will be hosted on ExoFOP-TESS. We will release a catalog of new planet candidates as well as other interesting astrophysical events, such as eclipsing binaries and RR Lyraes. This information will be open to the community. Although we are predominantly interested in finding new planet candidates, our hope is these light curves will yield scientific discoveries across many branches of astrophysics, including supernovae characterization and identification of stellar oscillations, to name a few. The open-source eleanor products will be available for a diverse set of scientific discoveries that will be achievable using the TESS FFIs.

We thank Doug Caldwell, Michael Fausnaugh, Jon Jenkins, and Roland Vanderspek for valuable discussions. We thank Patrick Vallely for his direction to recovering known supernovae in the TESS field of view. Work by B.T.M. was performed under contract with the Jet Propulsion Laboratory (JPL) funded by NASA through the Sagan Fellowship Program executed by the NASA Exoplanet Science Institute. We would like to acknowledge the live animal webcams of the Kansas City Zoo, and especially the emotional support of Elvis the emperor penguin during our hack weeks. This work was funded in part through the NASA TESS Guest Investigator Program, as a part of Program G011237 (PI Montet). \softwarenumpy (Van Der Walt et al., 2011), matplotlib (Hunter et al., 2007), scipy (Jones et al., 2001–) tensorflow (Abadi et al., 2015), lightkurve444https://doi.org/10.5281/zenodo.2557026, photutils555https://doi.org/10.5281/zenodo.2533376, astropy(Astropy Collaboration et al., 2013; Price-Whelan et al., 2018), eleanor666https://doi.org/10.5281/zenodo.2597620, sklearn(Pedregosa et al., 2011) TESS

Appendix A eleanor Software Demonstration

This represents an example of the basic commands for eleanor light curves for one target given a known TIC ID. The user has the ability to set the sector they wish to create a light curve for, if the target has been observed in multiple sectors.

1# Import the eleanor package
2import eleanor
3
4star = eleanor.Source(tic=38846515, sector=1)
5
6# PSF Modeling is a function that needs to explicitly be set to True
7# Default is no PSF modeled flux
8data = eleanor.TargetData(star, do_pca=True, do_psf=True)
9
10# Calling eleanor quality flags
11q = data.quality == 0
12
13# Calling the different flux options
14raw_flux  = data.raw_flux[q]
15corr_flux = data.corr_flux[q]
16pca_flux  = data.pca_flux[q]
17psf_flux  = data.psf_flux[q]
18
19# Light curve with the minimum CDPP
20best_aperture = data.aperture
21
22# Saving your TPF file and light curves
23data.save()

Users can additionally create eleanor light curves based on a given set of coordinates that have been observed in multiple sectors.

1# Calling a light curve for a set of coordinates
2from astropy.coordinates import SkyCoord, Angle
3from astropy import units as u
4
5ra = Angle(68.959732, u.deg)
6dec = Angle(-64.02704 , u.deg)
7
8# Obtaining target information for multiple sectors
9# Returns a list of eleanor.Source() objects
10stars = eleanor.multi_sectors(coords=SkyCoord(ra=ra, dec=dec), sectors=[1,2])
11
12# Extracting light curves
13data0 = eleanor.TargetData(stars[0])
14data1 = eleanor.TargetData(stars[1])
15
16# Creating a custom circular aperture of radius 1
17eleanor.TargetData.custom_aperture(data0, shape=’circle’, r=1)
18
19# Recreates default light curve with new aperture
20eleanor.TargetData.get_lightcurve(data0)
21
22# Creating a PCA Model; sets data0.psf_flux
23eleanor.TargetData.pca(data, flux=corr_flux, modes=4)
24
25# Creating a PSF Model; sets data1.psf_flux
26eleanor.TargetData.psf_lightcurve(data1, model=’gaussian’, likelihood=’poisson’)
27
28# Available light curves
29# No PSF modeled flux was created for this eleanor object
30raw_flux0 = data0.raw_flux # with new aperture
31corr_flux0 = data0.corr_flux # with new aperture
32pca_flux0 = data0.pca_flux
33
34# No PCA corrected flux was created for this eleanor object
35raw_flux1 = data1.raw_flux
36corr_flux1 = data1.corr_flux
37psf_flux1 = data1.psf_flux

References

  • Abadi et al. (2015) Abadi, M., Agarwal, A., Barham, P., et al. 2015, TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems, , , software available from tensorflow.org. https://www.tensorflow.org/
  • Aigrain et al. (2015) Aigrain, S., Hodgkin, S. T., Irwin, M. J., Lewis, J. R., & Roberts, S. J. 2015, MNRAS, 447, 2880
  • Armstrong et al. (2015) Armstrong, D. J., Kirk, J., Lam, K. W. F., et al. 2015, A&A, 579, A19
  • Astropy Collaboration et al. (2013) Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, A33
  • Barclay et al. (2018) Barclay, T., Pepper, J., & Quintana, E. V. 2018, ApJS, 239, 2
  • Barros et al. (2015) Barros, S. C. C., Almenara, J. M., Demangeon, O., et al. 2015, MNRAS, 454, 4267
  • Borucki et al. (2010) Borucki, W. J., Koch, D., Basri, G., et al. 2010, Science, 327, 977
  • Campante et al. (2016) Campante, T. L., Lund, M. N., Kuszlewicz, J. S., et al. 2016, ApJ, 819
  • Christiansen et al. (2012) Christiansen, J. L., Jenkins, J. M., Caldwell, D. A., et al. 2012, PASP, 124, 1279
  • Cloutier (2018) Cloutier, R. 2018, arXiv e-prints, arXiv:1812.08145
  • Dimitriadis et al. (2018) Dimitriadis, G., Foley, R. J., Rest, A., et al. 2018, ApJ, 870
  • Dragomir et al. (2019) Dragomir, D., Teske, J., Gunther, M. N., et al. 2019, arXiv e-prints, arXiv:1901.00051
  • Esposito et al. (2018) Esposito, M., Armstrong, D. J., Gandolfi, D., et al. 2018, arXiv e-prints, arXiv:1812.05881
  • Foreman-Mackey et al. (2013) Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman, J. 2013, PASP, 125, 306
  • Gandolfi et al. (2018) Gandolfi, D., Barragán, O., Livingston, J. H., et al. 2018, A&A, 619, L10
  • Goodman & Weare (2010) Goodman, J., & Weare, J. 2010, Communications in Applied Mathematics and Computational Science, 5, 65
  • Günther et al. (2019) Günther, M. N., Pozuelos, F. J., Dittmann, J. A., et al. 2019, arXiv e-prints, arXiv:1903.06107
  • Hellier et al. (2014) Hellier, C., Anderson, D. R., Cameron, A. C., et al. 2014, MNRAS, 440, 1982
  • Howell et al. (2014) Howell, S. B., Sobeck, C., Haas, M., et al. 2014, PASP, 126, 398
  • Huang et al. (2018) Huang, C. X., Burt, J., Vanderburg, A., et al. 2018, ApJ, 868, L39
  • Huber et al. (2019) Huber, D., Chaplin, W. J., Chontos, A., et al. 2019, arXiv e-prints, arXiv:1901.01643
  • Hunter et al. (2007) Hunter, J. D., et al. 2007, Computing in science and engineering, 9, 90
  • Jenkins et al. (2010) Jenkins, J. M., Chandrasekaran, H., McCauliff, S. D., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7740, Software and Cyberinfrastructure for Astronomy, 77400D
  • Jones et al. (2001–) Jones, E., Oliphant, T., Peterson, P., et al. 2001–, SciPy: Open source scientific tools for Python, , , [Online; accessed ¡today¿]. http://www.scipy.org/
  • Jones et al. (2018) Jones, M. I., Brahm, R., Espinoza, N., et al. 2018, arXiv e-prints, arXiv:1811.05518
  • Kostov et al. (2019) Kostov, V. B., Schlieder, J. E., Barclay, T., et al. 2019, arXiv e-prints, arXiv:1903.08017
  • Kreidberg (2015) Kreidberg, L. 2015, PASP, 127, 1161
  • Luger et al. (2018) Luger, R., Kruse, E., Foreman-Mackey, D., Agol, E., & Saunders, N. 2018, AJ, 156, 99
  • Mandel & Agol (2002) Mandel, K., & Agol, E. 2002, ApJ, 580, L171
  • Maxted et al. (2016) Maxted, P. F. L., Anderson, D. R., Collier Cameron, A., et al. 2016, A&A, 591, A55
  • Molnar et al. (2015) Molnar, L., Pal, A., Plachy, E., et al. 2015, ApJ, 812
  • Oelkers & Stassun (2019) Oelkers, R. J., & Stassun, K. G. 2019, Research Notes of the American Astronomical Society, 3, 8
  • Pedregosa et al. (2011) Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825
  • Price-Whelan et al. (2018) Price-Whelan, A. M., Sipőcz, B. M., Günther, H. M., et al. 2018, AJ, 156, 123
  • Quinn et al. (2019) Quinn, S. N., Becker, J. C., Rodriguez, J. E., et al. 2019, arXiv e-prints, arXiv:1901.09092
  • Ricker et al. (2015) Ricker, G. R., Winn, J. N., Vanderspek, R., et al. 2015, Journal of Astronomical Telescopes, Instruments, and Systems, 1, 014003
  • Rodriguez et al. (2019) Rodriguez, J. E., Quinn, S. N., Huang, C. X., et al. 2019, arXiv e-prints, arXiv:1901.09950
  • Shaya et al. (2015) Shaya, E. J., Olling, R., & Mushotzky, R. 2015, AJ, 150, 188
  • Smith et al. (2012) Smith, J. C., Stumpe, M. C., Van Cleve, J. E., et al. 2012, PASP, 124, 1000
  • Stassun et al. (2017) Stassun, K. G., Collins, K. A., & Gaudi, B. S. 2017, AJ, 153, 136
  • Stassun et al. (2018) Stassun, K. G., Oelkers, R. J., Pepper, J., et al. 2018, AJ, 156, 102
  • Stumpe et al. (2012) Stumpe, M. C., Smith, J. C., Van Cleve, J. E., et al. 2012, PASP, 124, 985
  • Sullivan et al. (2015) Sullivan, P. W., Winn, J. N., Berta-Thompson, Z. K., et al. 2015, ApJ, 809, 77
  • Tenenbaum & Jenkins (2018) Tenenbaum, P., & Jenkins, J. M. 2018. https://archive.stsci.edu/missions/tess/doc/EXP-TESS-ARC-ICD-TM-0014.pdf
  • Trifonov et al. (2019) Trifonov, T., Rybizki, J., & Kürster, M. 2019, A&A, 622, L7
  • Van Der Walt et al. (2011) Van Der Walt, S., Colbert, S. C., & Varoquaux, G. 2011, Computing in Science & Engineering, 13, 22
  • Vanderburg & Johnson (2014) Vanderburg, A., & Johnson, J. A. 2014, PASP, 126, 948
  • Vanderspek et al. (2019) Vanderspek, R., Huang, C. X., Vanderburg, A., et al. 2019, ApJ, 871, L24
  • Wang et al. (2019) Wang, S., Jones, M., Shporer, A., et al. 2019, AJ, 157, 51
  • White et al. (2017) White, T. R., Pope, B. J. S., Antoci, V., et al. 2017, MNRAS, 471, 2882
Planet P i e Reference
WASP-126b 0.0776 3.28880 7.634 87.9 0.18 Maxted et al. (2016)
WASP-126b 0.076 3.28692 6.080 86.972 0.183 This work
WASP-95b 0.1024 2.1847 6.51 88.4 0.0 Stassun et al. (2017)
WASP-95b 0.1255 2.1841 6.52 88.3 0.0 This work
WASP-124b 0.12 3.3727 9.434 86.3 0.017 Maxted et al. (2016)
WASP-124b 0.14 3.3721 9.290 86.6 0.07 This work
Men c 0.017 6.26790 13.38 87.456 0.0 Huang et al. (2018)
Men c 0.017 6.26834 13.10 87.31 0.0 Gandolfi et al. (2018)
Men c 0.018 6.27187 13.08 87.207 0.009 This work
Table 1: Transit Properties of Recovered Planets
TIC ID RA Dec T Period Depth Disposition
(J2000) (J2000) (BTJD-2457000) (Days)
234503282 00:46:22.92 -63:28:23.07 1336.1498 0.0008 1.4376 0.0002 0.0171 0.0003 Planet candidate
234504626 00:47:45.69 -62:25:23.28 1335.5641 0.0013 0.5341 0.0043 0.0035 0.0002 Planet candidate
299780329 02:30:07.20 -79:45:23.23 1335.5707 0.0022 1.6022 0.0074 0.0008 0.0001 Planet candidate
394340319 02:37:27.78 -79:49:22.90 1336.3655 0.0024 3.0371 0.0009 0.0068 0.0005 Planet candidate
350844139 06:00:07.22 -57:19:24.02 1335.5270 0.0025 0.5709 0.0001 0.0038 0.0002 Planet candidate
350930938 06:02:40.74 -54:50:10.85 1335.5272 0.0026 2.6489 0.0013 0.0257 0.0009 Planet candidate
260304296 06:18:27.76 -57:00:48.85 1335.5179 0.0005 0.51261 0.00001 0.0072 0.0007 Planet candidate
349155660 07:13:24.42 -63:59:18.69 1335.1362 0.0014 0.69792 0.00002 0.0137 0.0006 Planet candidate
349832804 07:34:34.61 -64:55:30.45 1335.2914 0.0013 1.04140 0.00002 0.0102 0.0005 Planet candidate
300810086 07:47:15.36 -69:01:50.52 1335.4133 0.0016 0.66797 0.00007 0.0009 0.0004 Planet candidate
159835004 21:21:00.62 -40:42:46.90 1335.4052 0.0011 0.51606 0.00003 0.0058 0.0002 Planet candidate
139771134 21:36:30.74 -52:30:46.85 1335.2118 0.0011 0.91331 0.00002 0.0055 0.0003 Planet candidate
38813184 04:30:37.51 -62:16:01.48 1340.4600 0.0022 5.3516 0.0020 0.0123 0.0013 Eclipsing binary
231090180 04:35:54.59 -66:08:01.20 1335.5791 0.0011 1.2545 0.0002 0.0129 0.0007 Eclipsing binary
260003467 06:04:19.90 -57:18:09.50 1337.3377 0.0016 2.6046 0.0004 0.0043 0.0004 Eclipsing binary
349480507 07:23:44.80 -65:00:39.38 1335.2111 0.0028 1.5628 0.0004 0.1741 0.0075 Eclipsing binary
349575582 07:27:56.92 -64:23:25.67 1335.8312 0.0031 2.1383 0.0008 0.0936 0.0050 Eclipsing binary
350091587 07:40:30.45 -61:20:51.62 1335.9237 0.0063 0.98006 0.00006 0.0027 0.0003 Eclipsing binary
159834934 21:20:47.97 -40:54:39.78 1335.8908 0.0059 2.7720 0.0029 0.0182 0.0012 Eclipsing binary
53896097 21:52:59.01 -24:47:50.20 1335.2516 0.0022 0.74080 0.00004 0.0111 0.0005 Eclipsing binary
301941187 21:57:14.13 -23:58:04.85 1337.0845 0.0063 7.6676 0.0014 0.0111 0.0017 Eclipsing binary
121490917 22:54:24.27 -46:38:39.17 1336.3423 0.0020 2.4652 0.0045 0.0478 0.0017 Eclipsing binary
293525767 23:28:02.47 -73:38:44.76 1336.0060 0.0034 6.8491 0.0035 0.0290 0.0028 Eclipsing binary
234508527 00:50:16.11 -63:18:27.53 1335.18 1.231 0.173 Contact binary
38907808 04:37:47.57 -65:16:33.97 1335.28 0.413 0.120 Contact binary
349762067 07:30:33.93 -64:43:35.04 1335.31 0.858 0.029 Contact Binary
29577927 20:58:45.47 -28:58:18.16 1335.15 0.904 0.117 Contact binary
44625047 22:45:21.52 -46:49:41.88 1335.57 0.808 0.059 Contact Binary
261560715 23:16:07.06 -74:01:53.05 1335.13 0.692 0.049 Contact binary
293507177 23:26:10.70 -73:23:49.92 1335.56 1.140 0.010 Contact binary
349647488 07:28:38.71 -64:20:56.37 1338.82 3.687 0.336 Young stellar object
349647697 07:28:37.26 -64:38:11.58 1.104 0.011 Stellar variability
29716907 21:02:13.27 -28:50:33.83 0.916 0.015 Stellar variability
277874877 23:31:53.19 -73:01:33.11 1.521 0.010 Stellar variability
234507163 00:50:00.62 -62:38:07.70 0.415 0.489 RR Lyrae
29752683 21:03:16.64 -29:33:07.96 1.465 0.172 RR Lyrae
293526531 23:29:26.71 -72:31:57.86 1.097 0.071 RR Lyrae

Note. – The uncertainties from the depth are taken directly from the data; dilution from nearby sources would change the of the presented signals.

Table 2: BLS Properties of New Signal Candidates
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
""
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
   
Add comment
Cancel
Loading ...
354783
This is a comment super asjknd jkasnjk adsnkj
Upvote
Downvote
""
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters
Submit
Cancel

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test
Test description