3.1.

Sky Coverage

Although many transiting planets have been found, relatively few of them orbit stars bright enough to enable follow-up measurements of planet masses and atmospheres. Therefore, a prime objective of TESS is to monitor bright stars. Since the brightest stars are nearly evenly distributed over the entire sky, this desire led in the direction of an all-sky survey.

3.2.

Period Sensitivity

Ideally, the mission would detect planets with a broad range of orbital periods, with $P_{\min }$ near the Roche limit of a few hours, and $P_{\max }$ of a year or longer. However, the choice of $P_{\max }$ determines the mission duration and thereby has a disproportionate influence on cost. Furthermore, transit surveys are strongly and inherently biased toward short periods. (As shown in Refs. 7 and 8, in an idealized imaging survey limited only by photon-counting noise, the effective number of stars that can be searched for transiting planets of period $P$ varies as $P^{-5/3}$.) Indeed, the period distribution of Kepler detections² reaches a maximum at $\approx 10\text{days}$, taking into account the period dependencies of both the transit detection efficiency and the occurrence of planets. Hence, even a value of $P_{\max }$ as short as 10 days should yield many discoveries.

Choosing $P_{\max }$ as short as 10 days rules out the detection of habitable-zone planets around Sun-like stars. However, it is possible to gain access to habitable-zone planets around M stars if at least a portion of the sky is observed with $P_{\max }\gtrsim 40\text{days}$. (For an M0 star with $T_{\mathrm{eff}}=3800\mathrm{K}$, the range of orbital distances receiving a bolometric insolation within a factor of two of the Earth’s insolation corresponds roughly to $P=25–75$ days.) And if there is a portion of the sky to be searched more extensively, it would be advantageous for that portion to coincide with the zones of longest continuous visibility with the James Webb Space Telescope (JWST), which will likely be the best telescope for follow-up studies of planetary atmospheres. Those zones are centered on the ecliptic poles.⁹

Together these considerations led to the choices of $P_{\max }\approx 10\text{days}$ in general, and $P_{\max }\gtrsim 40\text{days}$ for as large an area as possible surrounding the ecliptic poles.

3.3.

Type of Stars Monitored

While it would be interesting to survey all types of stars, it is logical to concentrate on main-sequence dwarfs with spectral types F5 to M5. Evolved stars and early-type dwarfs are large, inhibiting the detection of small planets. Dwarfs with spectral types earlier than F5 also rotate rapidly, broadening their spectral lines and preventing precise radial-velocity (PRV) monitoring.

On the other side of the spectral sequence, M dwarfs are especially attractive targets. They are abundant: about three-quarters of the stars in the solar neighborhood are M0–M5 dwarfs. They are relatively unexplored for transiting planets, because they constituted only a small minority of the Kepler target list. Furthermore, the transit signal of a small planet is easier to detect for an M dwarf than it would be for a larger star of the same apparent magnitude, facilitating both planet discovery and follow-up observations with JWST and other telescopes.

However, stars with spectral types later than M5 are rarer and optically faint. They could be observed advantageously at near-infrared wavelengths, but this would greatly increase the mission’s cost, complexity, and risk. Furthermore, planets transiting the very latest-type stars can be detected with ground-based instruments, as demonstrated by the MEarth survey.¹⁰ For these reasons, the F5–M5 range of spectral types was considered to be most important for TESS.

3.4.

Detector and Bandpass

The best astronomical detectors in the optical band are silicon charge-coupled devices (CCDs), which have excellent linearity and dynamic range for bright objects, and a long spaceflight heritage. A wide bandpass is preferred to reduce photon-counting noise. However, very wide bands present challenges in constructing a wide-field and well-focused optical system. Enhanced sensitivity to red wavelengths is desired because small planets are more easily detected around small stars, which are cool and red.

These considerations led to the choice of a 600 to 1000 nm bandpass. The red end represents the red limit of silicon CCD sensitivity, and the width of 400 nm was the largest practical choice for the optical design. This bandpass is centered on the traditional $I_{\mathrm{C}}$ band but is much wider (see Fig. 1). It is comparable to the union of the $R_{\mathrm{C}}$, $I_{\mathrm{C}}$, and $z$ bands.

Fig. 1

The Transiting Exoplanet Survey Satellite (TESS) spectral response function (black line), defined as the product of the long-pass filter transmission curve and the detector quantum efficiency curve. Also plotted, for comparison, are the Johnson–Cousins $V$, $R_{\mathrm{C}}$, and $I_{\mathrm{C}}$ filter curves and the Sloan Digital Sky Survey $z$ filter curve. Each of the functions has been scaled to have a maximum value of unity.

3.5.

Number of Stars Monitored

Kepler’s transit detection rate² for “super-Earths” ($1.25-2R_{\oplus }$) with $P<10\text{days}$ was about 0.2%. This represents the product of the occurrence rate ($\sim 5\%$) and the geometric transit probability ($\sim 5\%$). Therefore, TESS needs to monitor at least 500 stars for each super-Earth detection. The desire to find hundreds of such planets leads to the requirement to monitor $\gtrsim {10}^5$ stars.

According to Galactic star count models,¹¹ the brightest ${10}^5$ dwarf stars have a limiting apparent magnitude of $I_{\mathrm{C}}\approx 10$. In fact, the TESS design employs a larger catalog ($\gtrsim 2\times {10}^5$ stars) selected according to transit detectability rather than strictly on apparent magnitude. This leads to a limiting magnitude of $I_{\mathrm{C}}\lesssim 10-13$, depending on spectral type. The faint M dwarfs, for example, have such small sizes that they can still be usefully searched for small planets.

3.6.

Aperture Size

Stars must be monitored with sufficient photometric precision to detect planetary transits. This leads to requirements on the collecting area and other properties of the cameras, although the derivation of those requirements is not straightforward because of the wide variation in the characteristics of the transit signals. As reviewed in Ref. 1, the transit depth varies with the radii of the star ($R_{\star }$) and planet ($R_{\mathrm{p}}$)

Monte Carlo simulations were performed¹² to establish the requirements for detecting a certain number of transiting planets, taking into account the likely distribution of stellar apparent magnitudes and radii; the occurrence of planets of various sizes and orbits; as well as noise due to sky background, instrumental readout, and other sources. The results of this study, summarized in Sec. 8, indicate that detecting hundreds of super-Earths requires $\approx 50{\mathrm{cm}}^2$ of effective collecting area (i.e., the actual collecting area multiplied by all throughput factors including transmission losses and quantum efficiency). This corresponds to an effective aperture diameter of $D\approx 10\mathrm{cm}$.

3.7.

Time Sampling

Since the typical transit durations are measured in hours, the time sampling of the photometric observations should be substantially less than an hour. The timescale of the partial transit phases (ingress and egress) is $\sim T(R_{\mathrm{p}}/R_{\star })$, which is typically several minutes for planets smaller than Neptune. Therefore, to avoid excessive smearing of the partial transit signals, a time sampling of a few minutes or shorter is preferred. For TESS, the brightness measurements of the preselected stars will be recorded every 2 min or shorter, which is consistent with these requirements and also enables asteroseismology (see Sec. 8.3).

3.8.

Orbit

The ideal orbit for the TESS spacecraft would allow for continuous observations lasting for weeks, and would be stable to perturbations over the multiyear duration of the survey. It would offer a low-radiation environment, to avoid high trapped-particle fluxes and resulting degradation of the CCDs and flight electronics. It would also offer a stable thermal environment and minimal attitude disturbance torques, to provide a stable platform for precise photometry. The orbit would be achievable with a moderate $\mathrm{\Delta }V$, avoiding the need for a costly secondary propulsion unit. Furthermore, to facilitate data transfer, the spacecraft would be close to the Earth during at least a portion of the orbit. As explained in Sec. 6, these requirements were met by choosing an elliptical orbit around the Earth, with approximate perigee and apogee distances of 17 and $59R_{\oplus }$, and a period of 13.7 days.

4.1.

Lens Assembly

Each of the four identical TESS lenses is an $f/1.4$ custom design consisting of seven optical elements, with an entrance pupil diameter of 10.5 cm (see Figs. 2 and 3 and Table 1). For ease of manufacture, all lens surfaces are spherical except for two mild aspheres. There are two separate aluminum lens barrels that are fastened and pinned together. All optical elements have antireflection coatings. The surface of one element also has a long-pass filter coating to enforce the cutoff at 600 nm. The red limit at 1000 nm is set by the quantum-efficiency curve of the CCDs (see Fig. 1).

Fig. 2

Diagram of the lens assembly, charge-coupled device (CCD) focal plane, and detector electronics for one of the four TESS cameras.

Fig. 3

(a) Two lens prototypes were constructed during phase A. One was subjected to thermal vacuum testing at the operational temperature; the other was subjected to vibration testing. (b) The detector assembly of one of the prototype lenses. The frame-store regions of the CCDs are covered.

Table 1

Characteristics of the TESS lenses. Ensquared energy is the fraction of the total energy of the point-spread function that is within a square of the given dimensions centered on the peak.

Each lens forms a $24\mathrm{deg}\times 24\mathrm{deg}$ unvignetted image on the four-CCD mosaic in its focal plane. The optical design was optimized to provide small image spots of a consistent size across the field of view (FOV), and produce undersampled images similar to those of Kepler.¹³ At nominal focus and flight temperature ($-{75}^{°}\mathrm{C}$), the 50% ensquared-energy half-width is $15\mu \mathrm{m}$ (1 pixel or 0.35 arc min) averaged over the FOV. Each lens is equipped with a lens hood, which reduces the effects of scattered light from the Earth and Moon.

4.2.

Detector Assembly

The focal plane consists of four back-illuminated MIT/Lincoln Laboratory CCID-80 devices. Each CCID-80 consists of a $2048\times 2048$ imaging array and a $2048\times 2048$ frame-store region, with $15\times 15\mu \mathrm{m}\text{pixels}$. The frame-store region allows rapid shutterless readout ($\approx 4\mathrm{ms}$). On each CCD there are four regions of 512 columns, each of which has a dedicated output register. The full well capacity is approximately 200,000 electrons. The four CCDs are abutted with a 2-mm gap, creating an effective $4096\times 4096$ imaging array contained within a $62\mathrm{mm}\times 62\mathrm{mm}$ square (see Fig. 3).

The CCDs have an operational temperature of $-{75}^{°}\mathrm{C}$ to reduce the dark current to a negligible level. They are read out at 625 kHz and the read noise is $<10e^{-}$. The detector electronics are packaged onto two compact double-sided printed circuit boards, each 12 cm in diameter, which are located beneath the CCD focal plane and transmit digitized video data over a serial LVDS link to a DHU.

4.3.

Data Handling Unit

The TESS DHU is a Space Micro Image Processing Computer (IPC-7000) which consists of six boards: an IPC, which contains two Virtex-7 field-programmable gate arrays (FPGAs) that serve as interfaces to the four cameras and perform high-speed data processing; a Proton 400K single board computer, which is responsible for commanding, communicating with the spacecraft master avionics unit, and interfacing with the Ka-band transmitter; two 192 GB solid-state buffer (SSB) cards for mass data storage; an analog I/O power switch board to control instrument power; and a power supply board for the DHU.

The CCDs produce a continuous stream of images with an exposure time of 2 s. These are received by the FPGAs on the IPC, and summed into consecutive groups of 60, giving an effective exposure time of 2 min. During science operations, the DHU performs real-time processing of data from the four cameras, converting CCD images into the data products required for ground postprocessing. A primary data product is a collection of subarrays (nominally $10\times 10\text{pixels}$) centered on preselected target stars. The Proton 400 K extracts these subarrays from each 2 min summed image, compresses them, and stores them in the SSB prior to encapsulation as CCSDS packets for the Ka-band transmitter. Full frame images (FFIs) are also stacked every 30 min and stored in the SSB. Data from the SSB are downlinked every 13.7 days at perigee.

7.1.

Star Selection

A catalog of preselected stars will be monitored at a cadence of 2 min or better. Ideally, the catalog will include $\gtrsim \mathrm{200,000}$ main-sequence FGKM stars that are sufficiently bright and small to enable the detection of transiting planets smaller than Neptune. This corresponds to approximately $I_{\mathrm{C}}<12$ for FGK stars and $I_{\mathrm{C}}<13$ for M dwarfs.¹² These stars are bright enough that it should be possible to assemble a list using currently available all-sky surveys. A draft catalog has been constructed using the 2MASS,¹⁵ Tycho-2,¹⁶ and UCAC4¹⁷ catalogs, supplemented with smaller catalogs of nearby M dwarfs. For the M stars, infrared magnitudes allow for discrimination between dwarfs and giants.¹⁸ G and K dwarfs are distinguished from giants through their relatively rapid reduced proper motion.¹⁹ It may also be possible to use trigonometric parallaxes measured by the ongoing European “Gaia” mission, if they become available early enough for TESS mission planning. The TESS target catalog will be publicly released prior to launch.

7.2.

Scanning Strategy

The four cameras act as a $1\times 4$ array, providing a combined FOV of $24\mathrm{deg}\times 96\mathrm{deg}$ or $2300{\mathrm{deg}}^2$ (see Fig. 7). The north and south ecliptic hemispheres are each divided into 13 partially overlapping sectors of $24\mathrm{deg}\times 96\mathrm{deg}$, extending from an ecliptic latitude of 6 deg to the ecliptic pole. Each sector is observed continuously for two spacecraft orbits (27.4 days), with the boresight of the four-camera array pointed nearly antisolar. After two orbits, the FOV is shifted eastward in ecliptic longitude by about 27 deg, to observe the next sector. Observing an entire hemisphere takes 1 year, and the all-sky survey takes 2 years. (A video illustrating the TESS survey strategy, along with the pathway to the spacecraft orbit, can be seen at http://www.youtube.com/watch?v=mpViVEO-ymc.).

Fig. 7

(a) The instantaneous combined field of view of the four TESS cameras. (b) Division of the celestial sphere into 26 observation sectors (13 per hemisphere). (c) Duration of observations on the celestial sphere taking into account the overlap between sectors. The dashed black circle enclosing the ecliptic pole shows the region which James Webb Space Telescope will be able to observe at any time.

The overlap of the sectors is illustrated in Fig. 7. Approximately $\mathrm{30,000}{\mathrm{deg}}^2$ are observed for at least 27 days. Close to the ecliptic poles, approximately $2800{\mathrm{deg}}^2$ are observed for more than 80 days. Surrounding the ecliptic poles, approximately $900{\mathrm{deg}}^2$ are observed for more than 300 days.

7.3.

Data Downlink and Housekeeping Operations

At perigee, science operations are interrupted for no more than 16 h to point TESS’s antenna toward Earth, downlink data, and resume observing. This includes a nominal 4-h period for Ka-band science data downlink using NASA’s Deep Space Network. In addition, momentum unloading is occasionally needed due to the $\approx 1.5\mathrm{N}\mathrm{m}$ of angular momentum build-up induced by solar radiation pressure. For this purpose, TESS uses its hydrazine thrusters.

7.4.

Ground-Based Data Analysis and Follow-Up

The TESS data will be processed with a data reduction pipeline based on software that was developed for the Kepler mission.²⁰ This includes pixel-level calibration, background subtraction, aperture photometry, identification and removal of systematic errors, and the search for transits with a wavelet-domain matched filter.

Once the data are processed and transits are identified, selected stars will be characterized with ground-based imaging and spectroscopy. These observations are used to establish reliable stellar parameters, confirm the existence of planets, and establish the sizes and masses of the planets. Observations will be performed with committed time on the Las Cumbres Observatory Global Telescope Network and the MEarth observatory. In addition, the TESS science team members have access to numerous other facilities (e.g., Keck, Magellan, Subaru, HARPS, HARPS-North, Automated Planet Finder) through the usual telescope time allocation processes at their home institutions. The TESS team includes a large group of collaborators for follow-up observations and welcomes additional participation.

8.1.

Photometric Performance

Figure 8 shows the anticipated photometric performance of the TESS cameras. The noise sources in this model are photon-counting noise from the star and the background (zodiacal light and faint unresolved stars), dark current (negligible), readout noise, and a term representing additional systematic errors that cannot be corrected by cotrending. The most important systematic error is expected to be due to random pointing variations (“spacecraft jitter”). Because of the nonuniform quantum efficiency of the CCD pixels, motion of the star image on the CCD will introduce changes in the measured brightness, as the weighting of the image PSF changes, and as parts of the image PSF enter and exit the summed array of pixels.

Fig. 8

(a) Expected $1\sigma$ photometric precision as a function of stellar apparent magnitude in the $I_{\mathrm{C}}$ band. Contributions are from photon-counting noise from the target star and background (zodiacal light and unresolved stars), detector read noise ($10e^{-}$), and an assumed 60 ppm of incorrigible noise on hourly timescales. (b) The number of pixels in the photometric aperture that optimizes the signal-to-noise ratio.

The central pixel of a stellar image will saturate at approximately $I_{\mathrm{C}}=7.5$. However, this does not represent the bright limit for precise photometry because the excess charge is spread across other CCD pixels and is conserved, until the excess charge reaches the boundary of the CCD. As long as the photometric aperture is large enough to encompass all of the charge, high photometric precision can still be obtained. The Kepler mission demonstrated that photon-noise–limited photometry can be obtained for stars 4 mag brighter than the single-pixel saturation limit.²¹ Since similar performance is expected for TESS, the bright limit is expected to be $I_{\mathrm{C}}\approx 4$ or perhaps even brighter.

8.2.

Transit Detections

Monte Carlo simulations are used to verify that the science objectives can be met and to anticipate the properties of the detected planetary systems.¹² These simulations are based on a model of the local neighborhood of main-sequence FGKM stars.¹¹ Simulated stars are populated randomly with planets, and “observed” by TESS. Those for which transits are observed with a sufficiently high signal-to-noise ratio are counted as detections. In addition, the simulated star catalog is populated with eclipsing binaries that may be blended with brighter stars to produce transit-like signals; the detections of these “false positives” are also recorded.

Among the features of the current simulations are: (1) a realistic distribution of stars and eclipsing binaries based on local census data; (2) probability distributions for planetary occurrence and orbital properties taken from the Kepler results;³ (3) variation in stellar surface density and zodiacal light with position on the celestial sphere; (4) variation in the duration of TESS observations depending on ecliptic coordinates.

Figure 9 illustrates some of the results. TESS is expected to find thousands of planets smaller than Neptune, including hundreds of super-Earths ($1.25-2R_{\oplus }$) and tens of planets comparable in size to Earth. These will be accompanied by a comparable number of false positives (as has been the case for the Kepler mission), a majority of which are background eclipsing binaries. Some false positives will be identifiable using TESS data alone, based on the detection of secondary eclipses, ellipsoidal flux modulation, or transit-specific image motion. In other cases, ground-based observations will be required to check for composite spectra, large radial-velocity variations, color-dependent transit depths, and resolved companions that are indicative of false positives.

Fig. 9

Sizes and orbital periods of planets with host stars brighter than $J=10$. The $J$ band was chosen for convenience, since the 2MASS survey provides $J$ magnitudes for all of the known planet-hosting stars. (a) Currently known planets, including those from the Kepler and CoRoT missions as well as ground-based surveys. (b) Including the simulated population of TESS detections.

8.3.

Asteroseismology

Observing photometric variations due to stellar oscillations (“asteroseismology”) provides sensitive diagnostics of the stellar mass, radius, and internal dynamics. Based on the Kepler experience with mode amplitudes as a function of stellar parameters,²² TESS can be expected to detect $p$-mode oscillations on about 6000 stars brighter than $V=7.5$, including (a) the majority of all stars brighter than $V=4.5$, (b) about 75 stars smaller than the Sun, (c) about 2000 upper-main-sequence and subgiant stars, and (d) virtually all the giant stars. Stars that are not suitable for planet searching but are appropriate for asteroseismology (such as giants) can be added to the TESS input catalog at minimal cost.

The key features enabling TESS’s advances in this area are all-sky coverage and relatively fine time sampling. Compared to similar stars studied by CoRoT and Kepler, the TESS stars will be more numerous, brighter, and nearer to the Earth. TESS stars will have accurately known parallaxes, will be much more amenable to interferometric and radial-velocity studies, and are themselves among the likely targets for future, imaging-based missions to study exoplanets.

8.4.

Full Frame Images

In addition to monitoring 200,000 preselected stars with a 2 min cadence, TESS will return a nearly continuous series of FFIs with an effective exposure time of 30 min. As Kepler has shown, a 30 min cadence still allows for the efficient discovery of transiting planets. The FFIs will expand the transit search to any star in the FOV that is sufficiently bright, regardless of whether it was selected ahead of time. This will reduce the impact of any imperfections in the target star catalog, and allow the search sample to extend beyond the FGKM dwarfs that are the focus of the mission. In addition, the transit candidates that are identified in the FFIs during the baseline TESS mission could be selected for shorter-cadence observations during an extended mission.

The FFIs will also enable a wide variety of other scientific investigations, such as the detection and monitoring or nearby supernovae and gamma-ray burst afterglows, bright AGN outbursts, near-Earth asteroids, eclipsing and close binaries, novae, flaring stars, and other variable stars. Each 30 min FFI will provide precise ($\approx 5\mathrm{mmag}$) photometry for approximately ${10}^6$ bright galaxies and stars ($I_{\mathrm{C}}<14-15$) within a $2300{\mathrm{deg}}^2$ field. Over 2 years, the TESS all-sky survey will provide such data for approximately 20 million bright objects, many of which will exhibit short-term variability. Thus, the TESS FFIs will provide broad-ranging fundamental data for time-domain astronomy. In particular, TESS will complement the Large Synoptic Survey Telescope, which is limited to observing objects fainter than 16th magnitude.²³