PLATO science requirements


See this document for a list of high level science requirements of PLATO, used as basic input for the current assessment study.

Basic observation strategy
The stellar sample
Surveyed field
Photometric noise level
Collecting area and overall throughput
Non photonic sources of noise
Duration of observations
Duty cycle
Colour information
Time sampling
Dynamical range





Summary of observational requirements


Basic observation strategy

The PLATO science objectives require long uninterrupted high precision photometric monitoring of large samples of stars. These observations will first allow us to detect and characterize planetary transits, allowing us to measure planet sizes and orbital periods, as well asto detect planet satellites and rings. They will provide us as well with measurements of frequencies, amplitudes and lifetimes of oscillation modes of the same sample of stars. The analysis of these asteroseismic measurements will yield precise information about the internal structure and rotation of these stars, and in particular will allow to determine accurately their masses, radii and ages.

The stellar sample

A first sample of PLATO targets will contain objects that are bright enough to reach photometric precision allowing us to use seismic tools and also to perform efficient ground-based follow-up, and still numerous enough to allow for a large number of planetary transit detections in spite of the low geometrical probability that the line of sight is favourably oriented with respect to the planet orbital plane. This geometrical probability is typically of the order of 1% for Sun-Earth analogs. As demonstrated e.g. in the framework of the CoRoT, Eddington, Kepler missions, a prerequisite for such observations is to survey a sample of about 100,000 stars, out of which about one half will be cool dwarfs, adequate for transit search programmes.

A scaling up of preliminary results of CoRoT shows that reasonably sized optical monitors, typically with 0.3 m2 collecting area, are capable of detecting and exploiting solar-like oscillations down to mV = 11. Such objects will also be efficiently observed from the ground in high resolution spectroscopy. Therefore, the primary stellar sample for PLATO will be a set of at least 100,000 stars brighter than mV = 11, observed at ultra-high precision.

In addition to this main sample, for which both transit search and asteroseismology will be performed, an extended sample of fainter stars needs to be observed, in order to extend the statistical analysis of exoplanetary transits. This second sample will include about 400,000 additional main sequence stars down to mV = 14, which will be observed to a sufficient precision to detect the transits of Earth-sized planets.

A third sample will include about 1,000 very bright and nearby stars, observed at ultra-high photometric precision, extending the brightest limit to about mV = 4.

Moreover, asteroseismic observations of stars that are members of open clusters, chosen to offer a complete sequence of ages, as well as old population II stars, will be of major interest, and such targets should definitely be included in the programme.

Surveyed field

We have examined existing stellar catalogues, in particular the 2MASS and USNO catalogues, in order to estimate the minimum field of view to survey to comply with the stellar sample requirements. Several regions were explored in the vicinity of the galactic plane. We find that the sky is very inhomogeneous at mV = 11, so that the field(s) to be monitored must be chosen with care. Typically, the most populated fields seem to be in the southern hemisphere. As an  example, the table below indicates the total number of stars counted from the 2MASS catalogue, down to mV = 11, 11.5, and 12 in one such  typical field. An estimate of the number of stars down to mV = 14 is also given, based on star counts from the USNO-B1 catalogue in several smaller regions within the same field.  From these star counts, we conclude that the minimum required field would be of the order of 1000 deg2 in total.

Number of stars counted in fields centered at ecliptic coordinates (210°, -60°).
The first row gives the total number of stars, while the second row indicates
an estimate of the number of cool dwarfs,  calculated using the fraction of cool dwarfs
from the Besançon model (Robin et al. 2003).
Photometric noise level

The depth of a planetary transit is given by the ratio of the areas of the planet and its transited star, which is of the order of in the case of Sun-Earth analogs, while transit durations are typically of the order of 12 hours. In order to detect such transits at more than 4 sigma, a dimensioning requirement, we need to obtain a photometric noise level lower than about 2.5 x 10-5 in about 12 hours, i.e. about 8 x 10-5 in one hour. However, this requirement must be considered as the minimum one, since the detection of planets with sizes smaller than the Earth's, the detection of Earth-sized planets in front of stars hotter than the Sun, and the measurement of several points across the transits would be of major interest. We will set as a goal specification a photometric noise level of 2.7 x 10-5 in one hour, allowing us for instance to measure about ten points across the transits, and therefore to characterize them with high reliability and precision. Such a low noise level would also allow us to detect Earth-sized planets in front of 2 solar radii stars, which are in principle less active and therefore less intrinsically variable than solar twins, so that the detection of such exoplanets may in fact be easier than for solar-Earth twin systems; we would therefore extend to hotter stars the statistics of terrestrial exoplanets.

The figure below shows simulated light curves with various levels of noise, all including a 12 hour transit with a depth of 10-4, illustrating the absolute need for a noise level better than 8 x 10-5 in one hour for detecting such a transit and the potential for precise transit characterization at a noise level of the order of 2.5 x 10-5 in one hour.


Simulated light curves assuming various noise levels, and including
one single transit with a depth of 10-4 and a duration of 12 hours.
With a noise level of 2.5 x 10-5 in one hour, the transit is seen with
no ambiguity and can be characterized, even at its first occurrence;
the reliability of detection and characterization decreases with increasing
noise level. A noise of 8 x 10-5 in one hour is the maximum acceptable
for a transit detection at first occurrence (click figure to enlarge).

The dimensioning requirement for seismic investigations is to detect and characterize oscillations of solar-type stars, which are the lowest amplitude oscillators. Ground-based spectroscopic observations have recently revealed solar-like oscillations for a handful of bright stars, whose amplitudes, when translated in terms of photometric variations, range from a few ppm to a few tens of ppm. These conclusions have been recently confirmed by preliminary results from CoRoT.

The corresponding requirements on the photometric noise require that we should be able to detect individual p-mode oscillations in cool dwarf stars as faint as mV=11. This means that the total photometric noise in the observations must remain below 1 ppm after approximately 30 days of observation, down to mV = 11, in the whole frequency range of interest, i.e. from 0.02 to 10 mHz. The figure below shows a portion of the simulated power spectrum for a 1.2 solar mass star observed for 50 days with such a photometric noise, as well as that expected for brighter stars leading to lower noise levels. The spectrum was generated with a simulator built in preparation for CoRoT, including photon noise and stellar granulation noise, and its results have been successfully compared to the early results of CoRoT.


Expected power spectrum for a 1.2 solar mass star, observed for 50 days, with a noise level of
1 ppm in 30 days (lower curve). Such a noise level will typically be obtained at mV=11.5 or 12 with
PLATO, depending on the details of the instrumental design. The next two curves correspond
to better noise levels of 0.5 and 0.25 ppm in 30 days, achievable with PLATO at mV= 10 and 8.5, respectively.
Note that the oscillation spectrum, clearly visible at low noise levels, is only barely visible on the lower one,
indicating that 1 ppm in 30 days is indeed the upper limit for the specified noise level. The upper
curve shows a preliminary result obtained with CoRoT in 50 days on a solar-type star with mV=5.7,
which translates to mV = 8 for PLATO when the ratio of collecting areas and efficiencies is taken into account.
This CoRoT result (see official CoRoT release) demonstrates that the noise levels introduced in the simulation are realistic
and that solar-like oscillations can indeed be easily observed in photometry from space (click figure to enlarge).
The noise levels requirements of 2.7 x 10-5 in 1 hour and 1 ppm in 30 days are very similar. These photometric noise requirements impose:
- that the photon flux of the  target stars is sufficiently high to ensure that photon noise complies with the final noise specifications, and
- that all other sources of noise remain well below photon noise in the frequency range of interest.

The required photon fluxes can be translated into requirements on optics collecting area and overall throughput described in the following section.


Collecting area and overall throughput

The above noise requirements, applied for stars with mV = 11, imply a photo-electron flux of the order 3.8 x105 sec-1 at this magnitude, which requires the use of an optical design with a high overall throughput and a large effective collective area.

The overall optics + detector throughput can reasonably be expected of the order of 0.6-0.7 in average over a bandpass of 500 nm, using high quantum efficiency detectors, such as thinned backside illuminated CCDs, as well as efficient, optimally coated optics. Under these assumptions, it can be calculated that an effective total collecting area of 0.3 m2 is necessary to yield the required photo-electron flux down to mV = 11.

Non photonic sources of noise

We must ensure that additional sources of noise remain significantly below photon noise. Until a full analysis of all environmental and instrumental perturbations is carried out, we will specify that any given source of noise must remain at least three times below photon noise for stars in our prime sample of stars, i.e. down to  mV=11 - 12. We provide below some hints concerning the expected two major sources of perturbations, satellite jitter and thermal instability, but a thorough study of all possible other sources of noise must be carried out when designing the instrument and the mission during assessment, phase-A and B1.

Satellite jitter introduces photometric noise, as drifts in the star's position will change the illumination pattern of the detector, whose pixels do not have exactly the same sensitivity. While a large fraction of the induced noise can be calibrated out, as demonstrated by CoRoT results, a residual noise will still be present, and impose pointing stability requirements. At mV= 11 - 12, the photon noise will be of the order of 2 x 10-4 per minute, and jitter noise must remain about three times below this value, i.e. below 6 x 10-5 per minute. A detailed study, depending on the instrument design, is necessary to figure out how this requirement translates into pointing stability specifications, but order of magnitude estimates, as well as experience gained with previous missions such as CoRoT, indicate that a pointing stability of about 0.2 arcsec on timescales between one minute and 15 hours will be needed.

Variation in temperature of both the telescopes and the CCD detectors will lead, directly or indirectly, to changes in the apparent signal level from the target stars. The top level requirement for these perturbations is that the equivalent noise level induced by thermal changes must be three times below the photon noise level across the magnitude range of interest. While detailed a posteriori corrections can drastically improve the resulting noise, as demonstrated by CoRoT preliminary results, a full study taking into account the details of the instrument design and environmental perturbations must be carried out in order to specify clearly the thermal stability requirements. Again, as a first estimate, we can take advantage of the CoRoT experience, and require that the chip temperature must be stable to 0.1 K/hr, while a thermal stability of a few tenths of a K per day will be required for the telescope structure.

Duration of observations

The duration of the observations on each one of the two initial fields needs to be longer than about three years, so that at least three consecutive transits for Sun-Earth analogs can be detected during the mission lifetime.

For the seismic analysis of the target stars, the total monitoring time must be at least 5 months, so that the inversion of the oscillation spectra is sufficiently accurate to provide the star mass within a few percent in a reliable way.


Duty cycle

The probability that N transits of the same planet are observed is given by pN = (df)N, where df is the fractional duty cycle of the instrument. In order to achieve an 80% probability that all transits of a three-transit sequence are observed, a duty cycle of 93% is needed, ignoring gaps that are much shorter than individual transits. The requirement for planet-finding is therefore that gaps which are longer than a few tens of minutes do not occur over more then 7% of the time, with a loss by gaps as small as 5% being desirable.

A similar requirement is also imposed for seismology. Gaps in the data produce sidelobes in the power spectrum, which make mode identification ambiguous. Periodic gaps in the data must be minimized, as they will produce the most severe sidelobes in the power spectra. It can be shown that periodic outages representing 5% of the total time produce aliases with a power of about 1.5% of that of the real signal. Such sidelobes are just acceptable, as they will remain within the noise for most of the stars observed. It is therefore required that periodic data gaps are below 5%, preferably 3%.

Non-periodic interruptions have a less catastrophic influence on the power spectrum, and can therefore be tolerated at a higher level, provided the time lost is compensated by a longer elapsed time for the observation. Random gaps in the data representing a total of 10% of the monitoring time yield sidelobes with a power lower than 1% of that of the real signal, which will be adequate for this mission. The requirement on random data gaps is therefore that they do not exceed 10% of the elapsed time.

Colour information

Intrinsic stellar variability due to activity or other stellar phenomena can lead to variations of the light curves that are similar to those produced by the transits of terrestrial exoplanets. Planet occultations can be discriminated from these intrinsic variations if several successive transits can be observed. In practice, planetary transits are definitely discriminated from intrinsic stellar variations only after three transits, so that the longest orbital period detectable is in average between one third and one half of the total monitoring time.

Light curves in several colours can improve significantly the situation. Planetary transits indeed are achromatic, while all intrinsic variability pattern show some degree of chromatism. Comparing the depths and shapes of the candidate transits in several colours will allow us to distinguish transits from stellar intrinsic variations. Similarly, colour information will allow to discriminate against false positives arising from of stellar configurations involving eclipsing binaries, whose vast majority generates a chromatic signal. The level of confidence for candidate transits will therefore be greatly enhanced at first occurrence, hence increasing significantly the range of reachable orbital periods.

In addition to the measurement of oscillation frequencies, asteroseismology requires the identification (l,m) of the detected modes. Knowing the l identification for the dominant modes of each of the 100,000 bright target stars of PLATO implies a significant reduction of the free parameter space of stellar models and is a requirement to guarantee successful seismic inference of their interior structure parameters and ages.

For oscillations in the asymptotic frequency regime, the derivation of frequency spacings suffices to identify the modes. For most main-sequence stars excited by the kappa mechanism, when the modes do not follow particular frequency patterns, the identification of l can be achieved by exploiting the difference in amplitude and phase of the mode at different wavelengths.

However, whatever the interest of colour information, we must not compromise photon noise, which is the most important requirement of this mission. The option to measure light curves in several colours must be brought at negligible cost in terms of photon loss.

Time sampling

The duration  of a transit of a planet with semi-majoraxis a and orbital period P in front of a star with radius R* is given by
. For true Earth analogs = 13 hours. More generally, the duration of a transit around a single star may last from about two hours (a ``hot giant'' planet around a low-mass star) to over one day, for planets on Jupiter-like orbits (5 AU distance). Planets in the habitable zone, however, will cause transits lasting between 5 hours (around M stars) and 15 hours (for F stars), for equatorial transits. Because individual transits have durations ≥ 2 hours, a time sampling of about 10 - 15 minutes is in principle sufficient to detect all types of transits, as well as to measure transit duration and period. However, a higher time resolution is needed in order to accurately time ingress and egress of the planet transits for which the S/N in the light curve will be sufficient. The accurate timing will allow the detection of third bodies, which cause offsets in transit times of a few seconds to about a minute, and will allow to solve ambiguities among possible transit configurations through the determination of ingress and egress time of the planet. In practice, a time sampling of about 30 sec will be necessary to analyse in such detail the detected transits.

The needed time sampling for the asteroseismology objectives can be derived directly from the frequency interval we need to explore, which is from 0.02 to 15 mHz. In order to reach 15 mHz, the time sampling must be at least twice this frequency, i.e. of the order of 30 sec.

In summary, a 30 sec sampling will be needed for the cool stars in the bright sample of 100,000 stars for precise transit analysis and asteroseismology, while a sampling of 10 minutes will be sufficient to search for transits in the fainter sample of 400,000 stars. For the small fraction of these faint stars for which a transit will have been detected, a 30 sec sampling will have to be applied.


Dynamical range

The planet search objectives of PLATO are based on the observation ofstars as faint as mV= 13 - 14, but stars as bright as mV= 4 - 5 also need to be monitored for the detection of stellar reflected light on planet atmospheres and also for astrometric detection of planets. Stars between mV=4 - 5 and mV= 11 - 12 also need to be observed at a sufficient level for seismic analysis.