HARPS-N White Paper

The HARPS-N Collaboration

HARPS-N is a collaboration between various institutes of different countries for the construction and operation of HARPS-N. The HARPS-N instrument is a high-precision spectrograph located in the Northern hemisphere. It was commissioned in 2012 and operates at the Telescopio Nazionale Galileo (TNG), located at the Spanish Roque de los Muchachos Observatory. Its location allows access to the northern survey elds of the NASA Kepler/K2 and TESS missions, and complements its southern counterpart HARPS on the 3.6-m ESO telescope in Chile in searches for rocky planets around bright stars.

Scientific Rationale for HARPS-N

The main scientific goal of the HARPS-N Collaboration is the discovery and characterization of terrestrial exoplanets by achieving long-term 1 meter per second median radial-velocity precision on V < 11-mag stars. HARPS-N is a high-resolution (R=115,000) optical spectrograph with broad wavelength coverage (378-691 nm). In its four years of operation to date it has demonstrated precision and stability comparable to its sister instrument HARPS on the ESO 3.6-m telescope. Its performance and stability are monitored by laser-comb calibration and a programme of daytime solar observations via a dedicated bre feed. In the rst ve years of the GTO agreement, the HARPS-N Collaboration has published masses for 14 super-Earth and sub-Neptune planets, representing about half of all planets smaller than 3.2 earth radii with mass determinations better than 30% precision (Buchhave et al. 2016, Christiansen et al. 2017, Dressing et al. 2015, Dumusque et al. 2014, Gettel et al. 2016, Lopez-Morales et al. 2016, Malavolta et al. 2017, Motalebi et al. 2015, Pepe et al. 2013, Vanderburg et al. 2015)

Primary Scientific Objectives

The following is a description of the projects the HARPS-N Collaboration plans to carry out during the 5-year period 2017-2022.

Primary Project - Synergy with Kepler/K2, TESS, CHEOPS

NASA's Kepler mission was launched in March 2009. Kepler monitored a single eld in Cygnus/Lyra (RA=19:23, Dec=+44.5) from December 2009 to July 2014. It has already identified over 2300 transiting planets confirmed by either followup or multiplicity, 114 of which orbit stars brighter than Kp = 12 mag. Following loss of 3-axis reaction-wheel pointing capability in 2014, the Kepler spacecraft was repurposed for a new mission named K2. K2 consists of a series of 16 3-month-long pointings in the ecliptic plane, where solar radiation pressure is sufficiently well-balanced to allow 2-wheel pointing. This 16-fold increase in survey area greatly increases the number of planet candi- dates orbiting stars bright enough to be followed up with HARPS-N. K2 targets will constitute the primary source of transiting super-Earth and mini-Neptune planets until planet candidates from the first year of the TESS mission in the accessible part of the southern ecliptic hemisphere become available. The K2 follow-up work has already begun with the 11.82+/-1.33 Earth-mass measurement of a super-Earth in a 9.1-day orbit about the V = 10 K1 dwarf HIP 116454 (Vanderburg et al 2015), and mass measurements of 5.03+/-0.38, and 9.8+/-1.3 Earth masses of the two transiting planets with periods of 0.96 and 29.8 days, and a non-transiting planets with a minimum mass of 6.90+/-0.71 Earth masses in a 8.5-day period around the nearby V = 8.9 K0 dwarf HD 3167 (Christiansen et al. 2017). We are currently observing about 10 more stars with planets identified by K2.

The Transiting Exoplanet Survey Satellite (TESS) is a NASA space mission to survey the entire sky to discover the nearest and brightest transiting exoplanet systems, namely the best ones for follow up with future missions such as the James Webb Space Telescope (JWST). TESS was selected for implementation and is in phase C/D, with a current working launch date in March 2018. TESS will survey the entire northern ecliptic hemisphere in its second year of operation. Each eld will be surveyed for at least 27 days, with coverage increasing toward the ecliptic pole.

The Swiss-led ESA S-class mission CHEOPS will be launched in 2018, and will serve as a photometric follow-up mission. Among its prime goals are to search for transits in bright systems with known RV planets, and to rene radius measurements for small planets discovered with TESS. HARPS-N will allow Doppler follow-up for the large number of Neptune-like and sub-Neptune planets that K2 and TESS will discover, and CHEOPS will characterize. Because the systems discovered by these missions will be much brighter than the vast majority of the Kepler planets, it will be possible to determine masses for smaller planets. The TESS/CHEOPS planets will be the best targets for future research on transiting planets, such as studies of the planetary atmospheres and weather with JWST. Both missions have expressed a strong interest in using HARPS-N for follow-up observations.

Primary Project Goals

The primary project of the HARPS-N Collaboration has three broad scientific goals that can be achieved by delivering planetary masses of certain precision:

A. characterizing Earth-like planets of 2-5 Earth masses ("super-Earths") in various orbits with enough precision to distinguish between volatile-rich and predominantly iron-silicate planets;

B. measuring accurate masses to a precision of 10% or better, for planets near 10 Earth masses, at which the transition between super-Earths and Ice Giants (hot Neptunes, for example) occurs.

C. confirming an Earth-twin planet in the habitable zone of a G5V star or later, with a precision of 30% in mass;

Our strategy to address goals B and C will involve careful selection of known transiting planets from K2 and TESS spanning the apparent 1.6 Earth-radius boundary below which rocky compositions appear to dominate (Rogers et al 2015). Lopez et al (2012, 2014) have shown that among close- orbiting planets, loss of volatiles via irradiation by the host star may determine whether a planet retains or loses its primary envelope. Our target selection strategy aims to sample planets on both sides of the 1.6 Earth-radius boundary at a variety of orbital distances and hence irradiation levels. HARPS-N will determine orbital periods for targets showing single transits within the 27-day span of a typical TESS observation, enabling CHEOPS to deliver precise measurements of subsequent transits. This will be important for exploration of the dependence of exoplanet radii on host-star metallicities in low-irradiation environments, and for extending the exploration of habitable zones to warmer and more massive host stars.

Secondary Project - Earth-Like Planets Around Nearby Stars

HARPS-N is capable of better than 40 cm/s precision and stability for very bright stars, as demonstrated by long-term monitoring of the solar spectrum with the HARPS-N solar telescope feed and laser comb system, both contributed by the collaboration (Dumusque et al. 2015; Phillips et al, in prep.). The HARPS-N Collaboration will continue to observe a sample of the brightest, least active, and least noisy northern stars. The goal is to discover Earth-like planets and multiple planet systems. The value of such discoveries is that a number of exciting follow-up possibilities become possible for such nearby objects. This was demonstrated with the discovery of the compact system of rocky planets orbiting HD 219134 (Motalebi et al. 2015). In 75 hours of observing time, HARPS-N determined the masses and orbital periods of four planets with reflex velocity amplitudes between 1 and 5 m/s. Photometric followup to the HARPS-N observations with the Spitzer space telescope showed two of the planets to be transiting, and to have bulk densities indicating rocky compositions. The TransitFind project of the CHEOPS mission will perform such followup after launch in 2018.

While bright stars do not require very long exposure times, in order to achieve long-term precision below 1 m/s we have to devote additional observing time to characterize the stellar p-mode oscillations, granulation noise power, and rotation induced variations. The latter in particular dictate an optimal temporal sampling pattern, necessitating time-sharing across the TNG/HARPS-N user community (See Section 4).

Judging from our current experience, this project will require at least 20 clear nights per year, in order to be scientifically viable. The primary targets for this project have already been identified, and HARPS-N observations have been in progress since 2012.

Observing time requirements and queue-scheduled observing

The K2 and TESS targets best-suited for radial-velocity follow-up with HARPS-N will be K-type main-sequence stars with rotational velocities low enough to ensure that their spectral lines are sharp enough to yield precise radial velocities. Our experience with Kepler and K2 targets with orbital periods of order 5-10 days indicates that radial-velocity amplitudes of order 1.5 to 2.0 m/s can be measured to 5 or 6 precision in 50 hours at V magnitudes between 10 and 12 (Dressing et al 2015; Gettel et al 2016).

Table 1: With HARPS-N, mass determinations can be achieved at the level of precision required for the primary project in a known number of hours of observation under the following assumptions. The K2 or TESS light curve constrains the period and phase of the planet's orbit. We select targets brighter than V = 12, for which HARPS-N achieves 1.0 m/s precision in 1 hour. Our estimates of the number of hours required to achieve 10% precision in planet mass in a 0.10 AU orbit are tabulated here.

The numbers of effective clear observing nights required to accomplish each of the three goals of our primary project are therefore:

A. 250h = 25 nights/year for 20 planets in various orbits;

B. 210h = 21 nights/year for 20 planets in various orbits.

C. 160h = 16 nights/year for 2 planets over 3 years;

Stellar activity is ubiquitous in stars of spectral type mid-F or later. This affects the radial-velocity signal in two ways. The more severe of the two is magnetic suppression of granular convective blueshift in faculae and network regions, but the rotational flux imbalances caused by dark starspots and bright faculae rotating across the visible stellar hemisphere also perturb the radial velocity (Meunier et al 2010; Aigrain et al 2012; Haywood et al 2016). For these reasons, even slowly-rotating stars exhibit quasi-periodic radial-velocity modulation with amplitudes of a few m/s.

Studies of the Sun with the HARPS-N solar telescope confirm that the activity-related RV variations arise from by modulations in the full width at half maximum (FWHM) and bisector inverse slope (BIS) of the cross-correlation function. These latter modulations are unrelated to the effects of planetary reflex motion, and allow the stellar rotation period to be determined independently from the spectra themselves. Provided the stellar rotation cycle is well-sampled, these variations can be modelled simultaneously with the planetary orbits, down to the noise limits imposed by photon noise, stellar granulation and the long-term spectrograph calibration -- 40 cm/s in the solar case. Similar long-term stability is found for the brightest target stars, many of which are less active than the Sun. We already have planetary-system detections with residuals below the m/s level (including stellar and instrumental noise as well as undetected planets) over time scales of years, during a minimum of activity of the star (e.g. HD40307, Mayor et al. 2009). Longer-term activity- cycle dependent changes in the stellar RV are more difficult to account for, but these take place on timescales of years. A key goal of the long-term solar HARPS-N observation programme is to develop new spectral proxies that will allow these longer-term effects to be calibrated out.

The form of the activity modulation changes as new active regions emerge and decay at different stellar longitudes, while planetary signals remain constant in both amplitude and phase. If the stellar rotation period is known from the photometric or Ca II H & K emission modulation period, the radial-velocity signals caused by planets and activity can be separated by treating the activity signal as a form of correlated noise. Gaussian-process regression is particularly eective (see e.g. Haywood et al. 2014), using a 4-parameter quasi-periodic covariance model which can be trained on the radial-velocity data themselves if the stellar rotation cycle is adequately sampled by the observations (Faria et al. 2016). Adequate rotational phase sampling requires 6-10 observations well-dispersed in rotation phase. The samples may be obtained over several rotation cycles, but they must achieve the necessary running phase coverage on a timescale shorter than that on which the shape of the light curve, i.e. the star's variability, evolves.

It may not be possible to determine the photometric rotation periods of TESS host stars within the 27-day campaign duration at low ecliptic latitudes. Nonetheless, our experience with Kepler and RPS targets is that the rotation periods of suitable host stars are typically in the range 10-40 days (McQuillan et al 2014), while active-region lifetimes range from 20 days for late-F stars to over 100 days for late K and M dwarfs (Giles et al. 2017). The ideal radial-velocity sampling pattern for a star of unknown rotation period is therefore at least one visit every 2-3 days, maintained over the 100 days or so of a full observing season. For bright targets in which granulation noise is also an issue, or in cases where the planet has an orbital period less than 2 days, the established HARPS strategy of taking two or three 15-min or longer observations per night separated by at least 2 hours, should be followed. This strategy stems from the work of Dumusque et al (2011a,b) which demonstrated that a strategy with 3 points in a night separated by more than 2 hours every 3 nights gives a better RV "binned value" than with a strategy with 1 point per night every night. This allows the granulation noise amplitude to be estimated independently. The orbits of short-period planets can still be sampled on timescales much shorter than the activity variations in this way (see e.g. Pepe et al. 2013).

Conventional block-observing allocations based on the lunar cycle are not compatible with this ideal, because they typically sample only a small fraction of the stellar rotation cycle each month. The window function of the telescope schedule compromises our ability to characterise the stellar noise and to separate it unambiguously from a putative planetary signal. Keeping the current schedule means that many more hours of observations are then needed to achieve a planetary mass determination. Instead, we have found that spreading the observations over longer sets of nights, in which we can better sample the stellar activity signal improves our ability to model the stellar noise. The recent study by Lopez-Morales et al (2016) of the Kepler-21 system demonstrated the effectiveness of implementing such a change in the RV sampling strategy in the 2015 observing season, enabling the stellar activity signal to be decoupled from the planetary reflex motions.

A queue-scheduled observing strategy, with different observing programmes being carried out concurrently and continuously, oers the most efficient route to precise planetary mass determination in the presence of stellar activity. Queue-scheduled observing carries the additional advantage that in conditions of poor seeing, transparency or wind pointing restrictions, there is a greater pool of backup targets whose requirements are more robust. More integrated coordination with a large Italian project would achieve the changes we are envisaging in the renewal of the agreement. Cooperative exchanges of time between the GTO and GAPS programmes have already proven extremely effective for achieving sampling patterns that improve the efficiency of both projects.

References 

Aigrain, S. et al. 2012, MNRAS, 419, 3147

Buchhave, L. A. et al. 2016, AJ,152, 150

Christiansen, J. et al. 2017, arXiv:1706.01892

Dressing, C. et al. 2015, ApJ, 800, 135

Dumusque, X. et al. 2015, ApJ, 814L, 21

Dumusque, X. et al. 2014, ApJ, 789, 154

Dumusque, X. et al. 2011a, A&A, 527, 82

Dumusque, X. et al. 2011b, A&A, 525, 140

Faria, J. P. et al. 2016, A & A, 588, 31

Gettel, S. et al. 2016, ApJ, 816, 95

Giles, H. et al. 2017, arXiv:1707.08583

Haywood, R. et al. 2016, MNRAS, 456, 3637

Haywood, R. et al. 2014, MNRAS, 443, 2517

Lopez, E. D. et al. 2014, ApJ, 792, 1

Lopez, E. D. et al. 2012, ApJ, 761, 59

Lopez-Morales, M. et al. 2016, AJ, 152, 204

Malavolta, L. et al. 2017, AJ, 153, 224

Mayor, M. et al. 2009, A&A, 493, 639

Meunier, N. et al. 2010, A&A, 512, 39

McQuillan, A. et al. 2014, ApJS, 211, 24

Motalebi, F. et al. 2015, A&A, 584, 72

Pepe, F. et al. 2013, Natur, 503, 377

Rogers, L.A. et al. 2015, ApJ, 801, 41

Vanderburg, A. et al. 2015, ApJ, 800, 59