KECK OBSERVATORY

The Keck Observatory Planet Search Program discovered most of the first 200 exoplanets, including the first neptune-mass planet and the first super-earth.

In this unusual view looking east from the Subaru Telescope catwalk 12 meters above the ground, the nearby Keck I telescope and dome appear deceptively larger than the Keck II twin farther back. Yet they are identical, each with a 10-meter mirror and 37-meter dome. © 2007 Laurie Hatch.com / image and text.

As with the Lick Planet Search, the heart of the Keck Planet Search is a spectrometer designed and built by Steve Vogt at the Lick Observatory optical shop. The design for the HIRES spectrometer dates to the late 1980s, long before any extrasolar planets had been discovered. Vogt was aware of the work that Marcy and Butler had been doing on the Hamilton spectrometer at Lick Observatory and commissioned them to build a custom Iodine cell for HIRES.

The Keck HIRES spectrometer. This is similar to the Hamilton spectrometer except it uses a grating to cross-disperse the light rather than prisms.

NASA had purchased a one-sixth share of the two Keck 10-m telescopes. With the discovery of extrasolar planets in 1995 and 1996, and NASAs increasing involvement in the search and characterization of extrasolar planets, the NASA time was heavily weighted toward planet search program. Marcy, Vogt, and Butler started the Keck Planet Search Program on July 10, 1996. The time for the first two night observing run was provided by Ben Zuckerman of UCLA. Shortly thereafter the team began regularly receiving Keck time from NASA and the University of California.

A new planet search program typically requires 2 to 3 years before it begins generating planets. A number of stars need to be surveyed, and about 20 observations covering a full orbit need to be taken.

The first planet to emerge from the Keck Planet Search was the hot jupiter orbiting HD187123, a sun-like star. Kevin Apps a physics and astronomy student at the University of Sussex in the UK advised that this and several other stars be observed. With only 10 planets known at the time, it was already obvious that stars enriched with Iron and other elements heavier than Helium were producing a higher yield of planets than less enriched stars. Kevin studied the literature and found a number of bright enriched stars that were added to the Keck program. Keven shared authorship credit with the Keck team. Within the Keck team, this discovery was known as the “Planet of the Apps”.

HD 187123 was the fourth hot jupiter discovered. Discussion in the community was rampant about the possibility of a transit. Assuming random orbital inclination, the chance of a hot jupiter transiting its star is about 10%.

Keck Doppler velocities for HD187123 from the 1998 July observing run. A 3 day periodicity is evident from the 5 night observing string (Butler et al 1998, PASP 110, 1389).

H209458 was the sixth hot jupiter discovered, and the first to transit its host star. The Keck Doppler velocity data (Henry et al 2000) revealed a planet in 3.5d orbit with a minimum mass of 0.62 jupiter-masses.

Greg Henry at Tennessee State University photometriclly followed up the transit prediction from the Keck Doppler velocity data.

Discovery data for the transit planet HD209458 from the Keck planet search (Henry et al. 2000, ApJ 529, L41).

A team headed by David Charbonneau and Tim Brown independently discovered this planet with observations made at the Lowell and Geneva Observatories. The discovery papers by the two groups were published in the same issue of the Astrophysical Journal.

Over the next several years the Keck planet search program repeatedly set the record for the exoplanet with the smallest mass. In 2000 the group published the first two sub-saturn mass planets. HD 16141 has an orbital period of 75.8 days, and a minimum mass of 0.22 Mjup. HD 46375 has a period of 3 days, and a minimum mass of 0.25 Mjup. For comparison, 1 saturn-mass equals 0.30 Mjup.

Keck Doppler velocities, with error bars, for HD 16141. The solid line through the points shows the best Keplerian fit. The residuals having rms = 3.24 m/s are shown at the bottom (crosses) with arbitrary zero point (horizontal dashed line).
Keck Doppler velocities for HD 46375 from the 2000 February observing run. A 3 day periodicity is evident for this six-night observing string. Measurement uncertainties are ∼2.2 m/s.

The Keck planet search program announced the first neptune-mass planet in 2004. GJ 436 b is a neptune-mass planet orbiting a nearby M (red) dwarf star in a 2.64d period. 1 neptune-mass mass equals 17 earth-masses, or 0.054 jupter masses (Mjup)

Measured velocities vs. orbital phase for GJ 436 (filled dots), with repeated points (outside phases 0–1) shown as open circles. The dotted line is the radial velocity curve from the best-fit orbital solution, P = 2.644 days, e = 0.12, M sin i = 0.067MJup. The rms of the residuals to this fit is 5.26 m/s. The error bars show the quadrature sum of the internal errors (median = 5.2 m/s) and jitter (3.3 m/s). A linear velocity trend is found to be 2.7 m/s per year (Butler et al 2004, ApJ 617, 580).


There are three classes of planets in the Solar System. The terrestrial planets are rocky. The Earth is the most massive example of this class. The ice giants, Neptune and Uranus, are about 15 times more massive than the Earth. Finally the gas giants, Jupiter and Saturn, have masses about 300 and 100 times larger than earth, respectively. In 2005 the Keck planet search announced the first of a new class of planets, super-earths, with masses between 2 and 10 times that of the Earth. The Lick and Keck planet search programs had already discovered two gas giants with orbital periods of 30 and 60 days orbiting the nearby M (red) dwarf star GJ 876. Residuals to the 2-planet fit motivated continuing observations. By 2005 the evidence for a planet in a 1.9d orbit with a minimum mass of 5.9 earth-masses (and a most likely mass of 7.5 earth-masses) had become overwhelming.

The massive outer two planets in this system are in a 2-to-1 orbital resonance. Their gravitational interaction causes their orbital elements to evolve on short timescales, as shown in the middle and bottom panels of the figure.

Triple-Newtonian orbital fit to the radial velocity observations for GJ 876. The observed and model velocities for each planet are shown separately by subtracting the effects of the other two planets. The panels show the velocities due to companions d (top), c (middle), and b (bottom). The lines show model velocities for the first orbital period beginning with the epoch of the first observation. Note the differences in scale in the three panels. The deviations shown for companions c and b clearly demonstrate that their orbital elements have been evolving over the time span of the observations. The colored numbers in the bottom panel indicate which points correspond to which observing season. Note that the points taken in 1997 most closely follow the lines in the bottom two panels, as expected.
Keck HIRES CCD detector. This consists of 3 CCDs to capture the entire near UV, visible, and near IR spectrum from HIRES. It was installed in August 2004.

The original Keck HIRES CCD, installed at first light in 1993 had relatively large pixels (24 microns), a convex surface, excessive charge diffusion (which broadens the point-spread-function), and a subtle nonlinearity in the charge transfer efficiency (CTE). In August 2004 Steve Vogt replaced the original CCD with a mosaic of three CCDs. These CCDs have smaller (15 micron) pixels, a very flat focal plane (improving the the point-spread-function), and more spectral coverage. These CCDs are also free of signal-dependent nonlinearities, and have much smaller charge diffusion.

The Keck planet search program has maintained precision at the level of 3 m/s or better over 20 years. Examples of published stable stars are included below.

Measured velocities vs. time for four representative Keck/HIRES G & K dwarf stars that show no significant velocity variation during 6 years of observations. The standard deviation of velocities ranges from 2.5 to 4.0 m/s, which includes both the errors and photospheric jitter.
Measured velocities vs. time for four representative G & K dwarf stars that show no significant velocity variation during 6 yr of observations. The standard deviation of velocities ranges from 2.9 to 3.7 m/s, which includes both the errors and photospheric jitter.
Radial velocities vs. time for stable M dwarfs from the Keck planet search program. These M dwarfs are representative of the middle-aged M1.5–M3 dwarfs on the program. These stars have 10+ observations over 4+ yr. The observed rms velocity scatter of these stars ranges from 2.32 to 4.65 m s-1, showing that the combined velocity errors and photospheric jitter is less than 5 m/s (Butler et al 2004, ApJ 617, 580).
Radial velocity vs. time for Keck M dwarfs spanning 7 yr. The Keck HIRES system achieves precision of 3–5 m/s for M dwarfs brighter than V = 11.
Radial velocity vs. time for Keck M dwarfs spanning 7 yr. The Keck HIRES system achieves precision of 3–5 m/s for M dwarfs brighter than V = 11.