The twin 6.5-m Magellan telescopes sit side by side on the 8,000 foot peak of Las Campanas in the southern Atacama desert.

Carnegie’s Las Campanas Observatory in Chile. Photo by Scott Sheppard, DTM.

Having successfully led the team that designed and built the twin 6.5-m Magellan telescopes, Stephen Shectman (Shec) then teamed with Rebecca Bernstein to build a general purpose echelle spectrometer for Magellan. The MIKE spectrometer was installed on Magellan in November 2002. It included an Iodine cell. The MIKE planet search was initiated on December 15 2002.

MIKE was primarily designed to observe faint objects to take advantage of the 6.5-m Magellan telescopes, and the spectacular quality of the Las Campanas sky. MIKE is not stabilized and was not designed to take spectra at extremely high resolution, both of which are desirable for a precision Doppler instrument.

Shectman and Butler began discussing a purpose-built echelle spectrometer explicitly designed to acquire the stellar Doppler velocity measurements with the highest possible precision in 1999. The primary attributes of a precision velocity echelle spectrometer are resolution, sampling, and stability, with cost always a looming factor. Previously we had used general purpose echelle spectrometers with a resolution of 45K to 60K and no stabilization.

Shectman produced the original design for the “Planet Finding Spectrometer” (PFS) in 2000. Light going through the spectromer makes a double pass through the optics before being captured by the CCD camera. This allows for a more compact design. Instead of being the size of living room, PFS has the footprint of a ping-pong table, and is encased in a thick insulating box.

This project was funded by the NSF Major Research Instrumentation program in 2002. The Carnegie Institution for Science and a private donor also provided significant support.

Jeff Crane and Shec carried out the refined design and construction of PFS, largely at the Carnegie Observatories optical shop in Pasadena. Ian Thompson oversaw the CCD dectector system. The Planet Finding Spectrometer commenced operation on January 1, 2010.

Jeff Crane showing the “innards” of PFS, the Planet Finding Spectrometer. The dewar containing the CCD detector is the gold cylinder to the left.
Jeff Crane showing the “innards” of PFS. The echelle grating is in the black cylinder at the far right.
The Iodine absorption cell is mounted in a coffee can.
Face-on view of the Iodine absorption cell. The cloth wrap around the sides of the Iodine cell is heat wrap, which is connected to a temperature controller that maintains the temperature of the cell at 60 C.
Jeff Crane replacing the front cover of PFS. The Planet Finding Spectrometer is encased in a thick insulating box. The inside of the box is temperature controlled at 27 C.
Shec directing sunlight into the Planet Finding Spectrometer in the Carnegie Optical Shop.

The PFS spectrometer commenced operation on 1 January 2010. The advantage of a purpose-built state-of-the-art design was immediately clear. While our previous programs at Lick, Keck, AAT, and MIKE had produced precision at the 3 m/s level, PFS generated 1 m/s precision. An example of a planet found with the MIKE program and followed up by PFS is shown here.

This planet was initially found in the MIKE planet search program. The improvement in precision with the PFS spectrometer, starting in 2010, is obvious.

The primary goal of the PFS team is to survey the ~500 nearest G to M dwarf stars (sun-like to red dwarf stars). In addition we have participated in several programs to follow up on fainter transit planet systems. In a transit planet system the planet actually crosses the face of the star as seen from the Earth. These planets can be detected both by Doppler velocities and photometry. An animation showing a transit planet is shown below:

Astronomical transit. By Silver Spoon – Own work, CC BY-SA 3.0.

As the planet crosses the face of the star, it blocks some of the light from the star. The fraction of the light blocked provides a direct measurement of the radius of the planet relative to the star. Doppler velocities provide a direct measurement of the mass of the planet. The combined photometry and Doppler velocity information gives the radius and mass of the planet, and hence the density. Gas giant planets have a density similar to water, while rocky planets like the earth are about 5 times more dense.

For the few hours that a planet is “transiting” the face of the host star, Doppler velocity measurements will “wiggle” as the planet first blocks light from the approaching and then the receding limb of the star. An example of this is shown from the transit system WASP-7 (Albrect et al. 2012).

The shape of the Doppler velocity “wiggle” while the planet is transiting the host star reveals the inclination of the orbit of the planet relative to the equator of the star. This is known as the Rossiter-McLaughlin (RM) effect. In the case of the Solar System the inclination of the planetary orbits are within a few degrees of the Sun’s equator, which is expected if planets form in a disk while the star is still coalesing from a cloud of gas and dust. Many of transiting planets have been found to have extreme inclinations, including some examples that are counter-orbiting their star. The giant planet orbiting WASP-7 is a particularly interesting example as it is rotating at an angle of ~90 degrees relative to the equator of the host star. At some point in the formation of this system, the planet was knocked sideways out the equatorial plane of the star.

The transit planet orbiting the M (red) dwarf star HATS-6 was photometrically discovered by the HAT-South team. This system is interesting as it is a rare example of a gas giant in a short period (3.32 day) orbit around an M dwarf. This star is particularly challenging because it is faint (V=15.2), about a factor of 100 fainter than we prefer to go with PFS. The HARPS, FEROS, and PFS spectrometers followed up on the HAT-South discovery. As the HAT-South team notes (Hartman et al. 2015), the planet is only detected in the PFS data.

Top panel: high-precision RV measurements for HATS-6 from Magellan/PFS (dark filled circles), MPG 2.2 m/FEROS (open triangles), and ESO 3.6 m/HARPS (stars), together with our best-fit circular orbit model. Zero phase corresponds to the time of mid-transit. The center-of-mass velocity has been subtracted. The orbital fit is primarily constrained by the PFS observations. Second panel: velocity $O-C$ residuals from the best-fit model. The error bars for each instrument include the jitter which is varied in the fit. Third panel: bisector spans (BS), with the mean value subtracted. Note the different vertical scales of the panels.

In 2012 Guillem Anglada began investigating an enhanced technique to analyze the HARPS spectra, resulting in the

HARPS-Terra velocity code (Anglada & Butler 2012). Several papers have resulted in the combination of PFS and publicly available HARPS spectra from the ESO archive. The first was a the discovery of a 5 earth-mass planet in the habitable zone of the M dwarf GJ667 (Anglada et al. 2012). A total of four planets ranging between 5 and 80 earth-masses were found to orbit this star.

Phase-folded RV measurements of the four signals with periods of 7.2 days, 28 days, 75days, and ~10 years . The planet in a 28 day orbit is ~5 earth-masses, and is in the habitable zone of the star. The 143 HARPS measurements are shown in red circles, 21 PFS measurements are shown in blue squares, and the green triangles correspond to the 20 HIRES observations. Each preferred Keplerian model is shown as a black line.

Another system found by the HARPS-Terra and PFS collaboration is the double planet orbiting GJ221 (Arriagada et al. 2013). This system consists of a 6.5 earth-mass planet in a 3.87 day orbit, and a 53 earth-mass planet in a 125 day orbit.

Keplerian solution for GJ 221. Left panel: phased Keplerian fit for the 125 day component. Right panel: phased Keplerian fit for the 3.87 day component. The red symbols correspond to HARPS velocities, while the blue circles correspond to PFS velocities. These planets are roughly 53 and 6.5 earth-masses respectively.

The Anglo-Australian Planet Search program discovered a gas giant orbiting the faint red M dwarf GJ832 in 2009 (Bailey et al.). Follow up observations from the HARPS-Terra program and PFS have revealed an additional 5.4 earth-mass planet in a 36 day orbit, near the habitable zone for this star (Wittenmyer et al. 2014).

Radial velocities and fit for GJ 832b; the signal of the second planet has been removed. AAT (green), HARPS (red), PFS (blue).
Left: radial velocities and fit for GJ 832c; the signal of the outer planet has been removed. AAT (green), HARPS (red), PFS (blue). Right: same as top panel, but the AAT data have been omitted from the plot to more clearly show the low-amplitude signal.

Kapteyn’s star is one of the very nearest M dwarfs. This star is so important that it has a name! The combination of PFS, HARPS-Terra, and Keck HIRES data have revealed two super-earths with orbital periods of 48 and 120 days (Anglada et al. 2014). The planets are 5 and 7 earth-masses respectively. The inner planet is in or near the potentially habitable zone for this star. The precision advantage of the modern purpose-built spectrometers HARPS and PFS are illustrated in the discovery figure.

Phase-folded Doppler curves of the planets signals with the other signal removed [HARPS are red circles, High Resolution Echelle Spectrometer (HIRES) are brown diamonds and Planet Finder Spectrograph (PFS) are blue squares]. The maximum-likelihood solution is depicted as a black line.

In early 2017 PFS will be taken off-line for a few months. The old 15 micron CCD detector will be replaced with a modern CCD with 9 micron pixels, dramatically improving our sampling. A fiber slicer will be installed, increasing our resolution from 80,000 to 130,000. The fiber slicer will also increase the stability of the seeing disc and the spectrometer point-spread-function. Our goal is to improve precision from the 1 m/s level to the 0.5 m/s (50 cm/s) level.