Entrevista com Paul Butler

Paul Butler é uma figura incontornável na área dos exoplanetas. O seu trabalho ao longo dos últimos 20 anos contribuiu de forma fundamental para o actual estado da arte em espectrógrafos de alta precisão e em software para tratamento dos dados provenientes destes sistemas. Esta tecnologia constitui a base do método da velocidade radial para a descoberta de exoplanetas. Durante os anos 90, Paul Butler e Geoffrey Marcy formaram uma dupla extremamente produtiva, com dezenas de descobertas de exoplanetas. Nos últimos 15 anos Paul Butler tem-se dedicado a implementar programas de observação de exoplanetas com espectrógrafos de alta precisão em vários instrumentos como o Anglo-Australian Telescope e os telescópios Keck, Magellan e VLT.

Recentemente o AstroPT contactou Paul Butler no sentido de realizar uma pequena entrevista a qual foi amavelmente concedida. Aqui fica o registo:

AstroPT – What is the Anglo-Australian Planet Search (AAPS), at the Anglo-Australian Observatory, and how does it compare with related projects like HARPS at La Silla, in terms of observing strategy and precision ?

Paul Butler – In 1997 the Anglo-Australian Telescope (AAT) was the largest telescope in the southern hemisphere, and was equally owned by the Australian and UK governments. Individuals were allowed to apply either to the UK (PPARC) or Australia (ATAC). Anglo-Australian Observatory staff astronomers had the special privilege of submitting proposals to the TACs of both nations. In August 1997 I took a position as an Anglo-Australian staff astronomer so that I could apply to both Australia and the UK for time to start the Anglo-Australian Planet Search. I hand carried the Iodine cell with me when I moved to Sydney.

We took our first data in January 1998 and were initially allocated 20 nights per year. We have maintained precision of 3 m/s over the past 12 years, a record matched only by our Keck program. Based on our continued improvement in velocity precision and track record for discovering planets, our time allocation was significantly increased in January 2005. In addition to our guaranteed time (currently 58 nights per year) we have received additional allocations of 48 consecutive nights in January 2007 and 47 consecutive nights in July 2009.

The HARPS program is producing excellent results, and doing a wonderful job of pushing forward the discovery of planets around nearby stars. While the Anglo-Australian and HARPS programs are obviously competitors, they are also symbiotically tied together.
There are the only two southern hemisphere programs that are producing Neptune and terrestrial mass planets, and the only two programs that can follow up the results of the other group. The ability to follow up is a crucial element of science, its importance can not be overstated. I am a huge fan of the HARPS program and the Geneva team. I am honored to be working competitively with the Geneva team in the quest to find planetary systems around the nearest stars.

AstroPT – In recent years the Anglo-Australian Telescope Rocky Planet Search (AATRPS) has been launched as a part of the larger AAPS. The program has already produced some amazing results such as the Neptune mass planet around Lambda-2 For, the recent Super-Earths and Neptunes around 61 Vir and HD1461 and an extra planet orbiting 23 Lib. Can you describe the rationale behind the AATRPS, its observing strategy and ultimate goal ?

Paul Butler – Rocky planets have have the lowest mass and smallest signals of any planets found to date. The velocity semi-amplitudes (~2 m/s) are only slightly larger than measurement precision. Extended observing runs offer two major advantages:
1) improved internal precision, and 2) observing multiple orbits. All of the lowest mass planets reported to date have emerged from extended observing runs.

The goal of the Anglo-Australian program is to search the 250 nearest and brightest sun-like stars for their hidden planetary systems. These are the stars that will be followed up for the next century with next generation advances in astrometry, interferometry, and direct imaging. The planetary systems discovered to date have been a big surprise, including planets in 4 day orbits and more distant planets that are typically in eccentric orbits. Neither of these classes of planets were predicted prior to their discovery.

While the discovery of rocky planets in small orbits makes most of the headlines these days, the discovery of jovian-mass planets with long period (>10 year) orbits is possibly the most interesting work being done today. Of the Solar System planets, only Jupiter and Saturn could be detected with current state-of-the-art work. Jupiter and Saturn analogs with orbital periods of 10 to 30 years are the only detectable signposts of Solar System analogs. These discoveries are emerging now from our long term programs at Keck and AAT. Recent discoveries that are particularly notable include the first Jupiter analog around an M dwarf (GJ 832) from the AAT, and the Jupiter analog orbiting 23 Lib from the AAT and Keck.

From our library of 1000+ stars from Keck, AAT, Lick, and Magellan, we are selecting a handful for extensive follow up to search for rocky planets, and continue the long term work needed to find Solar System analogs. We would like to know if terrestrial mass planets and Solar System analogs are common or rare.

AstroPT – The detection of the Super-Earths around 61 Vir and HD1461 is quite remarkable and requires state-of-the-art precision in the measurements. In fact, the discovery was done in cooperation with the Keck Planet Search program. Can you explain the role of each of these programs in this discovery ?

Paul Butler – We began observing 61 Vir with Keck in December 2004. As this is one of the very nearest and brightest sun-like stars in the sky, we added it to the AAT program in April 2005.

By early 2009 we had amassed nearly 200 observations of 61 Vir on the Keck and the AAT. The signals of the 3 planets were apparent in the data sets of both telescopes. On 1 July 2009 we began 47 consecutive nights on the AAT. The attached figure shows the results of this run.

The inner two planets, with orbital periods of 4.2 and 38 days are clearly seen. The RMS (median deviation) of the observations to the best-fit double Keplerian in this figure is 1.08 m/s, consistent with the internal measurement uncertainty of these observations.

The semi-amplitudes of the 3 planets in this system are 2.1, 3.6, and 3.2 m/s, among the smallest signals yet detected. We are always worried about making a false claim. Seeing the same signals at the same phase from two independent telescopes provides the confirmation needed to elevate these signals from interesting to reliably confirmed.

AstroPT – Can you describe in a nutshell the complexity of making radial velocity measurements of this magnitude (e.g. for the Super-Earth around 61 Vir) ? How difficult is it to acquire the raw data ? What kind of systematic errors, noise sources, must be taken into account ? Can you estimate the amount of work required to put a single data point into a radial velocity graph at this level of precision ?

Paul Butler – Generating a velocity measurement with a precision of 1 m/s has required 30+ years of effort. When we began working on precision velocity measurements, the state-of-the-art precision was ~300 m/s. This work is fundamentally based on the advances in designing and building high resolution échelle spectrometers. These instruments provide high resolution spectra across the entire visible spectrum. Prior to the pioneering work of people like Steve Vogt, high resolution spectroscopy could only provide a few angstroms of coverage. Other crucial advances that have made this work possible are CCDs and inexpensive high speed computing.

A single pixel on a typical high resolution spectrometer corresponds to about 2 km/s. Achieving 1 m/s precision requires measuring the shifts of stellar absorption lines at the level of 5-ten thousandths of a pixel, or about 30 silicon atoms on the CCD substrate. Small variations in the spectrometer point-spread-function (PSF), due to telescope guiding, telescope focus, spectrometer focus, and temperature variations can easily cause systematic velocity drifts of 50 m/s or more. Small non-linearities in the CCD can cause systematic velocity drifts of a few m/s.

I spent 8 years, from 1987 to 1995, designing hardware (an Iodine cell) and software (the point-spread-function recovery code) to reduce these systematic errors to the 3 m/s level. Computers in this era were slow, and there was no road map on how to do this. I spent many years wandering down blind alleys. The consensus of the few experts who knew of this effort was that it was hopeless and destined to fail. For many years I shared this fear.

In November 1994 Steve Vogt upgraded the resolution of the Hamilton spectrometer on the Lick 3-m telescope. With this improvement, I finally achieved 3 m/s precision in 1995, just prior to the discovery of 51 Peg by the Geneva group. This wonderful discovery led several research groups to lend us computing resources, which allowed us to reduce our 8 year backlog of data. By the summer of 1996 we had found 5 of the first 6 known planets.

Over the past 15 years I have worked to install Iodine precision velocity systems on several telescopes including Keck, AAT, VLT2, and Magellan. I have continued to work on improving measurement precision, and I am now achieving 1 m/s precision at Keck and AAT. Other groups in Texas, Harvard, Japan, Italy, and Germany now also run Iodine systems and data reduction packages based on this work.

AstroPT – The California-Carnegie Planet Search, of which you are a member, is a few months away from starting data acquisition from a unique facility — the Automated Planet Finder (APF). Is there any plan for a APF-South ? This would seem like the natural evolution for the AATRPS.

Paul Butler – The California-Carnegie program has been disbanded. I am working with Steve Vogt and Greg Laughlin on the Keck and APF programs, with Steve Shectman, Jeff Crane, and Dante Minniti on the Magellan program, with Chris Tinney, Hugh Jones, Jeremy Bailey, and Brad Carter on the Anglo-Australian program, and Greg Henry on the Las Campanas automated photometry program. These programs fall under the umbrella of the “Earthbound Planet Search”.

The APF is a robotic 2.4-m telescope with a state-of-the-art échelle spectrometer. Steve Vogt is the project scientist for the telescope and the spectrometer. Steve designed the spectrometer which has been built under his supervision at the Lick optical shop. We expect this facility will begin taking data in the next few months. Our group will get 45% of the time, about 164 nights per year, on this instrument. We will focus on observing the most interesting stars culled from our other programs to search for terrestrial mass planets.

Steve Shectman and Jeff Crane have worked for many years to design and build the new Planet Hunting Spectrometer for the 6.5-m Magellan telescope. Our first science run begins on 1 Jan 2010. We are initially gearing up for 20 nights per semester, with a goal of adding collaborators to reach 30+ nights per semester. In addition to broadly surveying hundreds of stars, we will choose a handful of stars from Keck and AAT for intensive high cadence observing with the goal of finding terrestrial mass planets.

Greg Henry and I are working to install two 0.8-m robotic photometry telescopes at Las Campanas with the goal of achieving 1 milli-magnitude photometric precision on our planet search stars. This will allow us to search for photometric periodicities in our program stars, find stellar rotation periods, and search for connections between stellar photometric variability and Doppler velocity variability.

AstroPT – What technological advancements do you think will be involved in planet detection using the radial velocity method in the near future ? At what level of precision will we cease to see any improvements ?

Paul Butler – Improving Doppler precision has been an incremental process. Part of this work is focused on better spectrometers which include higher resolution, higher throughput, and increased stability. The remaining improvements are in software, the ability to model the spectrometer, most importantly the variable point-spread-function. The later is the “smearing” function of the spectrometer.

Improving precision from 3 m/s to 1 m/s requires the signal-to-noise ratio (S/N) be improved by a factor of 3. Since S/N goes as the square root of the number of photons, this requires 9 times as many photons. Further improvements become daunting, both in terms of eliminating systematic errors and improving the S/N.

The Geneva team and our team are both committed to achieving better measurement precision. It is not clear what precision can ultimately be achieved. The stars themselves ultimately set the precision floor. Unfortunately stars are not billiard balls. They are heaving, gurgling gas balls that oscillate on time scales of minutes with amplitudes as much as 2 m/s, convect on time scales of hours, and differentially rotate with asymmetric spot patterns on time scales of weeks and months.

The earth causes the Sun to wobble with a semiamplitude of 0.1 m/s. It will be an enormous struggle to achieve long term precision at this level.

AstroPT – Finally, in a more personal tone, what fascinates you the most about your work and how do you perceive its impact for future generations ?

Paul Butler – When I began this work, there were no known planets other than our own Solar System. Everyone was certain that all planetary systems would look like our own, with small planets close in, giant planets further out, all in concentric circular orbits. Nature is apparently much more creative and imaginative than humans. I continue to look forward to the new surprises that nature has in store for us.

I greatly enjoy the technical challenges of putting ground based hardware systems together that can detect unseen planets orbiting distant stars. As Steve Vogt puts it, we are putting together glass, silicon, and steel, literally sand and rocks, to detect the smallest variations around stars that are light-years away.

When we started this business, there was no sub-field of extrasolar planets. Merely stating that you were working on planet detection in 1988 would trigger laughter among other astronomers. Today there are hundreds of people working on extrasolar planets, funding panels set up explicitly for extrasolar planets, graduate students who became interested in astronomy because of extrasolar planets, major space based initiatives within NASA and ESA dedicated to extrasolar planets. It has been a great joy to watch this blossoming from such humble origins.