Probing the planet formation environment

by Ian Czekala

Ian Czekala. (Courtesy I. Czekala.)

As an astronomer, I think I live in a spectacularly exciting time to be studying the process of planet formation. Little needs to be said about how dramatically the Kepler  satellite telescope and other exoplanet surveys have revolutionized our understanding of the exoplanet population, as these kinds of discoveries pop up in the news on a seemingly daily basis. We now know that the planet formation process produces a diverse set of final products, many of which (such as hot Jupiters and super Earths) look very different from the planets in our own solar system.

Detection and formation meet in the middle

What may be less appreciated, however, is that we are also making great strides in understanding exoplanet formation by studying their birthplaces in ever greater detail. High-resolution observations of protoplanetary disks, from the optical to sub-millimeter wavelengths, probe exoplanetary systems during their formation epochs and provide new evidence to help us understand the processes at work. What is exciting is that these two approaches of understanding planet formation—extrapolating backward in time from the exoplanet population and conjecturing forward in time from the initial conditions—are beginning to meet in the middle.

Figure 1. Artist’s impression of a circumbinary disk in the process of forming planets around a double star system. (Credit: NASA/JPL-Caltech/T. Pyle (SSC).)
Figure 1. Artist’s impression of a circumbinary disk in the process of forming planets around a double star system. (Credit: NASA/JPL-Caltech/T. Pyle (SSC).)


Observing the Initial Conditions of Planet Formation

To study the initial conditions of planet formation, I use the Atacama Large Millimeter Array (ALMA), a 66-antenna sub-millimeter interferometer in Chile, to directly measure the emissions from gas-rich protoplanetary disks. We use novel forward-modeling techniques with molecular tracers like carbon monoxide and its isotopologues (molecules of carbon monoxide built from different isotopes of carbon or oxygen) to reconstruct the velocity field of the system. Using this new forward-modeling approach, we can obtain a precise measurement of the central stellar mass, a crucial parameter for understanding the disk dynamics as well as the age of the star.

I have focused my attention on circumbinary disks (see Figure 1), because the presence of an additional star in the system creates an active dynamical environment and permits a more detailed characterization of the system architecture. To understand these systems better requires a multi-wavelength approach, and so we turn to high resolution optical and infrared spectroscopy from a variety of echelle instruments (“echelle” refers to the instruments’ specialized diffraction gratings), such as the TRES spectrograph on Mt. Hopkins, to study the stellar photospheres. Because we have also obtained radial velocity measurements(i.e. figuring out how rapidly the star is moving towards and away from us, as it orbits the other) to precisely constrain the dynamics of the stellar orbit, we can check whether or not our disk-based measurements agree, and in the process search for interesting things like misalignment between the disk and the stellar orbital plane. The stringent dynamical constraints in such a multiple system also allow us to test pre-Main Sequence evolutionary models—which predict fundamental stellar properties like effective temperature, radius, and composition as a function of stellar mass and age—by comparing with our measurements.

In the past, I’ve worked on addressing some of the difficulties inherent in matching “synthetic spectra” of these stars -- i.e. spectra generated from these stellar models—to actual data (see below for an illustration of the process).

Fitting solar spectra to models. (Credit: I. Czekala.)
Figure 2. A demonstration of how fitting stellar spectra is often not straightforward (a stellar spectrum is the intensity of light from a star as a function of wavelength). The top spectrum is the data and the second row is one of the best-fit synthetic models. The bottom spectrum shows the residuals (data minus the model), which have a red-noise structure to them. Obtaining proper estimates of stellar properties like temperature, mass, and radius requires accounting for this covariant structure. (Credit: I. Czekala.)


With binary stars, however, there is an added complication. As the stars move in their orbits, we only ever see the “tangled” spectra. I recently developed a new technique that uses Gaussian processes to disentangle binary star spectra and simultaneously infer the orbital parameters, which opens new avenues for spectral analysis of multiple star systems, including precise stellar mass measurements. Improvements in our understanding of stellar properties help us refine our theories of how stars form and protoplanetary disks evolve, because these constrain the timescale (stellar age) and gravitational environment (stellar mass) for planet formation.

When taking spectroscopic observations of close binary stars, we usually cannot spatially resolve the stars on the sky. This results in "composite" spectra that contain light from both stars. Because of  the Doppler shift as these stars move through their orbit, however, we have an additional axis along which we can disentangle their light, wavelength (see text below, also). (Credit: I. Czekala.)

Our new technique simultaneously models the stellar orbits (using Keplerian orbital dynamics) and the intrinsic stellar spectra using Gaussian processes. The video above shows the process in action. As the video progresses, the third panel on the left shows several narrow snippets of data of an (M4 class) spectroscopic binary star. Notice that as time goes on, the composite spectrum changes drastically (if this were only one star, it would be the same throughout time).

On the right are the radial velocities of the primary (blue) and secondary (red) stars given by the orbit as a function of time. The "disentangled" spectra are shown in the first and second panels on the left. This combination of spectra and orbit were simultaneously determined to provide the best fit to the existing data. Now that the spectra are disentangled, we can analyze them to determine the properties of each star!

A sensitive view of protoplanetary disks with GPI

Here at KIPAC, I’ve also started working with the Gemini Planet Imager (GPI) team. GPI is an integral field spectrograph behind an adaptive optics system living on the 8m Gemini South telescope in Chile. This amazing instrument delivers a spectrum everywhere within the field of view, and is designed to discover and characterize giant exoplanets orbiting other stars. GPI is also well-equipped to observed polarimetric signatures of scattered starlight from circumstellar disks. Scattered light observations are complementary to sub-millimeter observations with ALMA because they probe micron-sized grains in the surface layers of the disk, while the ALMA observations probe larger millimeter-sized grains in the the bulk of the disk.

I am looking forward to combining these types of observations to learn more about the environment in which planets form. Eventually, combining all the techniques, we will be learning much more not only about the conditions and circumstances of the birthplaces of planets, but how such a diverse population that looks quite different from the denizens of our own Solar System can arise.  

The author with the GPI during an observing run in Chile. (Courtesy I. Czekala.)
The author with the GPI during an observing run to Chile in November 2016. (Courtesy I. Czekala.)


Constructing a Flexible Likelihood Function for Spectroscopic Inference

A Disk-based Dynamical Mass Estimate for the Young Binary AK Sco

A Disk-based Dynamical Constraint on the Mass of the Young Binary DQ Tau

Disentangling Time-series Spectra with Gaussian Processes: Applications to Radial Velocity Analysis