CALIFORNIA STATE UNIVERSITY, NORTHRIDGE
A Unique Data Reduction Methodology to Obtain Full Resolution Spectra of Exoplanet
Atmospheres Observed Using the Hubble Space Telescope in Stare Mode
A thesis submitted in partial fulfillment of the requirements For the degree of Master of Science in Physics
by
Amanda L. Rowen
August 2018
ã Copyright by Amanda L. Rowen 2018
ii
The thesis of Amanda L. Rowen is approved:
______Dr. Mark R. Swain Date
______Dr. Gael M. Roudier Date
______Dr. Say-Peng Lim Date
______Dr. Damian J. Christian, Chair Date
California State University, Northridge
iii Dedication
To my Parents for their patience and continuous belief in me
iv Acknowledgements
I would like to express my utmost gratitude to my group at JPL, especially Dr. Mark Swain and Dr. Gael Roudier for their immense support and guidance throughout my research and for constantly pushing me to be a better scientist. I would also like to thank Dr. Damian Christian and Dr. Say-Peng Lim for their patience and support - especially during hard times - during my years at CSUN. If it were not for the continued support and guidance from each one of them, this thesis would not be possible.
This research is based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the Data Archive at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555.
This research has made use of the NASA Exoplanet Archive, which is operated by the California Institute of Technology, under contract with the National Aeronautics and Space Administration under the Exoplanet Exploration Program.
Some/all of the data presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.
v Table of Contents
Copyright ...... ii Signature Page ...... iii Dedication ...... iv Acknowledgements ...... v List of Figures ...... vii Abstract ...... ix
1 Introduction ...... 1 1.1 Detecting Exoplanets ...... 1 1.1.1 Pulsar Timing ...... 2 1.1.2 Astrometry ...... 3 1.1.3 Radial Velocity ...... 3 1.1.4 Gravitational Microlensing ...... 3 1.1.5 Direct Imaging ...... 4 1.1.6 Transits ...... 4
2 Characterizing Exoplanet Atmospheres ...... 5 2.1 Transmission Spectroscopy ...... 5 2.2 Hubble Space Telescope: Stare vs. Scan ...... 9
3 Methods ...... 11 3.1 Data Reduction and Background Subtraction ...... 11 3.2 Building an Instrument Model ...... 14 3.3 Bayesian Statistics: The MCMC Method ...... 15
4 Results and Analysis ...... 18
5 Discussions ...... 22
6 Conclusions ...... 24
References ...... 25
vi List of Figures
1.1 Mass – Period relationship of all confirmed exoplanets, to date, color coded by their detection method. Image Credit: NExScI Exoplanet Archive, March 26, 2018, Akeson et al. 2013...... 2
2.1 A transmission spectrum from Wakeford et al. 2017 of WASP-39 showing enriched band amplitudes at varying wavelengths due to chemical abundances causing atmospheric opaqueness...... 6 2.2 An illustration of an exoplanet transit and it’s corresponding light curve. The solid blue line shows what the transit would look like due to limb darkening and the red dotted line shows what the transit would look like if there was no limb darkening...... 6
2.3 Comparing performances of three different limb darkening laws for the star GJ 436. From top to bottom: A Linear, Quadratic, and Non-Linear Fit to the Phoenix grid model. The three sets of semi-empirical models for each graph were attributed errors coming from the uncertainties in temperature, metallicity, and logg of GJ 436. Three models were chosen per graph at different wavelengths to account for the wavelength dependence of the limb darkening laws...... 8
2.4 GJ 436 limb darkening spectrum. The four coefficients (and their errors) of the non-linear limb darkening law...... 9
2.5 A spatially scanned image of GJ1214 compared to a staring mode image (red insert) of the star. The 0th and 1st orders are labeled in red. Image Credit: P. McCullough & J. MacKenty, May 02, 2012...... 10
3.1 Calibration and data reduction steps for post processing the STScI data...... 11
3.2 A Linearity Plot showing the signal of WASP-12 data versus time for Nsamp 0, 1, and 2...... 11
3.3 The standard deviation of the data with varying box size. The solid blue line shows the background subtracted data, while the dotted red line shows the data before background subtraction...... 12
3.4 Spectrum of WASP-12 during each main step of the data reduction and background subtraction steps. The steps in these examples (clockwise from top left) are the original image pre-reduction, the trimmed image after removing dark current and artificial negative pixels, the background selection region, the select- ed spectrum from the background...... 13
vii
3.5 The data timing calibration sequence for WASP-12b. (Top Left): A diagram showing the orbital phase angle of the planet with respect to its host star and the observer. (Top Right): The orbital phase sampling for each frame. (Bottom Left): The orbit sampling to determine the orbit number for each frame. (Bottom Right): The visit sampling showing the total number of orbital visits...... 14
3.6 The White Light Curves of WASP-12b (Top) and WASP-17b (Bottom). The original data are represented as bl- ack circles. The black dotted line shows the theoretical model. The red triangles are the corrected data points...... 15
4.1 Comparing my stare spectrum of WASP-12b to Kreidberg et al. 2015 spectrum shot in scan mode, and Swain et al. 2013 spectrum shot in stare mode. All spectra have been mean shifted so as to be directly compared. The top images aren’t binned while the bottom images have been binned down so as to match the form- at of the published spectra...... 18
4.2 The spectrum of WASP-12b over plotted with previously published spectra from Swain and Kreidberg showing a mean shift in average transit depth...... 19 4.3 The retrieval of WASP-12 for both sets of data are plotted in order to show the mean shift in posteriors and the associated errors. From top to bottom: Average transit depth, inclination, and mid-transit time. Collins and Southworth parameters are over plotted along with their relative errors to show how WFC3 localizes each parameter due to the sensitivity of the instrument when run through the EXCALIBUR pipeline...... 21
viii Abstract
A Unique Data Reduction Methodology to Obtain Full Resolution Spectra of Exoplanet
Atmospheres Observed Using the Hubble Space Telescope in Stare Mode
By Amanda L. Rowen Master of Science in Physics
Transit spectroscopy was the method by which the existence of exoplanet atmospheres was definitely established (Charboneau et al. 2002) and the molecular composition first probed (Tinetti et al. 2007, Swain et al. 2008). In the last decade, transit spectroscopy has grown rapidly and revolutionized our ability to probe atmospheric composition and conditions in a wide range of exoplanet atmospheres (Sing et al. 2016, Iyer et al. 2016, Kreidberg et al. 2014). However, interpretations of spectral transit light curves are heterogeneous in the field with multiple teams reducing their spectra in unique ways. A homogeneous extraction process is needed to produce spectra from multiple Hubble instruments (COS, STIS, WFC3) in order to combine them for more wavelength coverage, giving a more complete picture of exoplanet atmospheres. I report a tool to extract and produce exoplanet spectra from Hubble Space Telescope’s Wide Field Camera 3 (WFC3) instrument, shot in Stare mode. A transmission spectrum of WASP-12b is extracted in the near-infrared from 1.1–1.7 µm and compared to previously obtained spectra shot in stare and scan mode.
ix Chapter 1
Introduction With the implementation of increasingly precise technology and instruments, our knowledge of exoplanets has been broadening exponentially. Understanding multiple aspects such as the formation processes, host star properties, orbital parameters, and atmospheric composition are all critical in our full understanding of exoplanets. Over the past two decades, multiple missions have been tasked with discovering and characterizing exoplanets in their own unique way. On April 24, 1990, NASA’s Hubble Space Telescope launched on what would become one of the longest ongoing missions in history, discovering and characterizing multiple exoplanets through varying methods such as radial velocity, direct imaging, and the transit method. With the rising number of confirmed exoplanets each day, the task of characterizing them in terms of density, insulation, and atmospheric composition has become a necessary step in fully understanding exoplanets and our place in the universe. Chapter 1 will discuss the most common methods for detecting exoplanets as well as noteworthy discoveries made throughout history using each. Chapter 2 will explain the concepts of transmission spectroscopy as well as go on to compare the two observing modes of the Hubble Space Telescope in terms of their significance to this research. Chapter 3 will list the specific data reduction and calibration steps used to extract the spectrum of WASP-12b and how that spectrum can be used to probe atmospheric characteristics of the planet. Markov chain Monte Carlo sampling methods and Bayesian Statistics will be discussed pertaining to their importance in converging on the most likely solution for instrumental and science parameters. Chapter 4 will reveal the results of the spectral retrieval of WASP-12b as well as compare it to previously published spectra shot in scan and stare mode. Chapter 5 will discuss the implications of these results as well as their significance. Finally, Chapter 6 will detail future work that is made possible by this new calibration tool as well as conclude scientific findings involving the calibration of WASP-12b spectrum via stare mode.
1.1 Detecting Exoplanets An exoplanet is a planet that orbits a star outside of our solar system. According to NASA’s exoplanet archive, as of July 2018, there are 3,772 confirmed exoplanets and 2,720 candidates. There are a range of detection methods used to find and confirm exoplanets. Each
1 method has a preferred category of planet they discover due to the nature of the method. For instance, the transit method has a bias towards higher radius, short-period planets because the larger the planet, the easier it is to detect a dip it induces in the light curve as it passes in front of its star. With the same reasoning, it is also biased towards smaller stars. Also, since the transit method needs to measure multiple transit events in order to confirm the existence of an exoplanet, short-period candidates get confirmed quicker. The preference of this method is shown on figure 1.1 as a surplus of transit detections tend to be towards the upper left region of the figure.
Figure 1.1: Mass – Period relationship of all confirmed exoplanets, to date, color coded by their detection method. Image Credit: NExScI Exoplanet Archive, March 26, 2018, Akeson et al. 2013.
1.1.1 Pulsar Timing In 1992, the first two exoplanets were detected by Aleksander Wolszczan and Dale Frail orbiting PSR B1257+12, a pulsar in the constellation Virgo, using a method known as pulsar timing (Wolszczan & Frail 1992). A pulsar is a rapidly spinning neutron star that emits beams of extreme electromagnetic radiation that, when viewed from Earth, appear in periodic regular pulses. The pulses are timed so precisely that they are more accurate than an atomic clock (Guinot & Petit 1991). When an exoplanet is orbiting a pulsar, it causes the center of gravity to shift outside of the center of the pulsar which in turn causes the pulsar to wobble. An observer viewing this phenomenon would measure a periodic variation in the pulsar’s beam of radiation being emitted and be able to determine a presence of a planetary companion. Through this time variation method, exoplanets around pulsars can be detected.
2 1.1.2 Astrometry As discussed previously, an orbiting exoplanet shifts the center of mass of the planetary system and in turn causes its host star to wobble. In extreme cases, this wobble may be able to be detected via a method known as Astrometry, the method in which a star’s relative position in comparison to nearby field stars appears to shift. Through very precise measurements this wobble may be physically observed and measured in the detection of exoplanets. A research team led by Johannes Sahlman took astrometric measurements over the course of two years to reveal the orbital motion of a nearby ultracool brown dwarf DENIS-P-J082303.1-491201 caused by a companion (Sahlman et al. 2013). This companion was found to be an exoplanet of 28 ± 2 Jupiter masses in a highly eccentric orbit and became the first confirmed exoplanet discovered via astrometry.
1.1.3 Radial Velocity Another way to view a star’s wobble is through the method of radial velocity. If the inclination of a planetary system is nearly parallel to the line of sight of an observer, one may be able to measure a star’s wobble through the use of Doppler shift. Doppler shift is the measurement of how much a wave is stretched or compressed when reflected back to an observer due to the relative movement of the star. When the star is wobbling towards the observer, the wavelengths become compressed towards the bluer end of the spectrum and when a star is wobbling away from the observer the wavelengths become stretched towards the red end of the spectrum. This allows one to measure the effects that an exoplanet has on its host star and in turn determine properties of the planet such as size, period, and distance from the star. Pegasi 51, a solar-type star, was found to have a Jupiter-mass companion with an orbital period of 4 days (Mayor & Queloz 1995). This planet, presently known as Pegasi 51b, was discovered using regular periodic variations in the radial velocity measurements of its host star.
1.1.4 Gravitational Microlensing When an exoplanet system passes between an observer and a background star, the background star’s light is bent towards the observer due to the gravitational pull from the exoplanet system bending the fabric of space-time, causing light to take the shortest possible path through space. This phenomenon causes the brightness of the background star to increase or ‘magnify’ as the planetary system passes in front. As the planet passes in front of the star,
3 a spike in brightness may be measured in the light curve for a shorter period of time than the actual full lensing event allowing for an exoplanet detection. A sub-Neptune planet of 5.5 earth masses, named OGLE-2005-BLG-390Lb, was discovered orbiting 2.6 astronomical units from a 0.22 solar mass M-dwarf star through a microlensing event (Beaulieu et al. 2006). Recently, it has even been suggested that quasar microlensing can be used to probe extragalactic planets within a lens galaxy (Dai & Guerras 2018).
1.1.5 Direct Imaging An exoplanet can be seen directly when imaged very carefully with precise instruments. This process, known as direct imaging, is done through the use of coronagraphs (commonly called ‘masks’) which are used to physically block out the light coming from an exoplanetary system’s host star into the telescope. This allows the image to detect the fainter light coming from the exoplanet itself, rather than the host star’s light saturating the field of view in the image. The first exoplanet was directly imaged by a team of researchers led by Paul Kalas of the University of California, Berkeley in which they imaged the planet Fomalhaut b, located 25 lightyears from Earth (Kalas et al. 2008).
1.1.6 Transits Encompassing the majority of exoplanet discoveries, the transit method is responsible for roughly 78% of all confirmed exoplanets (e.g., Charbonneau et al., 2000; Knutson et al. 2014; Sing et al. 2016). On April 1, 2008, the planet WASP-12b was discovered via the transit method by the ground-based SuperWASP observatory (Pollacco et al. 2006). The planet, with a radius of R = 1.79 ± 0.09 Jupiter Radius (RJ), and mass of M = 1.41 ± 0.1 Jupiter Masses
(MJ) (Hebb et al. 2009), has become an ideal candidate for exoplanet atmospheric studies (Swain et al. 2013). When a planet passes between an observer and its host star, the planet blocks some of the star’s light, therefore a decrease in brightness is measured. The time dependence of the light curve can tell many properties of the system such as planetary radius, inclination, semi-major axis, etc. By plotting its fractional planet to star radius versus wavelength (exoplanet spectrum) one can extract the properties of the planet’s terminator region atmosphere (Tinetti et al. 2010).
4 Chapter 2
Characterizing Exoplanet Atmospheres When a planet transits its host star during the primary eclipse (transit), a small fraction of the star’s light passes through the planetary atmosphere. Because the light from the star will be absorbed differently when passing through the exoplanet atmosphere, depending on its composition, the radius of the planet will change depending on the wavelength. The modulation in the planetary radius will follow the absorption lines of each element constituting the exoplanet atmosphere. Therefore, a transmission spectrum is obtained by plotting the fractional radius of the planet to that of the star with respect to varying wavelength. The spectral retrieval can give insight into chemical abundances and physical processes that comprise the planet’s atmosphere such as clouds, hazes, and (non) equilibrium chemistry allowing an in-depth study of exoplanetary atmospheres. NASA’s Hubble Space Telescope has the capability to probe exoplanetary atmospheres thanks to its on-board spectrograph (WFC3+grisms). In this chapter, the fundamentals of transmission spectroscopy are discussed focusing on the influence of stellar limb darkening. The two data-taking modes of Hubble Space Telescope are then discussed to demonstrate their relative performances in obtaining exoplanet transit spectra.
2.1 Transmission Spectroscopy Atmospheres are more or less opaque at different wavelengths, with the amount of opaqueness depending on the chemical composition of the atmosphere. Depending on the wavelength probed, we can learn many different things about an exoplanet’s atmosphere through transmission spectroscopy. At ultraviolet wavelengths, Lymann Alpha and ionized metals can be measured which can provide insight as to atmospheric mass loss. At visible wavelengths, sodium and potassium can be measured which can provide insight as to the physical composition of the atmosphere such as clouds, hazes, or alternatively a clear atmosphere (Knutson 2012). Probing at infrared wavelengths can determine the chemical composition of the atmosphere to determine if the relative abundances of H2O, CH4, CO, and
CO2 are in thermochemical equilibrium (Burrows et al. 2006).
5 Figure 2.1: A transmission spectrum from Wakeford et al. 2017 of WASP-39 showing enriched band amplitudes at varying wavelengths due to chemical abundances causing atmospheric opaqueness.
In order to properly understand a planet’s transmission spectrum, the limb darkening of its host star must be well known and represented. Limb darkening is an optical phenomenon which appears as the decrease in brightness across the visual disk of a star from its center to its edge or limb, which is responsible for light curves appearing rounded in shape rather than sharp linear dips as shown in figure 2.2. To accurately model the light curve, a limb darkening law must be chosen that can produce intensities over the entire stellar disk, while preserving the flux. Figure 2.3 compares the performances of three different limb darkening laws for the
Figure 2.2: An illustration of an exoplanet transit and it’s corresponding light curve. The solid blue line shows what the transit would look like due to limb darkening and the red dotted line shows what the transit would look like if there was no limb darkening.
6 star GJ 436. The best fit results show that unlike linear and quadratic, a non-linear limb darkening law is best suited to match the Phoenix grid model over the entire disk, therefore, a four-parameter non-linear limb darkening law is chosen for this study. GJ 436 is an M dwarf star that stress tests the limb darkening laws to their limits (Maness et al. 2007). If the limb darkening laws are well suited for this type of star it will work for all of the stellar candidates within the pipeline.
Assuming spherical symmetry, a non-linear limb-darkened source can be represented by equation 2.1.