Upgrade to Pro — share decks privately, control downloads, hide ads and more …

My work in Astronomy

Stefano Meschiari
December 31, 2015
68

My work in Astronomy

Stefano Meschiari

December 31, 2015
Tweet

Transcript

  1. Hi, I’m
    STEFANO MESCHIARI
    and I study
    EXOPLANETS
    Civitas Learning — 11/11/2015
    UT AUSTIN/MCDONALD OBSERVATORY

    View full-size slide

  2. My work: overview
    Systemic
    • A package to analyze exoplanetary data
    • A virtual lab for teachers & students
    SAVE/Point
    • A pilot platform for education
    • 4 games and apps ready to play
    • Discovering new exoplanets
    • Understand how they form
    • Creating apps for education & outreach
    0.0 0.5 1.0 1.5 2.0 2.5 3.0
    -5 0 5 10
    Time [JD]
    RV, Planet b [m/s]
    Q01 KECK APF
    0 1 2 3 4 5 6 7
    -10 -5 0 5
    Time [JD]
    RV, Planet c [m/s]
    Q01 KECK APF
    -10 -5 0 5 10
    RV, Planet f [m/s]
    Q01 KECK APF
    0 20 40 60 80
    -10 -5 0 5 10
    Time [JD]
    RV, Planet g [m/s]
    Q01 KECK APF
    0 500 1000 1500 2000
    -10 -5 0 5 10
    Time [JD]
    RV, Planet h [m/s]
    Q01 KECK APF
    -5 0 5
    Residuals [m/s]

    View full-size slide

  3. What are exoplanets?
    SINCE 1995, WE NOW KNOW APPROXIMATELY 5,000 PLANET CANDIDATES!
    51 Peg b Gliese 581d Kepler 16 ABb Earth twin Jupiter analog
    with exomoon

    View full-size slide

  4. What are exoplanets?
    SINCE 1995, WE NOW KNOW APPROXIMATELY 5,000 PLANET CANDIDATES!
    51 Peg b Gliese 581d Kepler 16 ABb Earth twin Jupiter analog
    with exomoon
    Exoplanetary systems don’t look like the Solar
    System. Why?
    PROBLEM:

    View full-size slide

  5. 10-1
    100
    101
    102
    103
    100 101 102 103 104 105
    Period [days]
    Mass [MEarth
    ]
    Discovery method
    RV
    Transit
    Imaging
    Timing
    Microlensing
    Are we unique ❄ snowflakes❄?
    COPERNICAN MEDIOCRITY PRINCIPLE: PROBABLY NOT. MASSIVE SELECTION EFFECT
    Earth
    Jupiter
    Saturn
    Uranus
    Harder to
    detect

    View full-size slide

  6. Are we unique ❄ snowflakes❄?
    COPERNICAN MEDIOCRITY PRINCIPLE: PROBABLY NOT, BUT…
    0.125 AU
    HD 219134
    Mercury
    Earth
    Vogt, Burt, Meschiari et al., 2015

    View full-size slide

  7. Answering the why
    (1) FIND MORE EXOPLANETS!
    Automated Planet
    Finder (APF)
    Keck
    Discovering new exoplanets through data
    0.0 0.5 1.0 1.5 2.0 2.5 3.0
    -5 0 5 10
    Time [JD]
    RV, Planet b [m/s]
    Q01 KECK APF
    0 1 2 3 4 5 6 7
    -10 -5 0 5
    Time [JD]
    RV, Planet c [m/s]
    Q01 KECK APF
    0 5 10 15 20
    -10 -5 0 5 10
    Time [JD]
    RV, Planet f [m/s]
    Q01 KECK APF
    0 5 10
    net d [m/s]
    Q01 KECK APF
    0 20 40 60 80
    -10 -5 0 5 10
    Time [JD]
    RV, Planet g [m/s]
    Q01 KECK APF
    0 500 1000 1500 2000
    -10 -5 0 5 10
    Time [JD]
    RV, Planet h [m/s]
    Q01 KECK APF
    2451000 2453000 2455000 2457000
    -5 0 5
    Time [JD]
    Residuals [m/s]
    0.0 0.5 1.0 1.5 2.0 2.5 3.0
    -5 0 5 10
    Time [JD]
    RV, Planet b [m/s]
    Q01 KECK APF
    0 1 2 3 4 5 6 7
    -10 -5 0 5
    Time [JD]
    RV, Planet c [m/s]
    Q01 KECK APF
    0 5 10 15 20
    -10 -5 0 5 10
    Time [JD]
    RV, Planet f [m/s]
    Q01 KECK APF
    0 10 20 30 40
    -10 -5 0 5 10
    Time [JD]
    RV, Planet d [m/s]
    Q01 KECK APF
    0 20 40 60 80
    -10 -5 0 5 10
    Time [JD]
    RV, Planet g [m/s]
    Q01 KECK APF
    0 500 1000 1500 2000
    -10 -5 0 5 10
    Time [JD]
    RV, Planet h [m/s]
    Q01 KECK APF
    2451000 2453000 2455000 2457000
    -5 0 5
    Time [JD]
    Residuals [m/s]
    Meschiari et al., 2009, 2010, 2011; Vogt, Burt, Meschiari+15; Rowan & Meschiari et al., Meschiari et al. 2015
    submitted; + 10 more papers

    View full-size slide

  8. Answering the why
    (2) HOW ROBUST IS PLANET FORMATION?
    Understanding how planets form in extreme conditions
    Meschiari et al., 2008, 2012, 2012, 2014; Meschiari et al., 2015, in prep

    View full-size slide

  9. Answering the why
    (2) HOW ROBUST IS PLANET FORMATION?
    Understanding how planets form in extreme conditions
    Meschiari et al., 2008, 2012, 2012, 2014; Meschiari et al., 2015, in prep

    View full-size slide

  10. Answering the why
    (3) EMPOWER OTHERS! "
    Designing apps & games for science and education
    Systemic Super Planet
    Crash
    Orbits GravityKit SAVE/Point
    kiosk
    Open sourced on GitHub
    http://github.com/stefano-meschiari

    View full-size slide

  11. My work in a nutshell
    THE SCIENCE SIDE...
    Getting astronomical data
    (telescope observations, new discoveries, …)
    Writing code
    (data analysis, physical simulations, …)
    2451000 2452000 2453000 2454000 2455000 2456000 2457000
    -30 -20 -10 0 10 20
    d[, 1]
    d[, 2]
    Science

    View full-size slide

  12. My work in a nutshell
    ...AND THE ASTRONOMY EDUCATION SIDE
    Writing code
    (HTML5/JavaScript web development)
    Education
    Designing apps & games
    (Systemic, Super Planet Crash, SAVE/Point, …)

    View full-size slide

  13. Radial Velocity data
    DOPPLER EFFECT + LOTS O’ STELLAR NOISE

    View full-size slide

  14. Radial Velocity data
    DOPPLER EFFECT + LOTS O’ STELLAR NOISE

    View full-size slide

  15. Radial Velocity data
    DOPPLER EFFECT + LOTS O’ STELLAR NOISE

    View full-size slide

  16. Radial Velocity data
    DOPPLER EFFECT + LOTS O’ STELLAR NOISE

    View full-size slide

  17. Radial Velocity data
    DOPPLER EFFECT + LOTS O’ STELLAR NOISE
    Cochran’s law (1992)
    “Thou shalt not embarrass thyself and
    thy colleagues by claiming false planets.”
    “Controversial” detections
    (about 10%)

    View full-size slide

  18. Systemic: a package/IDE for exoplanets
    LIVE DEMO TIME! HOPEFULLY NOTHING CRASHES ✌
    Systemic
    (the offline/science version)

    View full-size slide

  19. Systemic’s architecture & entry points
    MUST… NOT… FORGET… ANY OF THESE LANGUAGES…
    Java
    (UI)
    Pipe
    R
    (High-level
    scripting)
    Dyncall
    C
    (libsystemic)
    Multiple
    processes
    Multithreaded
    Linking
    GSL
    SWIFT
    (FORTRAN)

    View full-size slide

  20. Example: HD219134
    DATA SPANNING 11 YEARS
    -10
    0
    10
    20
    2004 2007 2010 2013 2016
    Date
    Radial velocity [m/s]
    Dataset
    APF
    KECK
    Q01
    Vogt, Burt, Meschiari+15

    View full-size slide

  21. Orbital plot
    INNER AND OUTER PLANETARY SYSTEM
    1 AU
    0.125 AU
    Inner planetary system Whole system

    View full-size slide

  22. 10
    Period of planet b
    Density
    3.0926 3.0930 3.0934
    0 1000 3000
    Mass of planet b
    Density
    0.006 0.010 0.014
    0 100 200 300
    Mean anomaly of planet b
    Density
    0 50 100 150
    0.000 0.010 0.020
    Semi-major axis of planet b
    Density
    0.038470 0.038474 0.038478
    0e+00 2e+05 4e+05
    Semiamplitude of planet b
    Density
    1.0 1.5 2.0 2.5
    0.0 0.5 1.0 1.5 2.0
    Period of planet c
    Density
    6.761 6.763 6.765 6.767
    0 200 400 600
    Mass of planet c
    Density
    0.005 0.010 0.015 0.020
    0 50 150 250
    Mean anomaly of planet c
    Density
    0 50 100 150 200
    0.000 0.005 0.010 0.015
    Semi-major axis of planet c
    Density
    0.06480 0.06482
    0 40000 80000
    Semiamplitude of planet c
    Density
    0.5 1.0 1.5 2.0
    0.0 0.5 1.0 1.5 2.0
    Period of planet d
    Density
    22.78 22.80 22.82
    0 20 40 60 80
    Mass of planet d
    Density
    0.015 0.025 0.035
    0 50 100 150
    Mean anomaly of planet d
    Density
    200 250 300 350
    0.000 0.010 0.020
    Semi-major axis of planet d
    Density
    0.14565 0.14575 0.14585
    0 5000 15000
    Semiamplitude of planet d
    Density
    1.5 2.0 2.5 3.0
    0.0 0.5 1.0 1.5
    Period of planet e
    Density
    46.66 46.70 46.74
    0 5 15 25 35
    Mass of planet e
    Density
    0.050 0.060 0.070 0.080
    0 20 60 100
    Mean anomaly of planet e
    Density
    240 260 280 300 320
    0.00 0.01 0.02 0.03
    Semi-major axis of planet e
    Density
    0.2349 0.2351
    0 4000 8000
    Semiamplitude of planet e
    Density
    3.5 4.0 4.5 5.0 5.5
    0.0 0.5 1.0 1.5
    Period of planet f
    Density
    93.6 94.0 94.4 94.8
    0.0 1.0 2.0
    Mass of planet f
    Density
    0.02 0.03 0.04 0.05
    0 20 40 60 80
    Mean anomaly of planet f
    Density
    0 50 100 150 200
    0.000 0.004 0.008
    Semi-major axis of planet f
    Density
    0.3735 0.3745 0.3755 0.3765
    0 200 600 1000
    Semiamplitude of planet f
    Density
    1.0 1.5 2.0 2.5
    0.0 0.5 1.0 1.5
    Period of planet g
    Density
    2050 2150 2250 2350
    0.000 0.004 0.008
    Mass of planet g
    Density
    0.25 0.30 0.35 0.40
    0 5 10 15 20
    Mean anomaly of planet g
    Density
    0 50 150 250 350
    0.000 0.003 0.006
    Semi-major axis of planet g
    Density
    2.95 3.05 3.15 3.25
    0 2 4 6 8 10
    Semiamplitude of planet g
    Density
    4.5 5.0 5.5 6.0 6.5 7.0 7.5
    0.0 0.4 0.8 1.2
    Eccentricity of planet g
    Density
    0.00 0.10 0.20 0.30
    0 2 4 6 8
    Long. of periastron of planet g
    Density
    0 50 150 250 350
    0.000 0.004 0.008
    Noise, Q01
    Density
    0.5 1.0 1.5 2.0
    0.0 0.5 1.0 1.5 2.0
    Noise, KECK
    Density
    2.0 2.5 3.0 3.5
    0.0 0.5 1.0 1.5 2.0
    Noise, APF
    Density
    1.5 2.0 2.5
    0.0 0.5 1.0 1.5 2.0
    MCMC
    NATIVE CODE + MULTITHREADED IMPLEMENTATIONS
    Markov-Chain
    Monte Carlo

    View full-size slide

  23. Residuals
    BEAUTIFULLY NORMAL & CORRELATED WITH STELLAR ACTIVITY
    -10
    -5
    0
    5
    0.20 0.25 0.30 0.35 0.40
    Activity
    Velocity residuals [m s-1]
    2000
    2005
    2010
    2015
    Year
    Dataset
    KECK
    APF
    Activity
    Velocity residuals [m/s]
    Q-Q plot of residuals Residuals vs. activity

    View full-size slide

  24. Systemic R package
    NATIVE CODE + MULTITHREADED IMPLEMENTATIONS
    - Bootstrap

    - Cross-validation

    - Local optimization 

    Simplex, gradient descent, Levenberg-Marquardt

    - Global optimization 

    differential evolution, simulated annealing, MCMC

    - Long-term integration over millions of years
    http://github.com/stefano-meschiari/Systemic2

    View full-size slide

  25. Widely used in the community
    …INCLUDING AS A CITIZEN SCIENCE TOOL
    More than 5,000 registrations, used to discover/
    characterize more than 40 planetary systems.
    Makes for a great citizen science/authentic science
    experience tool.
    Fully half of registrations from teachers, students,
    and non-academic users.
    Peter Jalowiczor Dominick Rowan

    View full-size slide

  26. Widely used in the community
    …INCLUDING IN THE CLASSROOM
    Top complaints:
    It’s too hard
    to install!
    It’s too hard
    to use!

    View full-size slide

  27. Systemic’s architecture for the Web
    MY MAKEFILE HAS A JAVASCRIPT TARGET
    Java
    (UI)
    Pipe
    R
    (High-level
    scripting)
    Dyncall
    C
    (libsystemic)
    Multiple
    processes
    Linking
    GSL
    SWIFT
    (FORTRAN)
    Emscripten
    “make -f Makefile.js”
    JavaScript
    (libsystemic.js)
    HTML5
    (UI)
    <br/>

    View full-size slide

  28. Systemic Live
    RUNS ON YOUR FAVORITE BROWSER (DESKTOP/MOBILE)

    View full-size slide

  29. “The online Systemic Console is a real gift to the community. […] I use this site
    to train both undergraduate and graduate students – they love the power of
    this program.” 

    — Debra Fischer, Yale University
    “Systemic is simple enough to use that it can provide a hand-on ‘virtual lab’ for
    a large general education class, […] students can get a taste of the scientific
    process even before they learn to program” – 

    — Eric Ford, Penn State
    “[...] After demonstrating the relevant equations in class, students have a good
    feel for the physics involved, but not until they start to try to fit real data and
    extract real planets do they realize the how subtle, complex, and fun the real
    process is. ”

    — Mike Brown, Caltech, Pluto-killer
    “I have used Systemic for several years in my class for advanced undergraduate
    physics majors. […] Systemic is extremely sophisticated, but easy to use, so it
    allows students to get a feeling for the tools used in real exoplanet research.”

    — Jonathan Fortney, UC Santa Cruz
    Testimonials
    RUNS ON YOUR FAVORITE BROWSER (DESKTOP/MOBILE)

    View full-size slide

  30. Systemic Live
    RUNS ON YOUR FAVORITE BROWSER (DESKTOP/MOBILE)
    Systemic is employed in classes at UT, Caltech, University
    of Florida, MIT, SJSU, Delaware, Yale, Columbia, UC Santa
    Cruz, and others, reaching ∼500 students to date, and is
    a homework set on the recurring MOOC “The Science of
    the Solar System”
    , taken by ∼13,000 Coursera students.
    Started to put a bit more serious thought about online
    astronomy education after that…

    View full-size slide

  31. Super Planet Crash
    A “MINIMUM VIABLE GAME” ABOUT GRAVITY

    View full-size slide

  32. Super Planet Crash
    A “MINIMUM VIABLE GAME” ABOUT GRAVITY

    View full-size slide

  33. ~15,000,000 games played by ~700,000 users
    since April
    Huffington Post
    The Verge
    Physics World
    SciAm News
    io9
    Space.com
    VICE
    The Creators Project
    & others

    View full-size slide

  34. Success!
    BUT… I BROKE MY WEBSITE $
    Imgur, io9
    The Verge,
    HuffPost, VICE,
    Hacker News,
    Twitter Google Analytics, April-May 2014
    CloudFront + some serious
    on-the-fly re-engineering of the
    game helped.

    View full-size slide

  35. What do 15,000,000 clicks look like?
    TAKE A GUESS.
    Orbital spin
    Habitable zone
    (get bonus)
    I recorded on the
    server the
    coordinates of
    every click on the
    game area (the
    circle).
    Where are most of
    the clicks?

    View full-size slide

  36. U/L bias
    Closer to the controls on the left-
    hand side of the window; People pay
    more attention to upper field of
    vision
    “Crosshair” bias
    Empirical preference for cardinal
    directions (related to “oblique
    effect”?)
    Orbital spin
    What do 15,000,000 clicks look like?
    [SHOWING A SUBSET OF 100,000]

    View full-size slide

  37. SAVE
    /Point
    http://save-point.io

    View full-size slide

  38. Stefano (PI)
    Design &
    development
    Randi
    Educational
    content
    Joel
    Music &
    story
    Greg
    Accessibility
    for VI people
    Pilot project received “seed” funding by UT for 2014-2015.

    View full-size slide

  39. Philosophy of the pilot project
    OR
    Touch Gesture Reference Guide
    Press
    Double tap
    Tap
    Press
    and
    drag
    Drag
    Press
    and tap,
    then drag
    1 2
    Press
    and tap
    Multi-finger
    tap
    1 2

    View full-size slide

  40. Philosophy of the pilot project
    • A suite of apps & games will
    provide students at all levels with a
    “virtual astronomy lab” in browser

    No installation hassle, can run on
    mobile platforms, easily accessible
    anywhere.
    OR
    Touch Gesture Reference Guide
    Press
    Double tap
    Tap
    Press
    and
    drag
    Drag
    Press
    and tap,
    then drag
    1 2
    Press
    and tap
    Multi-finger
    tap
    1 2

    View full-size slide

  41. Philosophy of the pilot project
    • A suite of apps & games will
    provide students at all levels with a
    “virtual astronomy lab” in browser

    No installation hassle, can run on
    mobile platforms, easily accessible
    anywhere.
    • Prioritize accurate physics, real
    astronomical datasets, game-level
    fun and engaging design.
    OR
    Touch Gesture Reference Guide
    Press
    Double tap
    Tap
    Press
    and
    drag
    Drag
    Press
    and tap,
    then drag
    1 2
    Press
    and tap
    Multi-finger
    tap
    1 2

    View full-size slide

  42. Philosophy of the pilot project
    • A suite of apps & games will
    provide students at all levels with a
    “virtual astronomy lab” in browser

    No installation hassle, can run on
    mobile platforms, easily accessible
    anywhere.
    • Prioritize accurate physics, real
    astronomical datasets, game-level
    fun and engaging design.
    • Easy scoring/grading systems for
    instructors.
    OR
    Touch Gesture Reference Guide
    Press
    Double tap
    Tap
    Press
    and
    drag
    Drag
    Press
    and tap,
    then drag
    1 2
    Press
    and tap
    Multi-finger
    tap
    1 2

    Classroom scores

    View full-size slide

  43. Philosophy of the pilot project
    • A suite of apps & games will
    provide students at all levels with a
    “virtual astronomy lab” in browser

    No installation hassle, can run on
    mobile platforms, easily accessible
    anywhere.
    • Prioritize accurate physics, real
    astronomical datasets, game-level
    fun and engaging design.
    • Easy scoring/grading systems for
    instructors.
    • Open source and easily
    customizable by schools and
    instructors.
    OR
    Touch Gesture Reference Guide
    Press
    Double tap
    Tap
    Press
    and
    drag
    Drag
    Press
    and tap,
    then drag
    1 2
    Press
    and tap
    Multi-finger
    tap
    1 2

    Classroom scores

    View full-size slide

  44. Orbits
    • An HTML5/JavaScript game (using
    Canvas)

    Backbone + Paper.js

    • All the level logic (story,
    interactions, properties) in an easy-
    to-edit YAML file

    Working on a customization interface
    for instructors

    View full-size slide

  45. Interactive small-scale exhibi
    A cut-down version of our
    applications could be ideal
    for small astronomy demos
    and exhibits.
    Our group just got a small
    department grant to outfit
    several floors of our
    buildings with wall-mounted
    iPads running educational
    apps.
    A lot of kids don’t know
    how to use a trackpad or a
    mouse!
    Small-scale exhibits, interactive kiosks,
    activities for Visually Impaired users
    Touch-based (+sound/haptic feedback)

    View full-size slide

  46. At SXSW Gaming Expo
    Seattle Museum of Flight
    Interactive kiosks at UT
    In classrooms

    View full-size slide

  47. 0.125 AU
    My work: overview
    Systemic
    • A tool to analyze exoplanetary data
    • A virtual lab for teachers & students
    SAVE/Point
    • A pilot platform for education
    • 4 games and apps ready to
    play
    • Discovering new exoplanets
    • Understand how they form
    • Creating apps for education & outreach
    0.0 0.5 1.0 1.5 2.0 2.5 3.0
    -5 0 5 10
    Time [JD]
    RV, Planet b [m/s]
    Q01 KECK APF
    0 1 2 3 4 5 6 7
    -10 -5 0 5
    Time [JD]
    RV, Planet c [m/s]
    Q01 KECK APF
    -10 -5 0 5 10
    RV, Planet f [m/s]
    Q01 KECK APF
    0 20 40 60 80
    -10 -5 0 5 10
    Time [JD]
    RV, Planet g [m/s]
    Q01 KECK APF
    0 500 1000 1500 2000
    -10 -5 0 5 10
    Time [JD]
    RV, Planet h [m/s]
    Q01 KECK APF
    -5 0 5
    Residuals [m/s]
    Thank you!

    View full-size slide

  48. Systemic R package
    1 10 100 1000 10000
    0 10 20 30
    [1] Residuals for 0-planet fit
    Period [d]
    Normalized power
    NATIVE CODE + MULTITHREADED IMPLEMENTATIONS
    Lomb-Scargle
    periodograms

    View full-size slide

  49. Case study: HD219134
    SUBSTANTIAL VARIABILITY

    10−3
    10−2
    10−1
    100
    101
    0.1 1.0
    S−index median
    S−index dispersion
    Observations






    0
    50
    100
    150
    200
    250


    HD 219134
    KECK stars
    Median activity
    Activity variability

    View full-size slide

  50. Check for stellar activity
    IS THE RV SIGNAL CORRELATED WITH THE ACTIVITY?
    -10
    0
    10
    0.20 0.25 0.30 0.35 0.40
    Activity
    Velocity [m s-1]
    2000
    2005
    2010
    2015
    Year
    Dataset
    KECK
    APF

    View full-size slide

  51. Check for stellar activity
    IS THE RV SIGNAL CORRELATED WITH THE ACTIVITY?
    [1997,2008] (2008,2012] (2012,2015]
    -10
    0
    10
    0.20 0.25 0.30 0.35 0.40 0.20 0.25 0.30 0.35 0.40 0.20 0.25 0.30 0.35 0.40
    Activity
    Velocity [m s-1]

    View full-size slide

  52. Planetary parameters
    EXTRACTED FROM THE MARKOV-CHAIN MONTE CARLO OUTPUT

    View full-size slide

  53. Check for stellar activity
    IS THE RV SIGNAL CORRELATED WITH THE ACTIVITY?
    -0.3
    -0.2
    -0.1
    0.0
    0.1
    0.2
    0.3
    0.4
    -400 -200 0 200 400
    Lag
    Correlation

    View full-size slide

  54. Model comparison

    View full-size slide

  55. Systemic R package - HD32963
    NATIVE CODE + MULTITHREADED IMPLEMENTATIONS
    P M MA e ω a K Noise Trend
    P
    Density
    M
    Density
    MA
    Density
    e
    Density
    ω
    Density
    a
    Density
    K
    Density
    Noise
    Density
    Trend
    Density
    Markov-Chain Monte Carlo modeling
    1 AU

    View full-size slide

  56. Kepler’s 2014 bonanza
    More than 700 new planets confirmed!
    Image Credit: NASA Ames/SETI/J Rowe

    View full-size slide

  57. Kepler’s 2014 bonanza
    More than 700 new planets confirmed!
    Image Credit: NASA Ames/SETI/J Rowe

    View full-size slide