$30 off During Our Annual Pro Sale. View Details »

Optical Galactic Plane Surveys

Optical Galactic Plane Surveys

a seminar for Cardiff University on 29 May 2013

Geert Barentsen

May 29, 2013
Tweet

More Decks by Geert Barentsen

Other Decks in Science

Transcript

  1. Optical Galactic Plane Surveys a seminar by Geert Barentsen for

    Cardiff University on 29 May 2013 Image: NGC 3293/3324 in VPHAS (credit: Hywel Farnhill)
  2. INT/IPHAS & UVEX VST/VPHAS UKIDSS/GPS VISTA/VVV SCUBA-2/JPS & SASSy Herschel/Hi-GAL

    Janet Drew Phil Lucas Mark Thompson In Hertfordshire, we <3 Milky Way surveys
  3. This talk 1. Introduction The IPHAS/UVEX/VPHAS surveys are mapping the

    entire Galactic Plane in visible light. 2. Scientific rationale These projects are a necessary counterpart to infrared surveys. Also a key complement to Gaia. 3. Demonstration Discovery and parameter inference of young stars using survey photometry.
  4. IPHAS+UVEX+VPHAS = EGAPS • EGAPS stands for “European Galactic Plane

    Surveys”; key members are from institutes in UK/NL/Spain. • Composed of two surveys in the North (IPHAS/UVEX) and one survey in the South (VPHAS+). • Covers the entire Galactic Plane at |b| < 5° in u’, g’, r’, i’, Hα (near-simultaneous). • 5σ-depth typically at g’ > 22 / r’ > 21. (complete to r’ ~ 19; saturated at r’ ~ 13.)
  5. Isaac Newton Telescope • Originally located in Herstmonceux, Sussex (1967-1979).

    • “Moved” to La Palma in 1984 (new dome, mount & primary). • Surveys apply for time each semester, like anyone else.
  6. VPHAS+ status • Data taking started early 2012. • Nearly

    20% done so far. • Data quality is looking superb • 0.8” median seeing in r’ (unguided); • 0.05 median ellipticity. • Reduced data of the first semester was handed over to ESO on 30 April - release imminent.
  7. IPHAS/UVEX status • Initial data releases in 2008 (IPHAS) and

    2011 (UVEX). • IPHAS Data Release 2 (DR2) is imminent • 95% of the Northern Plane; • 159 milion sources (80% detected at >2 epochs); • photometric calibration consistent with SDSS at the level of 3%.
  8. Single-band catalogues Band-merged catalogues Source catalogue Images Flag duplicate detections

    Cross-match Catalogue generation Quality control Reduction pipeline (Cambridge / CASU) Recalibrate
  9. Let Δij be the magnitude offset between exposures i and

    j; minimise ∑∑(Δij + ZPi - ZPj)2 Photometric re-calibration Exploit the overlaps between exposures (cf. Glazebrook et al. 1994)
  10. Let Δij be the magnitude offset between exposures i and

    j; minimise ∑∑(Δij + ZPi - ZPj)2 Photometric re-calibration Exploit the overlaps between exposures (cf. Glazebrook et al. 1994) Solve
  11. Secret weapon #!/usr/bin/env python import numpy import multiprocessing import astropy.io.fits

    import astropy.wcs cf. http:/ /github.com/barentsen/iphas-dr2
  12. r’-i’ / r’-Hα diagram H-alpha emission is common for a

    wide range of rare objects. (Corradi et al. 2008)
  13. r’-i’ / r’-Hα diagram 2005MNRAS.362..753D (Drew et al. 2005) Main

    sequence Reddening Breaks degeneracy between spectral type and reddening (e.g. Sale et al. 2009). Complement to Gaia: need to constrain extinction to get luminosities!
  14. This talk 1. Introduction The IPHAS/UVEX/VPHAS surveys are mapping the

    entire Galactic Plane in visible light. 2. Scientific rationale These projects are a necessary counterpart to infrared surveys. Also a key complement to Gaia. 3. Demonstration Discovery and parameter inference of young stars using survey photometry.
  15. Demonstration: T Tauri stars Shock emission (UV/optical excess) Hot inner

    disk (near-infrared excess) Warm dust & gas (infrared/radio) Hot gas; emission lines (including H-alpha) = young, solar-like stars; < 10 Myr; < 2 Msun
  16. 2 degrees = 30 pc (d = 900 pc) IC

    1396 in IPHAS Massive star (O6V)
  17. Image: Nick Risinger Observation bias: 88% of T Tauri stars

    known by SIMBAD are located at |b| > 5, where spectrocopic surveys are cheaper b = +5 b = -5 Orion
  18. Different environments at larger distances 150 pc 1000 pc >>

    1000 pc Taurus log N = 2 IC 1396 log N = 3 Tarantula Nebula log N = 6? Sun’s birth environment thought to be log N = 3-4? (Adams 2010)
  19. Spectroscopy is the gold standard, but survey photometry is cheap

    & deep: • Readily available up to 20th mag; • Homogeneous: few biases between regions; • Narrow-band filters provide “a low-res spectrum”. => Use photometric surveys to analyse objects across environments in a homogeneous way.
  20. r’ - i’ / i’ -J diagram: extinction can be

    constrained for low-mass stars (Black line: NextGen model track)
  21. Data Generative model Data Physics what you want: Physics what

    you know: Inference Parameter estimation
  22. P(physics | data) ∝ P(data|physics) P(physics) P(data | physics) Bayes’

    theorem “posterior” “likelihood” “prior”
  23. Application Given IPHAS/UKIDSS photometry {r', Hα, i', J, σr', σHα,

    σi', σJ} we aim to constrain {extinction, mass, age, accretion rate } taking “nuisance parameters” into account uncertain distance (760 ± 5 pc) uncertain inner disc radius (5 ± 2 R*) uncertain log LHα ~ log Laccretion (±0.43 dex)
  24. Application (cont.) We might assume that the model residuals are

    Gaussian, i.e.: P(data | physics) ∝ exp[ ∑(SEDmodel - SEDobs)2 / σ2 ] ... and assume a uniform prior: P(physics) ∝ 1 We want to know the parameter-space regions where the posterior is high... In this case, the peak correspond to a chi-squared fit.
  25. Chi-squared fitting is usually not what you want! • Your

    model is rarely ever “just Gaussian”. There are often a bunch of nuisance parameters with known distributions. • Astronomical data is sparse and hence there is often a family of degenerate solutions. A maximum-likelihood fit does not capture this. • Generic solution: write down your posterior and compute its distribution in full.
  26. MCMC Computing the posterior with “brute force” is often intractable.

    Instead, perform a “biased random walk” => Markov Chain Monte Carlo (MCMC) sampling methods
  27. Conclusions • IPHAS/UVEX/VPHAS are mapping the entire Galactic Plane in

    u’, g’, r’, i’, H-alpha. • Get in touch if you would like to use the data, or keep an eye out for “IPHAS DR2”. • Take-away message: Python and Graphical Probabilistic Models are key tools for mining surveys.
  28. Survey team Consortium: University of Hertfordshire (IPHAS/VPHAS PI) University of

    Nijmegen (UVEX PI) University of Cambridge (pipeline) University of Graz Other members: Instituto de Astrofísica de Canarias, Harvard/Smithsonian CfA, University College London, Imperial College London, University of Warwick, University of Manchester, University of Southampton, Armagh Observatory, Macquarie University, Tautenburg Observatory, ESTEC, University of Valencia. Key individuals: Janet Drew, Hywel Farnhill, Geert Barentsen, Robert Greimel, Mike Irwin, Eduardo Gonzalez-Solares, Romano Corradi, Paul Groot (UVEX lead), Danny Steeghs.