Ph.D. project

Lyman α radiative transfer in the high-redshift, dusty Universe

     Supervisors: Jesper Sommer-Larsen and Anja C. Andersen

Related links

  • Ph.D. thesis (13 Mb)
  • Ph.D. thesis (w/ papers; 21 Mb)
  • Ph.D. thesis (text only; 2 Mb)
  • Explanatory notes
    (in Danish)

  • Context

    The Lyman α (Lyα) emission line (λ = 1216 Å) is an essential diagnostic tool for probing galaxy formation and evolution. Not only is it usually the strongest observable line from the high-redshift Universe, but also from its shape and strength detailed information about its host galaxy can be revealed.

    Lyα is produced mainly by massive stars ionizing their surrounding neutral hydrogen, which subsequently recombines, but in young galaxies a significant fraction may also be caused by cooling radiation from accreting gas.

    However, due to the resonant scattering nature of Lyα photons, the radiation diffuses in a non-trivial way in both real and frequency space. This not only obscures the origins of the radiation and alters the spectrum, but also makes Lyα especially vulnerable to dust, as its path length may be increased by a large and unknown factor.

    Image of a typical Lyα emitting galaxy. It does not at all look like the beautiful spiral galaxies seen in the nearby Universe; nevertheless, just by looking at this little blob a plethora of informaiton about the conditions in the early Universe can be gained (credit: Fynbo et al. 2003).

    Numerical Lyα radiative transfer

    To be able to interpret correctly the information gained through Lyα observations, I have developed a numerical Lyα radiative transfer (RT) code (called MoCaLaTA), capable of treating an arbitrary three-dimensional and adaptively refinable distribution of source Lyα emission, neutral and ionized hydrogen density, temperature, peculiar velocity field, and dust density of the interstellar medium (ISM).

    The code is a "Monte Carlo" code, i.e. it follows the paths of a high number of photons as they scatter stochastically in real and frequency space until they either are absorbed by dust or escape the galaxy and travels through the intergalactic medium (IGM) toward a virtual observer, providing a detailed surface brightness (SB) map and spectrum of the radiation.

    The galaxies used for the RT are taken from high-resolution, fully cosmological/galaxy formation hydro-simulations (Sommer-Larsen et al. 2003, Sommer-Larsen 2006).

    All relevant quantum physics concerning the scattering processes is taken into account, and the dust is distributed in an environment-dependent fashion, taking into account both formation and destruction processes. Since the galaxies are simulated using smoothed particle hydrodynamics, while the RT assumes a cell-based structure, first the physical properties of the particles (density, temperature, etc.) are interpolated to an adaptively refined grid, capturing fully the high resolution of the SPH models.

    Animation of an adatively refined grid with five levels of refinement. The adative refinement makes it possible to reach extremely high resolution where needed (dense regions), while not wasting computer memory where the resolution is not needed (refresh page to animate).

    Schematic view of the Monte Carlo nature of the code. For every step in code which is governed physically by a probability distribution (e.g. direction of a scattered photon, velocity of the scattering agent, the absorption or reflection by dust grains), the outcome is determined by a random number generator.


    Extended and anisotropic Lyα emission

    The original motivation for what started as my master project was the observations that young galaxies are often observed to be more extended on the sky when observed in Lyα than when observed in continuum bands (e.g Fynbo et al. 2003).

    An early (non-adaptive, non-dusty) version code was applied to a simulated "Lyman-break galaxy" at z = 3.6 of representative physical properties (it evolves into a Milky Way-like galaxy at z = 0). It was found that proper treatment of the RT can reproduce the observations of Fynbo et al. beautifully, qualitatively and quantitatively (Laursen & Sommer-Larsen 2007).

    Furthermore, any anisotropic configuration of gas will also tend to produce an anisotropic emission of Lyα. Specifically, for a given galaxy the maximum observed SB can vary by an order of magnitude, and the total flux by a factor of 3–6, depending on the viewing angle. This may provide an explanation for differences in observed properties of high-redshift galaxies, and in particular a possible physical link between Lyman-break galaxies and regular Lyα emitters. This is discussed in Laursen et al. (2009a) where also the implementation of adaptive mesh refinement is introduced.

    Left: The Lyα emission of a simulated galaxy, i.e. how it would look if the Lyα photon did not scatter.
    Right: Surface brightness map of the Lyα radiation after having done the proper Lyα RT. In order to escape its host galaxy, the radiation must scatter its way out.

    Effects of dust

    Dust is ubiquitous in the ISM, even at high redshifts. The effects of dust on the Lyα RT are intensely debated, in particular the escape fraction (fesc), i.e. the fraction of emitted Lyα photons that is not absorbed by dust.

    Analytical approximations assuming homogeneous gas predict than close to none of the radiation should be able to escape. Since obviously some does escape, different scenarios have been invoked to explain this, e.g. galactic outflows, ionized paths, multi-phase medium, and viewing angle.

    Previous attempts to determine Lyα escape fractions from high-redshift galaxies have mainly been trying to match observed Lyα luminosities with expected, and different methods obtain quite different results; from a few percent to close to unity.

    To scrutinize the effects of dust in a more "ab initio"-ish way, I implemented a quite elaborate environment-dependent model of dust in my code, which lead to some interesting results:

  • As might be expected, but has not so far been applied, fesc is not a single number, but decreases with the size of the galaxy.
  • In spite of the dust being almost gray, the spectrum is affected in a highly wavelength dependent fashion, with the line often being severely narrowed. The reason is that different parts of the spectrum originates in physically distinct environments of its host galaxy.
  • For the same reason, the Lyα SB profiles of galaxies are "smoothed out", giving rise to an even more "flat" SB profile, which can then be interpreted as an even more extended SB profile.
  • The results, which are presented in Laursen et al. (2009b), seem quite insensitive to the various input parameters of the model. This is good, because we can rely on the results, but is also a nuisance since at least from Lyα observations, we should not expect to be able to learn much about the physical properties of the dust itself.

    Escape fraction fesc as a function of galactic mass Mvir.

    Lyα line from a simulated galaxy in the case of no dust (dotted), and in the more realistic case with dust (solid). The line is severaly narrowed.

    The impact of the intergalactic medium

    Once the Lyα radiation escapes the host galaxy, it travels through the IGM. Because of the AMR grid, the RT code is able to operate on a large volume while at the same time capturing the complex inhomogeneities of the individual galaxies. As radiation blueward of the Lyα line traverses the IGM, it is eventually redshifted into resonance and may thus be scattered out of the line of sight if there happens to be neutral hydrogen in that particular place of the Universe.

    A different RT code for the IGM RT, called IGMtransfer allowed us to quantify to which degree the Lyα line is skewed by the IGM (Laursen et al. 2010).

    Moreover, since the transmission of the Universe depends on the fraction of neutral hydrogen at differents redshifts, comparing average obtained transmission curves to observations we were able to put constraints on another highly popular subject --- the so-called Epoch of Reionization.

    Emission from a damped Lyα absorber

    Damped Lyα absorbers (DLAs) are huge clouds of neutral hydrogen, characterized by column densities NHI ≥ 1020.3 cm-2. The fact that DLAs are in their very nature self-shielded against the ubiquitous meta-galactic UV field implies that the gas is able to cool sufficiently to initiate star formation. This makes them obvious candidates for present-day galaxies, and a prime goal of modern cosmology should be to unveil the nature of DLAs. In spite of this, little is known about the relation between emission and absorption selected systems.

    At the NOT Summer School 2007, we (two other students and I) found tentative evidence (∼ 3.5σ) for the detection of emission in a DLA in the spectrum of quasar Q2348-011, a sign of ongoing star formation. In 2009, a succesful proposal for four more nights at the NOT allowed us to go somewhat deeper. Although the follow-up observations did not increase the significance of the detection to more than ∼3σ, other observations performed with the X-shooter at the VLT (PI: J. Fynbo, Dark; co-I: me), yielded an unequivocal detection. Moreover, by fitting synthetic spectra obtained with MoCaLaTA to the observations, we were able to constrain the physical properties of the galaxy counterpart of the DLA (Fynbo, Laursen, et al. 2010).

    Comparison of the Lyα line of a simulated galaxy with the observed. The observed blue peak is more suppressed than the simulated, but this could be explained by the observed galaxy counterpart having stronger outflow than the simulated.