I run cosmological hydrodynamic simulations using GIZMO and understand the effects of feedback on galaxy formation and evolution. Most of my research is conducted by collaborating with the Feedback in Realistic Environments (FIRE) team. I have been working on a broad range of topics, primarily including the following:

☞ Simulating galaxies in the reionization era with realistic feedback

I am leading the "Simulating Galaxies at the Epoch of Reionization" project in FIRE-2. The broad scientific goal of this project is to make more realistic predictions for next-generation observational facilities (e.g. JWST), and to understand their contributions to cosmic reionization. The simulations take advantages of the realistic feedback presciptions from FIRE.

See this video for a talk I gave on this work at 2017 Santa Cruz galaxy workshop.

  • NEW! Stellar mass functions, luminosity functions at high redshift

    We have finished an initial suite of 15 high-resolution cosmological zoom-in simulations at z ≥ 5, covering a z = 5 halo mass range Mhalo = 108‒1012 M. In Ma et al. (2017c), we predict the galaxy scaling relations, stellar mass functions, and multi-band luminosity functions at z = 5‒12.

    We make these predictions public (read more).

    Figure: Ultraviolet luminosity function (UVLF) at z = 6. With canonical dust model, the simulations agree very well with observations.

  • NEW! Morphologies and sizes of high-redshift galaxies

    In Ma et al. (2017d), we investigate the morphologies and sizes of galaxies in the reionization era, using the recently built simulation sample.

    Figure: Stellar mass map (left), rest-frame UV image (middle), and rest-frame B-band image (right) for three galaxies at MUV ~ -16.5. Their UV images are dominated by small bright clumps where stars formed recently.

    This warns that the small, point-source like galaxies detected in the Hubble Frontier Fields (Bouwens et al. 2017) are possibly single star-forming clumps, given a shallow surface brightness.

☞ Escape fraction of ionizing photons from dwarf galaxies

I use a Monte Carlo radiative transfer code to compute the escape fraction of ionizing photons in galaxy simulations. An early version of the code was in SEDONA base (Kasen et al. 2006), and was working on uniform grid (Ma et al. 2015, 2016b). Recently, I upgrad the code to make it work on octree structure, so it better preserves the high resolution adopted in hydrodynamic computations.

See this video for a talk I gave on this work at 2015 Santa Cruz galaxy workshop.

  • Why binary stars provide much more photons for reionization?

    In Ma et al. (2015), we used three high-resolution cosmological zoom-in simulations of z ≥ 5 galaxies from FIRE-1, and found that the escape fractions of ionizing photons from these galaxies are only 5% on average, far below what is required (20%) for cosmic reionization.

    This is because most of the ionizing photons are produced by the youngest stars, which are always embedded in dense molecular clouds, so that these photons cannot escape (see Figure below). Only after a few Myr can feedback clear sightlines to allow ionizing photons to escape. However, according to canonical single-star stellar population models like STARBURST99, the ionizing photon emissivity declines rapidly after 3 Myr.


    Figure: Left: Gas density around stars of different ages. Top right: Ionizing photon emissivities for single-star and binary stellar population models. Bottom right: The effect of binary stars on ionizng photon escape fraction.

    In Ma et al. (2016b), we revisit this calculation using binary stellar population models BPASS. Binary stars have prolonged ionizing photon production rates. This boosts the escape fraction by a factor of 4‒10, making it sufficient for reionization.

☞ Galactic chemical evolution
  • We reproduced the observed diversity of the gas-phase metallicity gradients and their complex relationship with kinematic properties in galaxies at z = 0.5‒3, using more than 30 simulations from the FIRE-1 project. We argued that such diversity is driven by stellar feedback on time-scales of a few 100 Myr, and the observed metallicity gradients only reflect the instantaneous state of a galaxy (Ma et al. 2017a, read more, online data).
  • Using a Milky Way-mass disk galaxy simulation, we studied the stellar abundances and metallicity gradients in the galactic disk, and showed that they can be naturally understood from the structure, formation history, and dynamical evolution of the disk (Ma et al. 2017b, read more)
  • We predicted the shape and evolution of the galaxy mass‒metallicity relation (MZR) from z = 0‒6 using the FIRE simulations, and explained why the MZR should behave in such way (Ma et al. 2016a, read more).

Data release

NEW! Stellar mass functions, luminosity functions in the reionization era

In Ma et al. (2017c), we present an initial suite of 15 high-resolution cosmological zoom-in simulations at z ≥ 5 from the "Simulating Galaxies at the Epoch of Reionization" project in FIRE-2, spanning a z = 5 halo mass range Mhalo = 108‒1012 M.

We predict the stellar mass‒halo mass relation (§3.2), SFR(Mhalo, z) and SFR(M, z) relations (§3.3), magnitude‒stellar mass relation, and the inverse stellar mass‒magnitude relation at z = 5‒12 (§3.4). We further predict the stellar mass functions (§4), multi-band luminosity functions (§5, §6.2), and cosmic star formation rate and stellar mass densities (§6.1) at these redshifts.

We make our predictions publicly available (see Appendix A or the README file inside for detailed descriptions). The files include (1) the stellar mass functions and lminosity functions at rest-frame UV, B band, and J band (intrinsic stellar continuum emission, without dust extinction and nebular line emission) derived from the simulated catalog (SMF_sim_zxx.txt, LF_UV_sim_zxx.txt, LF_B_sim_zxx.txt, and LF_J_sim_zxx.txt, where xx represents the redshift in two digits), (2) model stellar mass functions and luminosity functions obtained by convolving the halo mass function with the best-fit scaling relations (SMF_model_zxx.txt, LF_UV_model_zxx.txt, LF_B_model_zxx.txt, and LF_J_model_zxx.txt), and (3) UV lluminosity functions after accounting for dust extinction (LF_UV_red_zxx.txt).

Click here to obtain our predictions.

☞ Gas-phase metallicity gradient in z ≥ 0 galaxies

In Ma et al. (2017a), we show the relation between gas-phase metallicity gradient and stellar mass, specific star formation rate (sSFR), and kinematic properties using the full simulated sample. To expedite the comparison with future observations and other theoretical predictions, we provide all relevant data from our simulations in Tables A1 in the Appendix of our paper and make public a machine-readable version of these tables.

Click here to obtain a machine-readable version of Tables A1 from Ma et al. (2017a).