thor: a massively-parallel GPU-accelerated MCRT code

I am a postdoctoral researcher at the Institute for Theoretical Astrophysics in Heidelberg, where I lead theoretical studies of galaxies, their diffuse gas, and their embedding cosmic structure. I designed and ran some of the largest simulations of galaxy formation. To this end, I develop high-performance tools and algorithms to analyze these simulations, and to test our understanding of galaxy formation and evolution since the Big Bang. I am particularly interested in the diffuse gas surrounding galaxies, which is now increasingly mapped in emission and absorption with modern telescopes. In Heidelberg, I work with Dylan Nelson, running and analyzing cosmological galaxy formation simulations, as well as development of new high-performance codes. Before moving to Heidelberg in 2021, I obtained my PhD in Physical Cosmology at the Max Planck Institute for Astrophysics in Eiichiro Komatsu’s group.

For students: Learn more about open projects available for bachelors and masters theses in computational astrophysics at the University of Heidelberg.

Interests

Circumgalactic Medium and the Cosmic Web
Cosmological Galaxy Formation Simulations
Scientific Big Data
Lyman-alpha Radiative Transfer
Single-degenerate Type 1a Supernovae

Education

PhD in Physical Cosmology, 2021

Max Planck Institute for Astrophysics (LMU Munich)
MSc in Physics, 2017

University of Goettingen
BSc in Physics, 2015

University of Goettingen

Projects

thor: a massively-parallel GPU-accelerated MCRT code

Emission and absorption line features are important diagnostics for the physics underlying extragalactic astronomy. The interpretation of observed signatures involves comparing against forward modeled spectra from galaxy formation simulations as well as more simplified geometries, while including the complex scattering radiative transfer (RT) of resonant emission lines. We introduce thor, a modern C++ radiative transfer code focused initially on resonant emission lines. thor is a high-performance, distributed memory MPI-parallel, multi-target code, running on CPUs, GPUs and other accelerators, yielding large $\sim 10-50\rm{x}$ speed-ups compared to previous CPU-only codes.

cosmosTNG

cosmosTNG is a constrained cosmological galaxy formation simulation suite in the COSMOS field at Cosmic Noon. CLAMATO Lyman-alpha forest flux and zCOSMOS galaxy positions are used to reconstruct the initial density field, which is subsequently evolved with the TNG galaxy formation model with the AREPO code. The simulation suite consists of 8 variations of the unconstrained small-scale modes evolved down to $z\approx 2.0$, containing the well-studied zFIRE protocluster region, which furthermore overlaps with JWST treasury programs COSMOS-Web and PRIMER.

Simulating Lyman-alpha Emitters

Below plots show first results of our new radiative transfer code for meshless structures applied to individual halos in the IllustrisTNG simulations. Lyman-alpha emitter after radiative transfer. Surface brightness in erg/s/cm$^2$/arcsec$^2$. Lyman-alpha emitter after radiative transfer. Artificially lowered neutral hydrogen density by a factor of $10$, revealing the radiative transfer smoothing out the emission from the star forming regions. Surface brightness in erg/s/cm$^2$/arcsec$^2$. The neutral hydrogen column density responsible for scattering out the injected photons in star forming regions.

IGM Interaction of Lyman-alpha Emitters Spectra

Lyman-alpha emitters (LAEs) show a rich variety of spectral shapes due to the emission line’s resonant nature and typically high optical depths. While there is a large body of literature exploring how small-scale density and velocity distributions can explain this variety of features in spectra, the intergalactic medium (IGM) has often been neglecting as a contributing factor for such features. Above sketch helps visualizing how the IGM density and velocity structure along a line-of-sight give rise to an attenuation profile possibly shaping the arising spectrum.

Radiative Transfer on Voronoi Meshes

An increasing amount of astrophysical and cosmological simulations are carried out on a moving unstructed mesh defined by the Voronoi tessellation. Photons are spawned in a Monte Carlo fashion from emitting gas cells. At each scattering the contribution reaching the observer along specified lines of sight is computed. Lately, we expanded the priorly used code in Behrens et al., 2019 ( public version here) to work on such meshless structure. This will ensure the code’s relevance in the future and application to new simulations that would not have been able to be processed with prior code due to the larger memory requirement due to an intermediate interpolation step.

Single-Degenerate Type 1a Supernovae Simulations

Binary Chandrasekhar-mass white dwarfs accreting mass from non-degenerate stellar companions through the single-degenerate channel have reigned for decades as the leading explanation of Type Ia supernovae. Yet, a coherent theoretical explanation has not yet emerged to explain the expected properties of the canonical near-Chandrasekhar-mass white dwarf model. Such near-Chandrasekhar-mass SNe Ia are preceded by a simmering phase within the convective core of the white dwarf, leading to the ignition of one or more flame bubbles scattered across the core.

Impacts of the Large Scale Structure on Detected Lyman Alpha Emitters

In 1967 Partridge and Peebles theorized that young galaxies at high redshifts emitting Lyman-α photons might be a suitable tracer of large-scale structure. Those distant galaxies of high Lyman-α emission, so-called Lyman Alpha Emitters (LAEs) can be used to constrain the cosmological standard model at high redshifts and furthermore allow insight for the environment of those galaxies. The Lyman-α line corresponds to the energy difference from the ground state to the first excited state of neutral hydrogen.