Radiative Feedback in Pop III Star Formation
The first stars (so called Pop III stars) in the universe were born a few hundred million years after the Big Bang often in the center of Dark Matter "minihalos". They differ from present day stars in that they formed from "pristine gas" of hydrogen and helium left over from Big Bang nucleosynthesis. Being deprived of the heavier elements produced in supernova explosions, they did not form from gas that cools as efficiently as present day stars. As a result they are often more massive than present day stars. No Pop III star has yet been conclusively detected though there are some eligible candidates. The chance of actually observing one has increased with the recent launch of JWST. A crucial way to understand these elusive objects is therefore through analytic work and small-scale, full-physics, high-resolution simulation.
In Chen et al. 2026 we analyze the evolution of Pop III stars for the first ~50,000 years after their creation using a unique suite of simulations that self-consistently incorporate turbulence, magnetic fields, and radiation feedback. We analyze, among other things, how different radiation prescriptions affect the stellar mass, and how HI-Shielding influences the impact of Lyman–Werner (LW) radiation. In the top plot below we see the masses of the stars as a function of time from their creation for each of our runs. Only in the LW-Only run (pink curve) does the disk fragment and creates multiples stars. In our fiducial run (blue curve) which includes HI shielding, LW radiation is effectively suppressed and the protostar continues to grow throughout the simulation. In the run without HI shielding (yellow curve) we see a suppression of stellar mass growth. This is because in the absence of HI-shielding, the LW radiation can more effectively phtotodissociate H2. This increases the temperature near the protostar which provides an additional thermal pressure support and hinders stellar accretion. The grey bands indicate the expected outcome of the death of these Pop III stars. We see that the inclusion of HI-shielding may move stars to the pair instability supernova track. In the bottom plot we show various slices plots at different times in our fiducial run. We see that the HII region (left side of each panel in top row) remains confined, extending ∼100 AU beyond the sink accretion radius (75 AU), Dense gravitationally bound gas sustains high recombination rates and prevents sustained pressure-driven breakout. Turbulence and magnetic fields may also contribute to its confinement, even at high ionizing luminosities.
Turbulence Generated by The Stream Velocity
For the first ~380,000 years after Big Bang the universe was a dense primeval "fireball" where baryons (protons and neurons) were coupled to the photons via their mutual coupling to the electrons (the former through Thomson scattering and the latter through the Coulomb force). The radiation pressure of the photons prevented the baryons from fully gravitationally collapsing into potential wells and created standing oscillating waves known as Baryon Acoustic Oscillations (BAO). The Dark Matter (DM), which doesn't interact electromagnetically, did not feel the radiation pressure, and so gravitationally collapsed into "potential wells". As a result of the different physics governing the motions of the baryons and DM, they had a relative velocity with respect to each other. At the time of recombination the baryons decoupled from the photons and their velocity dropped from being highly relativistic to that of a thermal ideal gas, and the relative velocity between the baryons and DM became supersonic.
Using high-resolution cosmological simulations in Chen et. al. 2025 we explore how this relative velocity, which is often called the "Stream Velocity", affects turbulence inside DM halos at the epoch of the first stars (a.k.a Pop III Stars). We found that the Stream Velocity boosts the turbulence in low mass halos and, rather surprisingly, suppresses it in high mass halos. This is illustrated in the top figure where we show the turbulent Mach number as a function of halo mass at z = 20 for the no-Stream Velocity case (red curve) and the Stream Velocity case (blue curve). For low-mass halos the additional kinetic energy created by the Stream Velocity is transformed into turbulent motions, boosting the turbulence. While for high-mass halos the Stream Velocity actually suppresses accretion-driven turbulence and merger-driven turbulence, two other ways of generating turbulence in the primordial universe. This suppression of turbulence in high-mass halos is caused by how the Stream Velocity modulates the density field in addition to the velocity field. Read the paper for more info! The bottom figure shows typical low-mass halos for no-Stream Velocity (top panel) and Stream Velocity (bottom panel).
Measuring Magnetic Fields using VGT and VDA
Magnetized turbulence pervades the ISM and plays a central role in cosmic ray transport, star formation, and the regulation of heat and mass flows. There are multiple techniques for measuring these magnetic fields, including Faraday Rotation (which uses polarization of EM waves when interacting with a magnetic field), Zeeman Effect (which uses the shifting of spectral lines in the presence of a magnetic field), and dust grain alignment (which uses the perpendicular alignment of tiny filings with the magnet field). The Velocity Gradient Technique (VGT) is a relatively new method that can infer the plane-of-sky magnetic field orientation from spectroscopic HI channel maps. It uses turbulent scaling relations and the statistical properties of MHD turbulence. In brief, because turbulent eddies in magnetized flows are elongated along the local magnetic field (see Goldreich & Sridhar 1995), the velocity gradients are largest perpendicular to the field: for a given velocity fluctuation, the smaller perpendicular scale length produces a larger gradient. The Velocity Decomposition Algorithm (VDA) is a method developed by Ka Ho Yuen that separates out the velocity (aka velocity caustics) and density components of channel maps, which can, among other things, tell use the phase and turbulent properties of a region in addition to providing a cleaner map to preform VGT.
In Yuen, Chen et. al. 2023 we used the VDA to show that a recently discovered large filamentary feature on the outskirts of the Milky Way, perhaps the largest ever found, named Cattail, is likely not in the CNM (as the authors suggested) and consists of multiple turbulent systems. We showed this in part by first noting that the main body of Cattail (yellow boxes in left side of top figure), looks the same in channel maps. VDA showed that this only happens in the presence of strong thermal broadening. Since the CNM is supersonic it is unlikely to contain strong thermal broadening. Further when analyzing the velocity caustic map (right side of top figure) we found that the ratio of the power of the density to velocity fluctuation were less than 1, meaning that the turbulence contributes significantly to the line width (3 - 5 km/s), further precluding strong thermal broadening. We also found the density power spectra was different for the yellow region and the green region of Cattail. The former had a density spectral slope of approx. -3.7 a signature of weakly magnetized turbulence while the latter had a density spectral slope of -2, indicating the feature is magnetized and parallel to the magnetic field. The power spectra of the velocity field, decomposed by VDA, followed the Kolmogorov cascade of -5/3. The channel map itself did not reveal this scaling relation, highlighting the power of the VDA. In the bottom figure we show the magnetic field measure using VGT and VDA.
