This page is still under construction!
My research focuses on the dynamics and formation of cosmic structures. In the era of large sky surveys, the increasing availability of astrometric and spectroscopic data allows us to study cosmic structures in phase space. Accordingly, decoding the observation requires a theoretical understanding of the dynamics and appropriate techniques of modeling and analyzing.
Small-scale challenges to cosmology
-
Recent observations reported a puzzling dearth of dark matter (DM) in a significant fraction of dwarf galaxies and high-z massive galaxies, raising a major challenge to the galaxy formation theory within the standard cold DM (CDM) cosmology. Any plausible solution to the galactic-scale challenges demands a more thorough understanding of the interplay between the DM halos and the baryons (stars and gas) and galaxy mergers.
-
As a major step forward, our recent work proposed the first self-consistent analytical model, CuspCore2, for the halos response to fast gas outflow or accretion. We then show how to form dark-matter deficient cores in high-z massive galaxies by episodic gas outflows and dynamical heating by mergers .
Galaxy formation at cosmic dawn
- Recent observations by JWST reported a puzzling excess of luminous galaxies in the early Universe at z~10. We propose a solution involving feedback-free starbursts (FFB) as a potential paradigm shift for early galaxy formation . The model prediction shows a good consistency with the present observations in general .
Orbits and masses of satellite galaxies
-
Starting from cosmological simulations, we built a universal and accurate model for the initial orbit distribution and the specific merger rate of infalling satellite galaxies, unifying its mass and redshift dependence. This model can help us better understand the halo structure and serve as the initial condition for semi-analytic models of halo/galaxy formation.
-
We showed that satellite galaxies of different galactic halos approximately follow the same present-day orbit distribution when properly scaled. On this basis, we constructed a universal satellite phase-space distribution function (DF) that is fully consistent with simulations, providing a realistic starting point for dynamical modeling using satellites. , .
Dynamical mass and boundary of the Milky Way
-
The virial mass and density profile of our Milky Way (MW) halo are crucial to many astrophysical studies. Most of the methods that have been proposed to constrain the MW mass profile make use of dynamical tracers (see our review paper ).
-
Based on scaling relations of satellite galaxy kinematics learned from simulations , , we obtained the current best constraints of the MW mass profile up to 300 kpc . We also demonstrated how the observed satellite kinematics can be used to test the galaxy formation models. Specifically, we showed that the MW satellite kinematics matches the EAGLE simulation better than the semi-analytical models.
-
We proposed a novel data-driven DF method that can infer the dynamical mass with the optimal statistical efficiency. Unlike conventional DF-like methods, it is independent of ad-hoc functional forms or orbital libraries. We applied this new method to measure the MW potential using satellites and globular clusters as tracers .
-
Following recent progress in characterizing the halo boundary (Fong & Han 2020) and using the motion of nearby dwarf galaxies, we first measured the depletion radius of the MW halo, a natural boundary separating a growing halo from the receding environment . The turnaround radius and the masses enclosed within the two radii are also estimated.
-
We compared the information content of different dynamical tracers and found that the satellite galaxies are the best tracers for galactic halos and galaxy clusters , because they convey less redundant information.
Machine learning and astronomy
-
The Gaussian process (GP) is a popular data-driven method for nonlinear regression; however, it is vulnerable to outlier contamination. We proposed a novel robust GP algorithm based on iterative trimming [code], which significantly outperforms the standard GP and its popular robust variants in most test cases with contaminated data. The advantages of robustness, efficiency, and computational tractability make it useful to a wide range of problems.
-
The proposed ITGP algorithm can precisely determine the main-sequence ridgeline of observed star clusters in the color-magnitude diagram, crucial for calibrating theoretical stellar models and accurately inferring cluster properties. It enables precise modeling of unresolved binaries in open clusters with unprecedented accuracy and precision, providing clear observational evidence of dynamical evolution .
-
We proposed a random forest machine learning approach to determine the accreted stellar mass fractions of central galaxies, based on various dark matter halo and galaxy features .