Dr. Yuan, Zhen
My research focus is Deep Learning Galactic Archaeology and the Local Group Universe. I developed StarGO to search for stellar streams and substructures in the Milky Way halo. By taking the high-resolution spectra of these very old and metal-poor fossils, I try to decode galaxy formation and nucleosynthesis in the early Universe. Traveling away from the Milky Way, I am interested in dwarf galaxies in the Local group. Currently, I am developing ML tools to detect them from the mock images of the upcoming Chinese Space Station Telescope survey.
PhotCalib
I developed a neural network based method, PhotCalib, to calibrate CaHK narrow band photometric data from 11,000 images of the Pristine survey.
PhotCalib is designed to calibrate CaHK photometry from the ground-based telescope, CFHT based on the synthetic CaHK band generated from Gaia BP/RP spectra. The goal of the calibration is to fix the zero-point offsets of different fields and field-of-view variations (see the figures below) at the same time. The second correction is assumed to be constant during a MegaCam run, which starts with the remount of the camera once or twice a month. The FoV model is a three fully connect neuron layers, which is more flexible to remove subtle variations specific to each run and gives a flatter residual (right two panels of the 2nd figure below). Overall, the final calibration gives the systematic floor of 0.018 mag (histogram plot below).
During the summer of 2019, there was a hair left on one of the MegaCam lenses. The direct consequence is that all the photometric catalogues observed during a period of about 3 months suffer from this additional source of localized absorption. It causes the CaHK values to get fainter by up to ∼ 0.15 mag. PhotCalib trains a new nerual network in the affected region (region A below) after the standard approach to remove the hair feature.
Dwarf Galaxy Detection limit of the CSST survey
This is the first study to predict the dwarf galaxy detection limits for the upcoming Chinese Space Station Telescope (CSST) survey using the state-the-art method.
Chinese Space Station Telescope (CSST) survey that will cover 17,500 square deg of the sky with a wide field of view of 1.1 square deg. The point-source depth reaches 26.3 mag in the g band and 25.9 mag in the i band. The detection limit of Mv as a function of distance is fainter than most of the known Local Group dwarf galaxies. We expect CSST to open up an exciting discovery space for faint field dwarf galaxies. With its optical bands, wide survey footprint, and space resolution, CSST will undoubtedly expand our knowledge of low-mass dwarf galaxies to an unprecedented volume.
C-19 — A Globular Cluster or Dwarf Galaxy Stream ?
The C-19 stream is the most metal-poor structure ever discovered, and I have been working on searching its member stars in all-sky along the orbit, and characterizing its chemo-dynamical properties.
I obtained UVES spectra of seven member stars (including two new members) of C-19, the most metal-poor structure that has ever been discovered (Martin et al 2021, Nature). One newly confirmed member is located 30° away from the main body of C-19, implying that it is significantly more extended than previously known and that more members likely await discovery. Stay tuned! The elemental abundance of C-19 is consistent with a globular cluster (GC) origin, but dynamically it is hotter than most of the GC streams (see details in Yuan et al 2022, and the full member list can be downloaded here).
The comics I created on the left side are inspired by
the N-body simulations of the C-19-like system (Errani et al 2022) In order to have large velocity dispersion (~ 6km/s) consistent with the observation, the GC progenitor was embedded and pre-processed in a dark matter halo before the whole system accreted to the MW.
A brief History of the Cetus Stream
The Cetus stream is the first stream I identified after Gaia DR2 came out. It was discovered in SDSS (Newberg et al 2009) and named Cetus polar stream at that time. In this work (Yuan et al 2019), I identified ~ 150 members (green dots in the movie) from LAMOST K giant sample from their clustering in dynamical space. The new members show the orbital pole is around 45 degrees so I suggested dropping "polar" and named it Cetus stream. I also confirmed the association with NGC 5824 (green star) based on the similarity of their orbits and suggested the globular cluster is likely the core of the Cetus progenitor dwarf galaxy. To verify this scenario, Jiang Chang (PMO) did a series of N-body simulations (Chang & Yuan et al 2020). We found NGC 5824 was not the core, but located off-center in the Cetus progenitor in order to reproduce the stream (blue dots). Our fiducial model also predicts a vast extension of the stream in the South, surprisingly coincident with the Palca stream discovered in the DES (Shipp et al 2018).
The Cetus and Palca streams are confirmed to be the same structure and named the Cetus-Palca stream by Thomas et al 2022, using spectro-photometric distances of the 300 000 SEGUE stars derived from an artificial neural network. Around the same time, I reported our work of the searches of the southern Cetus in Gaia EDR3 (Yuan et al 2022b) from the fusion of two stream-searching tools, STREAMFINDER and StarGO. I found two wraps that extend over 100 degrees and belong to the Cetus system, one of which is the Cetus-Palca and the other is named Cetus-New. Our estimation of the Cetus progenitor mass is around 10^6 Msun, similar to the estimation from Thomas et al 2022.
Combining the bright Cetus members (G < 16) I have collected from all these works, I have performed a series of high-resolution spectroscopic followups and led HR-GO team to systematically analyze their elemental abunances. The Cetus progenitor dwarf galaxy has a similar size as Sextans and Ursa Minor, but is located much closer, thus, their spectra is much more achievable. Our on-going work will uncover the enrichemnt of elements in a low-mass dwarf galaxy and stay tuned!
The Low-Mass Stream-1
Blue horizontal- branch (BHB) and RR Lyrae (RRL) stars are excellent tracers of substructures of the ancient stellar halo. By augmenting the RRL and BHB catalog from LAMOST and SDSS and cross-matching with Gaia DR3, we get ~ 7600 stars with full 6-D kinematic information. I applied StarGO to this dataset and found a significant cluster from the Self-Organzing-Map, which reveals a dwarf galaxy stream (Yuan et al. 2020b). Compared to the Sagittarius stream, the new stream is made up of much fewer members and is dominated by a very metal-poor population. We thus name this substructure as the low-mass stellar-debris stream (LMS-1) shown as the purple scatter below. We also found two globular clusters M53 (NGC 5024) and NGC 5023 (star symbols) are dynamically associated with the LMS-1 based on their similar orbits. Moreover, this pair of GCs are embedded in the stream in 3D space, which strongly favors the scenario that they were stripped from the same parent dwarf galaxy. We did an N-body simulation following the GC orbit and were able to recover the structure of LMS-1.
After Gaia EDR3, the StreamFinder group re-discovered LMS-1 (Malhan & Yuan et al 2021), and found the main part of the stream, shown as the dense purple wrap below. Based on their members with full 6-D information, we fit the stream orbit and re-simulated the disruption history of its progenitor system.
Dynamical Relics of the Ancient Galactic Halo
What are the origins of the CEMP and r-process enhanced stars in the Milky Way Halo? Where do they come from and what are their progenitor systems like?
The motivation of this work is to find the linkages between chemically peculiar stars and the substructures in the nearby stellar halo. In particular, we believe CEMP and r-process enhanced stars carry the primordial nuclesynthesis information from the early Universe, and their birthplaces are most likely small dwarf galaxies dominated by very metal-poor population. Finding more halo stars originally from the same system will give us precious fossils to uncover the chemical evolution in ancient dwarf galaxies. With this aim, I applied StarGO to the LAMOST DR3 VMP catalogue and identified a series of existing and new substructures. I then identified their dynamically associated chemically peculiar stars, which very likely come from same disrupted progenitor systems. For example the dynamically tagged group (DTG) Rg5 (magenta triangles in the plot below) are associated to two known r-process enhanced halo stars (green stars). Their progenitor is possibly a low-mass dwarf galaxy which has experienced prolific r-process enrichment event, such as neutron star merger in the early Universe. In order to verify this scenario, I am leading systematic high-resolution spectroscopic followups on these VMP DTGs. Stay tuned ！
StarGO
StarGO stands for Stars' Galactic Origin. It is originally built to find halo stars sharing a common origin in the Milky Way (Yuan et al 2018). The core of StarGO is Self-Organizing-Map (SOM), and we developed a group identification approach based on the visualization of SOM.I applied StarGO to halo samples from LAMOST and SDSS in the dynamical space and discovered a new dwarf galaxy stream, LMS-1, new components of the Cetus stream, as well as a series of dynamically tagged groups (DTG) in the nearby very metal-poor stellar halo. We also applied StarGO to the nearby Gaia sample in the 5D astrometric space and discovered the tidal tails of the Coma Berencies open cluster (Tang et al 2019). StarGO can be applied to general clustering problems.
