Vortices, magnetic fields, convolutional neural networks, and dust rings

Here is a summary of a backlog of papers led by my collaborators.


Spotlights in protoplanetary disks

Hammer, Lin, Kratter, & Pinilla (2021, MNRAS, 504, 3963)

Michael Hammer, now a postdoc at ASIAA, took a detailed look into the evolution of protoplanetary disks harboring massive planets. We were motivated by the following conundrum. Recent theoretical models and some observations suggest that protoplanetary disks are only weakly turbulent. Planets formed in such disks are likely to carve out gaps that become unstable to vortex formation, which would translate to asymmetric observations of the dust distribution. See the above figure for synthetic images of what one would see. However, the majority of substructures observed in protoplanetary disks are axisymmetric dust rings and gaps. One possibility is that disks are simply more turbulent than previously thought, but the morphology also depends on the planet mass and the disk temperature. For example, sub-Saturn mass planets in cold disks sustain asymmetries as they reform after disappearing. On the other hand, in hot disks, a vortex can survive for a long time.


Ups and downs in a magnetized disk

Cui and Lin (2021, MNRAS, 505, 2983)

Can Cui, a postdoc at Cambridge University, investigated how the vertical shear instability (VSI), operates in magnetized protoplanetary disks. The VSI is recognized to drive low levels of turbulence in protoplanetary disks and could be important for dust evolution. While the VSI was discovered in a hydrodynamic setting, real protoplanetary disks are magnetized, albeit subject to non-ideal effects. We found that magnetic fields tend to stabilize the VSI. This can be seen in the figure above, where growth rates (in color) diminish as one moves from the right to the left as the gas-to-magnetic pressure ratio (β) drops.


Fun fact: we also found that an axisymmetric isothermal disk threaded by a magnetic field is mathematically equivalent to an adiabatic, unmagnetized disk with the same geometry.


A picture is worth a thousand planets

Auddy, Dey, Lin, & Hall (2021, MNRAS, 920, 3)

In our previous work (Auddy & Lin, 2020), led by former ASIAA postdoc Sayantan Auddy, we applied artificial neural networks to model dust gaps produced by planets embedded in protoplanetary disks. This allows us to quickly predict the mass of a gap-opening planet based on the morphology it induces. However, this required us to manually process images to extract gap features, such as its width.


Here, we extend this approach to convolutional neural networks. The framework can now take entire images as input. This circumvents the ambiguity in the description of dust gaps, for which different authors adopt different definitions. In other words, we automate the process of image analyses. When supplied with additional disk parameters (see the figure above), this hybrid approach can predict more accurate planet masses than our first model.


Survival of the dustiest

Lehmann & Lin (2022, A&A, in press)

ASIAA postdoc Marius Lehmann carried out detailed 2D and 3D simulations of dusty protoplanetary disks. These models also include a pressure bump, which has been invoked as preferential sites for planetesimal formation due to its ability to concentrate dust. But what happens when the disk is turbulent?


Consistent with previous work (Lin 2019), dust settling is hampered by turbulence, but can overcome it if the dust abundance (Z) is sufficiently enhanced. This is shown in the top panels in the above figure. However, it appears that pressure bumps alone cannot sustain a dust concentration in the midst of turbulence. We also find that dusty vortices form readily at low dust abundances, but are suppressed at sufficiently high dust-loading. Instead, dust rings persist in the pressure bump, as shown in the bottom panels.


For this work, we took full advantage of the TAIWANIA-2 GPU cluster, a.k.a. the Taiwan Computing Cloud, hosted by the National Center for High-performance Computing.

36 views0 comments

Recent Posts

See All