Dr. Károly Zsolnai-Fehér reviews magnetic simulations that now handle rigid and deformable bodies with two-way coupling. A yin-yang ferrofluid animation took nearly a month to render on consumer hardware, while newer surface-only methods run much faster than earlier full 3D approaches.
Questions this video answers
March 5, 2021.
LARRUZ (Dr. Károly Zsolnai-Fehér).
Nearly a month.
On a consumer machine's processor (not on supercomputers or networks).
Quadro RTX 6000, RTX 8000, and V100; claims availability of 48 GB RTX 8000.
Chapters
Dear fellow scholars, this is Two Minute Papers with Dr. Zsolnai. Have a look at these beautiful ferrofluid simulations from a previous paper. These are fluids that have magnetic properties and thus respond to an external magnetic field. And yes, you are seeing correctly, they are able to even climb things. And the best part is that the simulator was so accurate that we could run it side by side with with real life footage and see that they run very similarly. Excellent. Now running these simulations took a considerable amount of time to address this. A follow up paper appeared that showcased a surface only formulation. What does that mean? Well, a key observation here was that for a class of ferrofluids, we don't have to compute how the magnetic forces act on the entirety of the 3D fluid domain. We only have to compute them on the surface of the model. So what does this get us? Well, these amazing fluid labyrinths and all of these ferrofluid simulations, but faster. So remember the first work did something new but took a very long time. And the second work improved it to make it faster and more practical. Please remember this for later in this video. And now let's fast forward to today's paper. And this new work can also simulate ferrofluids. And not only that, but also supports a broader range of magnetic phenomena including rigid and deformable magnetic bodies and two way coupling too. Oh my, that is sensational. But first, what do these terms mean exactly? Let's perform four experiments and, and after you watch them, I promise that you will understand all about them. Let's look at the rigid bodies first in experiment number one, iron box versus magnet. We are starting out slow and now we are waiting for the attraction to kick in. And there we go. Wonderful. Experiment two, deformable magnetic bodies. In other words, magnetic lotus versus a moving magnet. This one is absolutely beautiful. Look at how the petals here are modeled as thin elastic sheets that dance around in the presence of a moving magnet. And if you think this is dancing, stay tuned. There will be an example with even better dance moves in a moment. And experiment number three, two way coupling. We noted this coupling thing earlier. So what does that mean? What coupling means is that here the water can have an effect on the magnet. And the two way part means that in return the magnet can also have an effect on the water as well. This is excellent because we don't have to think about the limitations of the simulation. We can just drop in nearly anything into our simulation domain, be it a fluid solid magnetic or not. And we can expect that their interactions are going to be modeled properly. Outstanding. And I promised some more dancing. So here goes Experiment number four, the Dancing Ferrofluid. I love how informative this compass is. Here it is a simple object that tells us how an external magnetic field evolves over time. I love this elegant solution. Normally we have to visualize the magnetic induction lines so we can better see why the tentacles of a magnetic octopus move or why two ferrofluid droplets repel or attract each other. In this case, the authors opted for a much more concise and elegant solution. And I also liked that the compass is not just a 2D overlay, but but a properly shaded 3D object with specular reflections as well. Excellent attention to detail. This is really my kind of paper. Now, these simulations were not run on any kind of supercomputer or a network of computers. This runs on the processor of your consumer machine at home. However, simulating even the simpler scenes takes hours, for more complex scenes, even days. And that's not all. The ferrofluid with the yin yang symbol took nearly a month to compute. So is that a problem? No, no, of course not. Not in the slightest. Because thanks to this paper, general magnetic simulations that were previously impossible are now possible. And don't forget, research is a process. As you saw in the example at the start of this video, with the surface only ferrofluid formulation, it may become much faster. Just one more paper down the line. I wanted to show you the first two papers in this video to demonstrate how quickly that can happen. And two more papers down the line. Oh my, then the sky is the limit. What a time to be alive.
This episode has been supported by Lambda GPU Cloud. If you're looking for inexpensive cloud GPUs for AI, check out Lamda GPU Cloud. They've recently launched Quadro RTX 6000, RTX 8000 and V100 instances. And hold on to your papers because Lambda GPU Cloud can cost less than half of AWS and Azure. Plus they are the only cloud service with 48gb RTX 8000. Join researchers at organizations like Apple, MIT and Caltech in using Lambda Cloud instances, workstations or servers. Make sure to go to lambdalabs.compapers to sign up for one of their amazing GPU instances today. Our thanks to Lambda for their long standing support and for helping us make.
Better videos for you. Thanks for watching and for your generous support and I'll see you next time.
