Arthur T Knackerbracket has processed the following story:
We intuitively understand that the sand pouring through an hourglass, for example, forms a neat roughly pyramid-shaped pile at the bottom, in which the grains near the surface flow over an underlying base of stationary particles. Avalanches and sand dunes exhibit similar dynamics. But scientists at Lehigh University in Pennsylvania have discovered that applying a magnetic torque can actually cause sand-like particles to collectively flow uphill in seeming defiance of gravity, according to a September paper published in the journal Nature Communications.
Sand is pretty fascinating stuff from a physics standpoint. It's an example of a granular material, since it acts both like a liquid and a solid. Dry sand collected in a bucket pours like a fluid, yet it can support the weight of a rock placed on top of it, like a solid, even though the rock is technically denser than the sand. So sand defies all those tidy equations describing various phases of matter, and the transition from flowing "liquid" to a rigid "solid" happens quite rapidly. It's as if the grains act as individuals in the fluid form, but are capable of suddenly banding together when solidarity is needed, achieving a weird kind of "strength in numbers" effect.
Nor can physicists precisely predict an avalanche. That's partly because of the sheer number of grains of sand in even a small pile, each of which will interact with several of its immediate neighboring grains simultaneously—and those neighbors shift from one moment to the next. Not even a supercomputer can track the movements of individual grains over time, so the physics of flow in granular media remains a vital area of research.
But grains of sand that collectively flow uphill? That is simply bizarre behavior. Lehigh University engineer James Gilchrist manages the Laboratory for Particle Mixing and Self-Organization and stumbled upon this odd phenomenon while experimenting with "micro-rollers": polymer particles coated in iron oxide (a process called micro-encapsulation). He was rotating a magnet under a vial of micro-rollers one day and noticed they started to pile uphill. Naturally he and his colleagues had to investigate further.
[...] “Up until now, no one would have used these terms,” said Gilchrist. “They didn’t exist. But to understand how these grains are flowing uphill, we calculated what the stresses are that cause them to move in that direction. If you have a negative angle of repose, then you must have cohesion to give a negative coefficient of friction. These granular flow equations were never derived to consider these things, but after calculating it, what came out is an apparent coefficient of friction that is negative.”
It's an intriguing proof of principle that could one day lead to new ways to control how substances mix or separate, as well as potential swarming microrobotics applications. The scientists have already started building tiny staircases with laser cutters and videotaping the micro-rollers climbing up and down the other. One micro-roller can't overcome the height of each step, but many working collectively can do so, per Gilchrist.
DOI: Nature Communications, 2023. 10.1038/s41467-023-41327-1 (About DOIs).
(Score: 4, Interesting) by VLM on Thursday December 28, @08:19PM (3 children)
Every grain in a cement company sand pit, sure. F-ing around at home making piles of virtual sand is much more trivial.
Luckily piles of sand are pretty self similar and a big pile is mostly the same as a little pile, mostly, so no need to simulate a trillion sand grains if you just want pretty pix on the screen.
I would say granularity simulation is a step above porosity simulation. I'd start with porosity simulation before trying granularity studies. You can do pretty good simulated porosity studies using 70s hardware and software. I had to do this in school (about 20 years later LOL). Take a 2-d (or 3-d) rock. Randomly eat pixels/voxels. depth first search along one axis or one plane to the other side. Did you make it? if so its porous. If not, it aint. Make a dot graph of the percent of randomly removed pixels vs porosity status. You'll get some weird ass remarkably near-constant number for a porosity threshold I can't remember that seems to come out of nowhere, like its a constant of the universe or something.
Granularity studies are not "that" much more complicated. I'd suggest starting with a dumb simulation where more than 50% overhang means it tips over and falls off (aka the x-axis delta between the current position of the grain and the grain its on top of, is larger than half the particle radius). You'll get nice piles with simple code. The logical next step is simulating momentum in the x-axis and then simulating rotating torques and rolling after landing and stuff like that.
A cousin of "grains fall from the sky into a pile" is the more complicated liquefaction studies. So you got a stable-ish plop of sand. What happens when it's vibrated with various intensities of force or soaked with water so the average density changes? Again you can do a lot of fun stuff with msbasic back in the 1980s or scale up to supercomputer nonsense in 2023 (soon to be 2024).
This kind of shit used to be the stereotypical "how to teach the noobs how to PVM" or whatever your cluster system of choice. I suppose they still teach using this technique for noob Apache Spark (ab)users.
This is kind of the visualspatial analogy of F-ing around with prime numbers. You can do a remarkably good job with msbasic in 1981, although crude, all the way up to some pretty fancy supercomputer cluster simulations in 2023 and its a pretty smooth continuous function of F-ing around and a shallow-ish learning curve.
I bet, its possible to do a shitty sand pile simulator in a dozen LoC using Python, and that can be scaled up to pretty elaborate 3-d simulations of sand piles if you want. Its kind of "fun" as a programming challenge if you're bored with project euler type learning tasks.
(Score: 2) by crafoo on Thursday December 28, @08:37PM
Project Euler is pretty fun.
generally agree. they're talking about 10s of thousands of encapsulated grains interacting with each other, which seems quite do-able on current hardware. Probably out of the box Bullet3D, Bullet3D on GPU and modified for their purposes would be just fine on a desktop.
(Score: 3, Interesting) by Unixnut on Thursday December 28, @10:44PM (1 child)
Your post reminded me about an old game I used to enjoy playing called "The Powder Toy [powdertoy.co.uk]." At its core is physics simulations, such as the grains of sand you mentioned, but it simulates other substances as well as their interactions with each other. It was hours of fun for me, especially the Nuclear power stations!
However it does show that within a limited domain, you can simulate a lot of physics, and that is just with a single CPU, let alone the multiprocessor machines nowadays. I even remember that before the AI craze, GPUs started being used for physics simulations, giving even workstation size machines the ability to do a good amount of simulation using SIMD calculations on GPUs.
So I agree with you, given a limited domain you should be able to simulate sand interactions. Sure not a sand pit or a beach, but I can imagine a small hourglass should be doable.
(Score: 3, Informative) by VLM on Friday December 29, @12:07AM
Ironically won't matter for sand. I haven't been keeping up on research but almost exactly a decade ago there was a "famous" journal article that proved once and for all that granular media is fractally self similar at all scales larger than "pretty small", so a big pile of sand behaves like a small pile of sand, so ironically stuff we were doing in the 80s was pretty accurate. I mean, yeah, people always "knew" you can't quantitatively determine the size of a picture of a sandpile unless there's a ruler in the pix because they all look the same, but these dudes 'proved' it using fractal math geometry tools or whatever.
So, the angle of repose you get from simulating a pile of 10K sand grains won't be much different than 10M grains or 10B grains.
Your powder toy link looks pretty cool like a ultra high resolution modded minecraft sort of, but not really.
The genre of nuclear power plant simulator was much more popular in the 80s and 90s. The fickle changing tastes of gamers I guess. I remember several from the command line and CLI UI era. IIRC there was an "80-Micro" cover story nuclear power plant simulator. And I found it while typing this post, its a cover story from March of 1984. The author built a very nice simulator panel that I'm sure could be replicated by a small herd of Arduinos.