The surface of a riverbed is typically lined by relatively large rocks, which protect the layers of finer sand and gravel beneath from erosion. Geologists have long thought that fluid mechanics control this pattern; the idea being that the flow of the river washes away the finer particles from the bed’s surface, leaving the larger particles behind.

But in a new study, geophysicists from Penn’s Department of Earth and Environmental Science in the School of Arts and Sciences, led by Associate Professor Douglas Jerolmack, took a different approach, considering the problem through the lens of granular physics. What they found was that riverbeds share something in common with a can of mixed nuts: larger particles tend to rise to the top.

“In granular physics, there’s a popular term people use, the ‘Brazil nut effect,’” says Jerolmack. “Basically, if you open up a container of mixed nuts, the Brazil nuts are the big ones and, somehow, all the shaking and jostling of the can causes them to be on the surface. We were able to show that this same particle segregation occurs in a model riverbed.”

The findings, published in Nature Communications, shed light on the mechanisms by which riverbeds form, and have implications for how they may erode. The work also makes new insights into the fundamental physics of particle segregation, which apply to all sorts of granular materials, from riverbeds and soils, to industrial and pharmaceutical substances.

In their work, the Penn team wanted to see if the Brazil nut effect might be at play in a riverbed, conceiving of it as a granular system, not just one subject to fluid dynamics. To do so, the researchers turned to a laboratory stand-in for a river: a doughnut-shaped channel filled with large and small spherical particles. The lid of the channel pushes the fluid atop the particles, replicating the flow of a river.

As Jerolmack and his team had shown in an earlier study, particles move along the riverbed in two ways: those at the top are pushed by the flow of liquid, while those deeper down creep along slowly due to the interaction among particles.

In the new work, the Penn researchers wanted to understand how these particles moved not just horizontally, but also vertically in the bed.

Using their custom-built channel and fluid embedded with a fluorescent dye, the researchers could scan through the entire depth of the channel, visualizing an entire plane of particles, even those buried under several dozen other particles.

“It’s almost like taking an X-ray of our granular sample, but with pictures,” Jerolmack says.

With the help of a software program, they were able to then track the horizontal and vertical positions of all of these particles through time. And they saw the Brazil nut effect in action.

“In this laboratory experiment of a very simplified river,” Jerolmack says, “we saw that, when we have a liquid pushing grains on the riverbed, those grains push grains underneath them that push grains that are underneath them, and so on, and it creates this jostling motion that allows large particles to kind of float up. So we confirmed that this general behavior that is seen in granular systems seems to also occur in rivers.”

Researchers have found it difficult to predict when rivers erode, or when hillsides dissolve into landslides, and these findings may help explain why these predictions have proved so elusive.

“We’ve been working on these problems for 100 years, and we still can’t predict with much certainty what fluid force is going to cause grains to start eroding,” Jerolmack says. “And that point changes through time. Riverengineering projects, bridges, and buildings all rely on estimates of the erosion threshold. I think we need to start from scratch with a new framework that incorporates granular physics.”

While these experiments and simulations can’t exactly replicate the complex conditions seen in rivers, such as turbulence, Jerolmack notes that the findings point to a need for integrating Earth science with fundamental physics research to advance knowledge in both spheres.

“Our inability to predict when erosion will occur, our inability to predict when a slow, oozing pile of dirt on a hill will suddenly become a landslide, is because we are up against our limit of the fundamental
understanding of how disordered materials behave,” he says. “We need to advance our understanding of the fundamental physics of disordered materials in order to have any shot at making predictions in the Earth-materials realm. And this is one problem where I think we’ve made a start at doing that.”