Four scientists could soon live and work beneath the Florida Keys for days.The properties of water are unusual among liquids—it expands when it freezes, reaches its highest density just above freezing, and does not behave like the simple liquids chemists often use to understand matter.
Now, a new study in Communications Chemistry uses artificial intelligence to compare competing methods for measuring water’s hidden molecular order and determine which are truly effective.
Researchers at the University of Osaka developed a neural network to test 16 different ‘structural descriptors,’ which are mathematical tools for measuring how ordered or disordered water molecules are at the microscopic scale. Over the past 30 years, scientists have independently developed these descriptors, each with its own approach and scale, all aimed at explaining the same underlying puzzle.
Until now, researchers lacked a standard method for comparing them directly—a gap this study fills.
To establish a common standard, the Osaka team trained a neural network using molecular dynamics simulations of water at temperatures from 200 to 300 Kelvin. Using only the values from a single structural descriptor, it had to predict the temperature at which each molecular configuration was generated. If a descriptor captured meaningful structural information, the network could easily distinguish hot from cold.
“Past studies have shown that using machine learning to classify and understand structural data is effective,” says corresponding author Kang Kim. “We specifically wanted to incorporate a neural network model into this study to evaluate how accurate the descriptors were at capturing key structural information, in a way that is like human cognition.”
The network assessed each descriptor using a standard classification measure. To deepen their analysis, the team used an explainable AI technique called LIME. This method looks at how the model makes its decisions and checks whether its reasoning matches established physical principles, instead of relying only on statistical patterns.
Out of the 16 descriptors tested, two stood out. The Local Structure Index (LSI) and another, referred to as zeta (ζ), both of which measure the distance between the first and second layers of neighboring molecules, achieved nearly perfect accuracy across the tested temperatures. Another group, based on how hydrogen bonds connect, also scored well but was not as accurate as the others. Older descriptors that rely on tetrahedral angles or interaction energies became less accurate at higher temperatures, where molecules move more, and the differences blur.
“The network used what it had learned to compare how 16 descriptors differentiated between LDL and HDL structures at different temperatures,” says senior author Nobuyuki Matubayasi. “In this way, we determined the most efficient descriptors.”
The LIME analysis backed up the network’s results with evidence consistent with scientists’ expectations. Molecules with high Local Structure Index and zeta values led the model to predict lower temperatures, which is consistent with the idea that more ordered, LDL-like structures appear as water cools. This suggests that the network captured more than just statistical correlations.
With a common framework for comparing descriptors in place, researchers could attempt to develop descriptors that directly identify HDL and LDL environments rather than infer them indirectly. Whether these new tools will finally solve one of water’s biggest mysteries, the point where its two proposed liquid forms fully separate, is something scientists are just starting to investigate.
Austin Burgess is a writer and researcher with a background in sales, marketing, and data analytics. He holds an MBA, a Bachelor of Science in Business Administration, and a data analytics certification. His work focuses on breaking scientific developments, with an emphasis on emerging biology, cognitive neuroscience, and archaeological discoveries.
A team of physicists set out to recreate a 60-year-old quantum mystery using a large water tank. While their experiment produced the expected results, it also revealed an unexpected behavior in the water.
The experiment centered on the Aharonov–Bohm effect, a strange quantum phenomenon where electrons respond to a magnetic field even when they never actually pass through it. First predicted back in 1959, this effect has been notoriously difficult to confirm because the changes it causes in electron waves are incredibly subtle.
Instead of using electrons and a particle accelerator, researchers from the Okinawa Institute of Science and Technology (OIST), along with collaborators in Oslo and Chile, recreated the effect using water waves, making an otherwise invisible quantum effect visible in a tabletop experiment. The team reported their findings in Communications Physics.
“With waves traveling the opposite direction, you see a mirror image pattern,” says Jonas Rønning, co- author and former postdoc in the OIST unit. “The question for us was, what happens if you send waves from both directions at the same time? We thought that the patterns might cancel each other out, or both pitchfork-like patterns would be visible, but our intuition was completely wrong.”
The researchers constructed a custom tank, created a vortex at its center, and sent waves from opposite sides to meet in the middle. They used a high-speed camera to observe how the waves interacted on the water’s surface.
Without a vortex, the peaks and troughs remain in place. When the vortex is there, the standing wave breaks down. The vortex shifts the phase of the waves as they travel around it, altering the standing-wave interference pattern and creating rotating nodal lines. These are lines of momentarily flat water that spiral outward and turn in the opposite direction of the vortex.
“This was something new and unexpected,” says Aditya Singh, a PhD student in the Nonlinear and Non-equilibrium Physics Unit and co-first author of the study. “That’s what makes this fluid analogue system so valuable. It reveals topological effects — wave behaviors that occur across the whole system — that can’t be seen in quantum experiments.”
At first, the team suspected their equipment was causing the rotating lines. However, the same pattern also appeared in their computer simulations.
“When we first saw these lines, we thought they were an experimental artifact,” Singh says. “But when we also saw them in our simulations, we dropped everything and quickly worked out the mathematics underlying how they arise.”
As the strength of the vortex increased, more rotating nodal lines appeared, always moving in the direction opposite to the vortex. The study showed that the number of lines is quantized, meaning it is fixed to whole-number values determined by the vortex’s circulation. In other words, the topology of the vortex determines how many rotating nodal lines appear. This effect was observed across the entire wave field, not just near the vortex core as earlier models suggested.
This finding differs from Berry’s original experiment, which observed changes only near the vortex. In this case, the researchers discovered a non-local effect, a pattern that covers the entire system and depends on the overall setup. Instead of being limited to the area near the vortex, the rotating nodal lines appeared throughout the whole standing-wave pattern.
Although the practical uses of these newly discovered nodal patterns are not yet clear, the study shows that simple classical systems can reveal behaviors that are challenging to observe in quantum experiments. By making these effects visible, the researchers plan to explore more complex wave systems that could offer new insights into superconductivity and other quantum phenomena.
“One direction is to make the system more complex by introducing multiple vortices and arranging them into a lattice,” senior author Mahesh Bandi says. “That setup would mirror conditions in some superconducting materials, with the water waves behaving like a supercurrent. We don’t yet know what we’ll see — and that’s exactly what makes it worth doing.”
Austin Burgess is a writer and researcher with a background in sales, marketing, and data analytics. He holds an MBA, a Bachelor of Science in Business Administration, and a data analytics certification. His work focuses on breaking scientific developments, with an emphasis on emerging biology, cognitive neuroscience, and archaeological discoveries.
Nearly 170 billion tons of ice melt off the Greenland Ice Sheet every year, but scientists don’t fully understand the process by which this melting occurs. Researchers in the past suggested that liquid water forms at the glacier’s surface, where warm air and sunlight melt the ice. This meltwater travels down through cracks in the glacier and along the bedrock beneath it and flows out to sea.
However, some scientists view this model as too simple because it assumes water only flows downward and ignores how temperature affects water movement. They argue that an oversimplified melting model is unreliable for predicting how future ice sheets will behave.
One group of researchers tested this by investigating how water movement across Greenland is controlled by lakes beneath the ice sheet, called subglacial lakes. They monitored a newly discovered subglacial lake using high-resolution satellite images taken from 2012 to 2019 and 3D surface maps generated from the images. They focused on ice sheet changes from July 22 to August 1, 2014, during a suspected subglacial lake drainage event. A drainage event is like a water balloon that’s too full, bursting from excess pressure.
They examined images of the Greenland ice sheet taken from Landsat-8, 3D surface maps from the Polar Geospatial Center, and data from ICESat and ICESat-2 taken in the summer of 2014. Based on these data, the researchers identified a 2-square-kilometer (0.8-square-mile) region of ice rising 10 to 15 meters (about 30 to 50 feet), forming a dome on the ice sheet’s surface. They suggested that this dome formed when a large, meltwater-fed lake developed on the bedrock beneath the ice sheet, pushing the ice upward.
They also saw that on July 22, 2014, the dome began to collapse. Over the next 10 days, it dropped by 85 meters (about 280 feet), forming a basin. Using the dome’s dimensions, the researchers estimated that 90 million cubic meters, or about 3 billion cubic feet, of water drained from the lake at an average rate of 100 cubic meters, or about 3500 cubic feet, per second over the 10 days. This is roughly equivalent to 36,000 Olympic swimming pools draining at a rate of 1 pool every 25 seconds!
The researchers also found that during the collapse, 40-meter-high (130-foot) ice blocks were uprooted about 1 kilometer (0.6 mile) downstream from the basin, followed by 6 square kilometers (about 2 square miles) of smooth ice. The researchers hypothesized that these features formed when water burst through the ice, flowed along its surface, and then reentered the ice sheet.
Next, they used data from Landsat 5, Landsat 9, the National Snow and Ice Data Center, and the U.S. Geological Survey to demonstrate that the drainage event also affected the surrounding environment. When the water reentered the ice sheet, it flowed beneath the Harding Glacier downstream. This influx of fast-moving water reduced pressure at the glacier’s base, slowing its movement. Simultaneously, the increased drainage beneath the Harding Glacier caused 500 to 600 meters (about 1,600 to 2,000 feet) of ice to break off its end.
In response to these changes, the researchers hypothesized that subglacial waters rose to the surface rather than sinking to the bedrock because the ice sheet was frozen to it. To test this, they simulated temperatures at the base of the ice sheet using computer-generated thermal models. They input a range of likely bedrock temperatures and ice thicknesses to simulate possible conditions under the ice sheet. They found that all the simulations predicted the glacier’s base would remain below -5°C (23°F). At such low temperatures, the ice would be frozen to the bedrock before subglacial water could drain out. Since water couldn’t flow under the glacier, it was forced to travel upward.
The researchers used these findings to develop a new conceptual model of how glacial meltwater moves. First, surface ice melts and flows into a subglacial lake. As meltwater accumulates, pressure builds at the base of the ice sheet, forming a dome on the surface. A drainage event causes the dome to collapse. As water moves downward through the glacier, it is blocked from reaching the base because the ice is frozen to the bedrock. The water then moves upward, bursts through the surface, travels across it, and eventually reenters the glacier before draining to the ocean.
The team concluded that water movement above, through, and below glaciers is interconnected. Destructive upward flow of meltwater, followed by downward flow beneath glaciers, can weaken ice sheet structure and alter downstream glacier dynamics. Their work highlights the need for more scientists to consider the processes driving glacial ice loss.
The post How do ice sheets melt? appeared first on Sciworthy.
Login