Science takes a shot at your afternoon pick-me-up

Perfecting pour-over coffee

Coffee is a ubiquitous pick-me-up, but climate change is making Arabica coffee plants—which are temperature sensitive—more challenging to grow. And that means coffee could become a liquid luxury. Physicists at the University of Pennsylvania wondered if they could use science to get more out of pour-over coffee.

“The main idea is whether it might be possible to use those coffee beans just a little bit more efficiently so that you can still get the same strength or the same extraction but maybe using slightly fewer beans,” says Arnold J. T. M. Mathijssen, who led the group.

The pour-over method for making coffee involves streaming hot water over coffee grounds in a cone-shaped filter that sits within a funnel and collecting the extraction liquid. Chemists will recognize the process as gravity filtration, and some pour-over equipment even uses chem-lab aesthetics reminiscent of scientific glassware.

Mathijssen and colleagues Ernest Park and Margot Young began their studies by mimicking the pour-over process with silica particles. Because silica is somewhat transparent but still scatters light, the team was able to use a laser and a high-speed camera to observe the physics taking place within the funnel. They saw that the water stream disrupts the silica to create a tiny avalanche within the cone. Streams of water poured from different heights produced varying amounts of avalanche activity. The physicists hypothesized that more avalanche activity produces better coffee extraction because there’s more contact between the water and the grounds.

“To test that with real coffee, basically what you do is you make a whole lot of pour-overs” with the kettle at different heights, Mathijssen tells Newscripts. Evaporating the liquid from each brewed cup and weighing the solids left behind shows which kettle height produced the best extraction (Phys. Fluids 2025, DOI: 10.1063/5.0257924).

Mathijssen has three tips for making the best pour-over. First, he says, pour slowly enough to give the grounds time to come into contact with the water but quickly enough that the volume of water can push the grounds around well. Second, the higher you hold the kettle above the coffee filter, the more velocity you’ll get from the stream of water. Third, don’t hold the kettle so high that the stream of water breaks up into droplets. The extraction data collected by the Penn team supports the avalanche effect, and droplets don’t do the trick.

The work has applications beyond baristas: it can inform how waterfalls impact soil erosion and could influence how chemists do gravity filtrations.

ChatGPT’s peculiar taste

If your afternoon coffee break also includes a sweet treat, like a brownie, it might interest you to know that generative artificial intelligence (genAI) also has opinions on this chocolate delight. University of Illinois Urbana-Champaign food scientist and perception expert Damir D. Torrico recently used the genAI tool ChatGPT as a sensory evaluator of 15 chocolate brownie formulations, some with common ingredients and some with unusual substitutions, including fish oil and worm meal (Foods 2025, DOI: 10.3390/foods14030464).

Torrico is interested in using AI to model sensory responses based on physical and chemical inputs. The idea is that genAI could help food product developers winnow hundreds of formulations to just a few for a human sensory panel to test.

To Torrico’s surprise, ChatGPT turned out not to be particularly discriminating when it comes to brownie recipes. It gave all 15 formulations a score of 8.5–9.5 out of 10. “ChatGPT was being, I think, polite in some cases, just saying that it really liked the food,” he tells Newscripts. But he notes that this positive bias is also commonly observed in human sensory panels.

The results suggest that ChatGPT isn’t great at evaluating recipes, but Torrico thinks that with some customization, genAI could speed up food product screening. Even so, he says, “I don’t think we’re going to replace humans.”

Please send comments and suggestions to newscripts@acs.org.