An espresso puck is still mostly dry one second into the shot

A shot of espresso reads as one event: the pump starts, the coffee comes out. Inside the basket it is at least two. A group working with a rotating x-ray scanner has now filmed the first one, and the footage shows an espresso puck still largely dry a full second into the shot, with water standing on top of the coffee grounds because what is underneath cannot take it fast enough.

What the x-ray scanner saw inside the coffee bed

The work was done on the rotating x-ray system at the Fraunhofer EZRT development centre in Germany, a source and a detector riding a ring that turns around the object instead of turning the object itself. Spinning a brewing machine at a revolution a second would fling the water around by centrifugal force. The team ran the source at 160 kV, took 1000 frames a second through one full rotation, and split 30 seconds of brewing into 30 one-second reconstructions.

The machine was an ordinary domestic espresso machine, a De’Longhi EC685 Dedica, stripped of the metal parts that spoiled the reconstruction, without changing how it brewed. The coffee was 100% arabica, ground on a Demoka mill, 10 g in the basket, tamped by hand until a scale below read 12 kg.

The first drops go almost nowhere

On the fine grind, timed from the moment the first drop touched the surface, very little happened. The first drop sank only a short distance. Over the next 2 seconds the water spread out across the top and stayed there, barely penetrating. From second 4 it began to push in properly, and by second 7 the bed was wet all the way through. The paper, published in Physics of Fluids in January 2025, fits a model to the same run: water starts ponding on the surface at 0.823 s, and the bed reaches full saturation at 6.669 s.

Call that a warm-up if you like. The authors do not: they put the wetting stage at up to a third of the time a shot takes, and note that models of espresso extraction have mostly skipped it.

Ponding is a word borrowed from soil science

Hydrologists use it for rain on dry ground: the soil drinks everything at first, then loses the capacity to keep up, and a puddle forms. A coffee bed behaves the same way. The pump on this machine, measured in steady flow with no portafilter fitted, moved 317 ml a minute. The bed, 9.975 mm deep in a 59 mm basket, cannot pass anything close to that while it is still filling, so water stacks up in the 7.8 mm above it.

That standing water is not a nuisance, it is the mechanism: the deeper it gets, the harder it presses, and that pressure drives the wetting front down through the grounds. The 15 bar quoted for this machine is the manufacturer figure the authors handed their model, not something anyone measured, and the pressure a machine advertises was never a description of what the bed feels at second two.

A coarser grind changed the shape of the front

The experiment ran twice, once with grounds mostly between 300 and 1000 micrometres and once with everything below 300. Same 10 g, same 12 kg on the tamp. The coarse bed still came out deeper and more porous, which is the first thing worth noticing: grind size changes the shape of the puck before any water arrives.

Water then went through that coarse bed faster, and less evenly. At second 5 the front was further ahead near the walls of the basket than through the middle, uneven enough that the authors kept their one-dimensional model to the fine grind only. Patchy wetting is what channelling grows out of, though this work stops short of that claim and says so plainly. Grinders shift the picture in other ways too: chilling the beans before grinding narrows the spread of particle sizes without making them finer.

What the scans still cannot settle

The fitted porosity came out at 0.322, which the authors themselves call a little low against the 0.4 to 0.6 the literature suggests. One reading of that gap is that some of the pore space inside the grains never fills during the wetting stage at all, either because the water has not reached it or because those pores are closed to it.

The literature on swelling does not agree with itself either. One study cited in the paper has particles swelling by 15%, with 60 to 80% of that inside 30 seconds. Another finds nothing significant between 30 seconds and 5 minutes. The shape of the front was a surprise too. Earlier work assumed constant flow or constant pressure, which would give a front moving in a straight line or as the square root of time. This one ran fast, slowed as the bed pushed back, then sped up again before reaching the bottom.

So watch the gap tomorrow morning between the pump starting and the first dark liquid in the cup. A pre-infusion setting stretches exactly that stage, and nothing here measured what that does in the cup. What the scans show is that those seconds are not dead time before espresso extraction begins. The authors tie early wetting to the first rapid release of solubles, so part of the shot is decided while half the bed is still dry.

Does the coffee in the basket deserve those seconds?

None of this rescues coffee that went stale in a cupboard. A bed only gives up what is still in it, and beans roasted months ago have less to give however evenly the water arrives. Santo Café is a single origin Mexican arabica, roasted fresh, which is the part of a shot that no machine setting can correct for.