Espresso machine pressure: 60 shots found the ceiling at 5 bar
Sixty measured shots put the flow ceiling at about 5 bar, and the coffee bed itself is what puts it there.
Push harder, get more: the promise written into every pressure gauge on an espresso machine. A physics group in Warsaw spent 60 shots testing it and found a ceiling instead. Past roughly 5 bar the espresso puck stopped passing more water, however hard the pump leaned on it, and at the top of the range pushing harder made the shot slower. The coffee bed is not a still filter but a wet elastic solid that squeezes itself shut under pressure.
Where espresso machine pressure stops buying flow
The rig: a two-group Sanremo Zoe Competition, a load cell under the cup, a pressure sensor at the pump outlet, both logging at 10 Hz. Sixty coffees went through at 11 basket pressures from 1 bar to 12 bar. Every shot ran at least 120 seconds, so the bed could settle, and the espresso flow rate on record is the average of the last 10 seconds.
Below 5 bar the bed behaved as a rigid porous medium should: more pressure, proportionally more flow, a straight Darcy line. Above 5 bar it bent over and flattened. The cup filled fastest at about 5 bar, then the trend reversed outright: more pressure, less flow. It is a preprint on arXiv, version 2, dated 9 May 2026, not yet peer-reviewed.
The espresso puck is elastic, not a rigid filter
Wet coffee swells. A spent puck knocks out in one piece, nothing like the loose grounds that went in. Once the bed is saturated, water pressure and the stress in the grounds share one fixed total: every bar the pump adds partly compresses the bed instead of travelling through it. The fitted equilibrium porosity came out at 0.122, with a calibration flow of 1.90 g/s, both effective values from a model fit, not instrument readings, as the paper stresses.
Notice what nobody varied. An automatic tamper held 20 kg on every single shot, so espresso tamping could not contaminate the result. That choice has backing: a 2017 study in the Journal of Food Engineering tracked caffeine and trigonelline out of a commercial machine across a range of particle sizes and tamping pressures, and found particle size drives extraction while tamping pressure has no detectable effect. The arm on the tamper is not what compacts the bed. The pump is, and the 9 bar espresso standard sits well above where that takes over.
Pores of about 6 microns, not 200
The usual packed-bed arithmetic gives an absurd answer. Take particles of 200 micrometres, feed them into the Carman-Kozeny estimate engineers reach for, and predicted throughput lands near 2000 g/s. A real shot delivers around 1 g/s. Calculating espresso flow rate from the size of the grounds is out by a factor of a thousand.
That gap closes if the working pores are about 30 times narrower than the grounds imply, near 6 micrometres across. Swelling and confinement do it: the particles drink, expand against a basket that cannot give, and the gaps collapse. Push harder and the added pressure is cancelled by the resistance it creates. It lands in the same place as the grind setting where finer stopped raising espresso extraction yield, approached from the opposite side.
What the scans found under the espresso basket
The same pucks went through an X-ray microtomograph at 20 micrometre resolution, scanned in the basket just before brewing and a couple of minutes after. Dose 18.50 g in a 58 mm basket, 14 mm deep. Before the shot, a tidy cylinder of grounds.
After it, nothing of the kind. Delaminations run across the bed, cracks split it into slabs, and at the bottom the puck has peeled clean off the filter mesh with a layer of air in the gap.
Why a pause in the shot is the real risk
One side experiment matters most. The team held 8 bar for about 75 seconds, then cycled the brewer off and on, the pump falling to about 3 bar between bursts. Each time pressure returned, flow jumped without any additional dissolution: more water through the bed, no more coffee with it. The puck had reorganised in the gap, opening paths the water then took for free.
That is channelling, from inside the basket. The authors' conclusion is blunt: any brewing process with a temporary pause or repeated application of pressure might be prone to channelling and should be avoided. The limits belong in the same breath: this is the long-time flow regime, not the first seconds, it does not resolve how a channel forms, and all of it ran on one machine, one grind and one Brazilian roast. Espresso pre-infusion that ramps gently is a different animal from a profile that kills the pump and restarts it.
Watch the last ten seconds of the next shot. If the stream thins and speeds up rather than slowing down, the bed has already given way somewhere under there.
Which coffee shows this most clearly?
A bed that compacts this hard rewards beans that grind evenly, because an even bed resists in the same way right across its width. Santo Café works with a single origin Mexican arabica rather than a blend, which keeps one variable out of the basket: beans of one origin and one roast pack and swell more alike than a mix of several ever will. What is left is the grinder, the dose, and a pump nobody interrupts.
Taste our coffee from Mexico
Freshly roasted, delivered to your door.
