How Does a Siphon Work: Gravity, Not Atmospheric Pressure

Water That Climbs Over the Wall of a Tank

Drop one end of a water-filled garden hose into a brimming fish tank, run the other end down to a bucket on the floor, and the water moves. It climbs over the rim, over the top of the loop, and pours into the bucket until the tank is empty. No pump, no electricity. The only rule is that the outlet must sit lower than the water surface in the tank; stand the bucket on a table above the tank and the flow stops dead.

That is a siphon, and Egyptian reliefs from around 1500 BC already show them drawing liquid out of jars. The explanation almost everybody is taught goes like this: the air outside presses on the tank surface with about 101 kPa of atmospheric pressure, pushing the water up the rising leg; gravity takes over on the way down. It is tidy, it is in a great many textbooks, and it sat in the Oxford English Dictionary for ninety-nine years. It is also wrong in exactly the part that matters.

The Dictionary That Got It Wrong

In May 2010 Stephen Hughes, a physicist at Queensland University of Technology, published a short paper in Physics Education (volume 45, pages 162–166) about a siphon used to refill a lake in South Australia. In it he noted that the OED entry for siphon — written in 1911 and still unrevised — credited the device’s operation to atmospheric pressure. He went looking through other dictionaries and found none that named gravity instead.

The story travelled, and Oxford’s editors answered carefully rather than conceding the point. They observed that there is continuing debate among scientists about which account is correct, and said the revised entry would reflect that debate. A letter in Physics Today the following year argued the opposite case, that gravity is already implicit in any mention of atmospheric pressure, since without gravity there would be no atmosphere to exert pressure at all.

What Actually Drives the Flow

Here is the mechanism. The tube is full of liquid hanging as two columns either side of the crest. The descending column, from crest down to outlet, is longer than the rising column from source surface up to crest, because the outlet sits below the source surface. Gravity pulls harder on the longer one, and the resulting pressure difference between the two ends of the tube is what drives the flow.

You can get the outlet speed in two lines. Apply Bernoulli’s equation between the source surface, where the liquid is nearly stationary at atmospheric pressure, and the open outlet, where it leaves as a free jet at atmospheric pressure too. The two pressure terms are identical, so they cancel — which is already the whole argument against air pressure being the driver. What remains is ½ρv² = ρgh, so

v = √(2gh)

where h is the vertical drop from the source surface to the outlet. That is Torricelli’s law, the same expression as for water jetting from a hole in a tank wall. Notice what is absent: the height of the hump. A 2.5 m drop gives v = √(2 × 9.81 × 2.5) ≈ 7.0 m s⁻¹ whether the crest stands 10 cm above the source surface or 4 m above it.

This is the single most useful fact about siphons. Raising the hump does not slow the flow. It only decides whether the siphon can run at all.

Source surface3.0 m Crest (hump)5.0 m Outlet height0.5 m Tube bore12 mm LiquidWater, 20 °CEthanol, 20 °CMercury, 20 °C

Flowing. Outlet speed 7.00 m s⁻¹, delivering 0.79 L s⁻¹. Crest pressure 57.1 kPa, which is 54.7 kPa of headroom above the vapour pressure of water.

Atmospheric Pressure Sets the Ceiling, Not the Speed

So atmospheric pressure is not the engine. It is the roof. Raise the crest and the pressure at the top of the tube falls. Keep raising it and the crest pressure reaches the vapour pressure of the liquid, the liquid boils at room temperature, a bubble forms, and the column separates. The siphon stops.

Where that happens follows from one line of hydrostatics. Set the air pressure equal to the weight of liquid it can support, p = ρgh, and solve for h. With p = 101325 Pa, ρ = 1000 kg m⁻³ and g = 9.81 m s⁻², h = 10.33 m. It is Torricelli’s number, and why no suction pump can draw water from a well deeper than about ten metres. The real ceiling is lower: water’s vapour pressure at 20 °C is 2.34 kPa, giving a limit of 10.11 m, and dissolved air breaks the column earlier still. That failure is cavitation, the instability that also erodes ship propellers.

Run the same arithmetic with mercury, 13.5 times denser, and you get 0.76 m — 760 mm, precisely the height of a mercury barometer. A barometer is a siphon pushed past breaking point on purpose: the column has snapped, leaving a vacuum at the top, and what stands is as much liquid as the weight of the atmosphere can hold up.

Cohesion or Pressure? A Live Argument

Hughes’s paper also proposed a replacement, the chain model: the molecules are linked by intermolecular forces into a chain draped over a pulley, the heavier side dragging the lighter side up. Richert and Binder went after it in The Physics Teacher in 2011 (volume 49, pages 78–80) with four cheap experiments and concluded it was disproved; their own account rested on Bernoulli’s equation, with gravity and pressure doing real work and intermolecular forces none.

Sharper evidence came in 2018, when Jumper and Dy siphoned carbon dioxide gas. A gas has no cohesive chain to pull on, yet dense CO₂ in a tube runs from one vessel to a lower one just as water does, because what matters is that the fluid has weight. Whatever drives an ordinary siphon cannot require cohesion.

And yet siphons do run in a vacuum. Nokes reported it in 1948 with water, dibutyl phthalate and mercury, though later critics argued those tubes sat at the liquids’ own vapour pressure, not true vacuum. Boatwright, Puttick and Licence closed the question in 2011: a siphon can run in vacuo, provided the liquid is degassed and the surfaces clean. In 2015 Boatwright, Hughes and Barry ran a water siphon 15 m tall at sea level, far above the barometric limit, by degassing the water so that it could not cavitate.

Treat this as two regimes rather than one contradiction. In an ordinary siphon full of tap water the pressure inside the tube stays positive, the liquid is never in tension, and atmospheric pressure is what stops it cavitating — so the everyday siphon genuinely does depend on the atmosphere being there. It just does not depend on it for the push. In the tall degassed experiments the liquid sits at genuinely negative pressure and its own cohesion does that job instead. Both are right about their own regime; neither is the general reason a siphon flows, which remains the height difference between the columns.

Where You Actually Meet One

Most people operate a siphon several times a day without noticing: the trap behind a toilet bowl is one, and the flush works by dumping water in fast enough to fill that trap, at which point the siphon primes itself and empties the bowl far more thoroughly than pouring water in could. Aquarium keepers use a hose for water changes, and the same trick empties fuel tanks. In karst country whole springs behave this way: a water-filled conduit in the limestone acts as a natural siphon, running intermittently as the feeding reservoir drains and refills.

A relative worth knowing is the thermosiphon, which runs on density rather than geometry: in a solar hot-water panel or passive cooling loop, heated fluid becomes less dense, rises, and is replaced by cooler fluid from below, circulating without a pump. That is buoyancy driving a loop, not column height, so the usual 10 m ceiling does not apply. Real siphons also lose head to viscous drag, which is why a long narrow hose delivers noticeably less than √(2gh) predicts.

One famous case is not a siphon at all. Water in the xylem of a tall tree rises far past 10 m — a redwood lifts it 100 m — and this is sometimes called siphoning. It is not: there is no return leg and no downhill outlet. The water is pulled up under tension by evaporation at the leaves, held by cohesion in vessels narrow enough to resist cavitation. The heights exceed the barometric limit precisely because pressure is not doing the lifting.

The Part Worth Keeping

If you want one sentence, make it this: a siphon runs on the difference in height between its two liquid columns, and atmospheric pressure sets the maximum height of the hump, not the speed of the flow. An explanation can predict the right behaviour in every case you normally meet and still name the wrong cause — and it took a dispute with a dictionary, a gas siphon and a 15 m tube of degassed water to pull the two apart.

Sources and Further Reading

• Stephen Hughes, “A practical example of a siphon at work”, Physics Education 45(2), 162–166 (2010)
• Oxford University Press, “On the definition of siphon” (OUPblog, 2010)
• A. J. W. Richert and P. M. Binder, “Siphons, Revisited”, The Physics Teacher 49(2), 78–80 (2011)
• J. D. Boatwright, S. Puttick and P. Licence, “Can a Siphon Work In Vacuo?”, Journal of Chemical Education 88(11), 1547 (2011)
• A. Boatwright, S. Hughes and J. Barry, “The height limit of a siphon”, Scientific Reports 5, 16790 (2015)
• Corrigendum: The height limit of a siphon, Scientific Reports 7, 46792 (2017)
• A. Vera et al., “Negative Pressures and the First Water Siphon Taller than 10.33 Meters”, PLOS ONE 11(4), e0153055 (2016)
• J. Jumper and E. Dy, “What the CO2 Siphon Tells Us”, The Physics Teacher 56(6), 373 (2018)
• S. Hughes and S. Gurung, “Exploring the boundary between a siphon and barometer in a hypobaric chamber”, Scientific Reports 4, 4741 (2014)