The Verification Venue · Boston · 15 January 1919
Fast Because of Gravity, Deadly Because of Cold
Just after 12:40 on an unseasonably warm January afternoon, a tank holding 2.3 million US gallons of molasses tore open in Boston's North End. The wave killed 21 people and injured about 150. A century of retelling says it was fast and it was sticky — and both are true, but they are two different pieces of physics, and only one of them cares about the cold.
Here is the surprising part, and you can watch it happen. The wave was fast for a completely ordinary reason: a tall column of liquid let go, and gravity did what gravity does to a burst dam. The speed of the front is set by the height of the molasses and by g — nothing else. What made it lethal was separate: as the molasses cooled toward a Boston January, its viscosity climbed steeply, and the thickening set the trap. Cold did not slow the wave. It arrived afterward and held people fast. Drag the temperature below and watch the two readouts refuse to move together.
Front speed of the wave
38 mph
energy speed √(2gH); dam-break front 2√(gH) = 54 mph
unchanged by coldThe grip — molasses viscosity
×3 vs warm
μ = 12 Pa·s · about 12,000× water
climbs as it coolsThe fresh batch went in warm two days before. Drag toward 0 °C — a Boston January — and the grip triples, then triples again. The speed doesn't flinch.
How tall the column of molasses stood. The 2.3 M gal filled the 50 ft tank to about 14.7 m. This — and only this — sets the speed.
The speed readout is computed from two textbook free-surface flow laws, and you can read the whole illusion off them: neither one contains viscosity. A parcel of liquid that falls the height of the column arrives moving at the energy speed U = √(2gH); the very tip of a collapsing column — the classic 1892 Ritter dam-break front — runs faster still at U = 2√(gH). Plug in the current head:
That is the Gödel–Goodstein move this page is built on. The disputed hundred-year-old eyewitness figure — “the wave came on at 35 miles an hour” — is not a measurement anyone took. It is a memory. And yet the dam-break law regenerates it from g and a height: 35 mph is exactly the energy speed of a column about 12.5 m tall, comfortably inside a 15.24 m tank. The physics does not prove the witnesses right; it shows their number was never unreasonable. In 2016 Nicole Sharp with Jordan Kennedy and Shmuel Rubinstein put the modern gravity-current numbers at 15–17 m/s (34–38 mph) — the same band.
Now the second law of the disaster. Molasses viscosity rises steeply as it cools — roughly threefold for every 10 °C it drops. That is a property of the liquid, present even for a perfectly ordinary Newtonian fluid; it has nothing to do with speed. Model it as an Arrhenius climb anchored on that sourced figure:
So the two readouts answer two different questions. Why was it fast? Gravity and geometry, with viscosity nowhere in the equation — the Reynolds number at the front stays in the thousands across the entire temperature slider, so inertia rules and the cold cannot slow the front. Why was it deadly? Because once the wave stopped, the molasses cooled and thickened around whatever it had caught, and a person cannot climb out of a fluid that is tens of thousands of times more viscous than water. Two regimes. One ran on g; the other ran on T.
The check — every number recomputed in front of you
The speeds below fall out of g and the head alone; the viscosities fall out of the sourced “triples per 10 °C” law. The green column is the cross-check, not a claim. The row in amber is your current slider setting.
| estimate | law | m/s | mph |
|---|
| T °C | μ (Pa·s) | × vs warm | × water | Re at front |
|---|
Re stays far above 1 at every temperature, so the flow is inertia-dominated and the inviscid speed holds — cooling multiplies the viscosity sixteenfold and the front speed does not care.
| quantity | from | computed | published |
|---|
Run it yourself: node research/the-great-molasses-flood/verify.mjs. The speed laws are exact inviscid results and carry no viscosity term; the viscosity law is a representative fit anchored on the published ×3-per-10 °C figure, and the apparatus below names exactly where it is a model.
The cold ran the disaster at both ends
There is a second, entirely separate cold effect — and it comes from a different analyst. The structural engineer Ronald Mayville, reviewing the tank in 2014, found the walls were roughly half the thickness the stored load required, and that the steel was low in manganese, giving it a high ductile-to-brittle transition temperature: it turned brittle below about 59 °F (15 °C). A Boston January is well under that. So cold plausibly helped fracture the tank — Mayville's finding — and then cold helped the spilled molasses trap its victims — Sharp's finding. Two cold effects, two independent analyses, at opposite ends of the same afternoon. Neither is the single “cause;” the root was a tank built too thin and never properly water-tested — filled with only about six inches of water instead of a full hydrostatic fill.
What the record supports, what is estimate, and what is a model
Exactly true (the hard record)
Date, place, scale, toll. 15 January 1919, about 12:40 pm, Commercial Street, North End, Boston. The tank held 2.3 million US gallons (8,700 m³), stood 50 ft tall and 90 ft across, and was owned by the United States Industrial Alcohol Company through its Purity Distilling subsidiary (bought 1917). 21 people died; about 150 were injured. The deaths were by drowning and suffocation in the molasses, by crushing debris, and — in at least one case — by asphyxiation hours later. We state the count and causes plainly and leave it there.
The speed laws. U = √(2gH) and U = 2√(gH) are exact results for inviscid free-surface flow (the energy/Torricelli speed and the Ritter 1892 dam-break front). They contain no viscosity, so the computed front speed is independent of temperature — that independence is exact in the model, not a fudge, and the large Reynolds number is what licenses the inviscid model here.
The viscosity climb. Molasses viscosity really does rise roughly threefold for each 10 °C of cooling in the relevant range, and blackstrap molasses is on the order of thousands of times more viscous than water — the property that made the aftermath lethal.
Estimate, not measurement
“25 ft high” and “35 mph.” These are contemporary eyewitness estimates, not instrument readings — no one clocked the wave. Modern modelling finds them physically plausible (Sharp, Kennedy & Rubinstein: 15–17 m/s), and this page brackets them from first principles. That is the honest status: not measured facts, and not myth either. The page reproduces the number; it does not certify the memory.
The provenance caveat
Not a journal paper. The widely-cited “2016 Harvard study” is a conference talk and abstract — 69th Annual Meeting of the APS Division of Fluid Dynamics, abstract L27.8 — presented by Nicole Sharp (of the fluid-dynamics site FYFD) with Jordan Kennedy and Shmuel Rubinstein of Harvard, together with a rheological survey of molasses. It is not a peer-reviewed journal publication, and we do not dress it up as one.
The overclaim we refuse to repeat
“It was non-Newtonian, so it trapped them” overshoots. Molasses is shear-thinning (non-Newtonian), but the study found that effect was extremely small during the flood. The deadly viscosity rise was temperature-driven — an effect present even for an ordinary Newtonian liquid. Saying the non-Newtonian nature did the killing swaps one tidy wrong story for another; the real killer was simply cold, thick liquid.
The trigger is not settled
Root cause vs contributing hypotheses. The root cause is structural: under-designed walls (about half the required thickness) on a tank that was never hydrostatically tested — Arthur Jell filled it with only about six inches of water rather than a full load, so leaks and weld faults went undetected. Contributing hypotheses — not proven single causes — include CO₂ pressure from fermentation and thermal expansion from the warm fresh batch added two days prior; Mayville's brittle-steel analysis is a further, separate structural finding. No one of these should be stated as the settled trigger, and we don't.
Free choices in the instrument (named)
The viscosity anchor. We take μ ≈ 4 Pa·s at 25 °C (Sharp's “about 4,000× water”, with water ≈ 1 mPa·s) and let it climb by the sourced factor of 3 per 10 °C. Real molasses is super-Arrhenius — the factor grows at lower temperatures — so this simple law, if anything, understates the cold's grip below about 10 °C. Density is Sharp's 1.5× water (1,500 kg/m³); some sources give ≈1.4× — the choice barely moves the Reynolds number and does not enter the speed at all. The energy and dam-break speeds are idealised inviscid free-surface flows; the real front sits between a slower gravity-current spreading front (roughly 23 mph on the 7.6 m wave head, with reduced gravity g′ ≈ g because air is negligibly light) and these upper laws — the eyewitness 35 mph lands inside that whole bracket. The “effective head” that launches the front is somewhat less than the full fill.