Between 1942 and 1946 the United States emergency shipbuilding program saw nearly 1,500 significant hull fractures across its all-welded merchant fleet, and on 24 November 1943 the worst-case form of that failure killed: the Liberty ship SS John P. Gaines broke clean in two and sank off the Aleutian Islands in the cold North Pacific, with the loss of 10 lives. The cause was not enemy action, not overloading alone, and not bad seamanship. It was brittle (cleavage) fracture of low-toughness steel that, below its ductile-to-brittle transition temperature, snapped without yielding — a crack that initiated at a stress raiser and ran the length of the hull through continuous welded plate that gave it nothing to stop against.
The Liberty ship and its tanker counterpart, the T2, were welded rather than riveted because welding was faster, used less steel, and could be done by an unskilled wartime workforce trained in weeks. That choice met the production target — 2,710 Liberty ships in under four years — but introduced a fracture mode the riveted hull did not have: a welded hull is metallurgically continuous, effectively one sheet of steel, so a crack that starts anywhere can propagate uninterrupted from gunwale to keel, where a riveted seam would have blunted and arrested it. The steel was the second half of the problem: rolled to a chemistry high in sulfur and carbon and low in manganese, it had poor notch toughness and a ductile-brittle transition temperature that, in winter North Atlantic and North Pacific water, the ship routinely operated below.
The U.S. Board of Investigation convened by the Secretary of the Navy in April 1943, whose third and final report issued on 15 July 1946, fixed the mechanism in service-wide terms: of 4,694 merchant ships welded during the emergency program, 970 sustained fractures, attributable to notches in steel that was notch-sensitive at low operating temperatures. The cracks favoured a specific detail — the square corner of a cargo hatch, often coinciding with a welded seam, where two stress concentrators stacked. From that notch, in cold water, a cleavage crack could initiate at a stress far below the steel’s nominal strength and run the full beam of the ship. The remedy was material and structural, not operational, and the work — with Constance Tipper’s Cambridge demonstration of the transition-temperature mechanism — grew into modern fracture mechanics. The case did not produce a trial; it produced a discipline.
At roughly 11:00 p.m. on 16 January 1943, while moored at the fitting-out dock of the Kaiser Swan Island shipyard on the Willamette River at Portland, Oregon, the brand-new T2 tanker SS Schenectady cracked almost in two with a report heard a mile away; no one was killed and no one was hurt, because the ship lay in still water under no sea load, but the hull failed by brittle fracture of notch-sensitive steel in near-freezing weather — a crack that ran across the deck, down both sides, and nearly through the bottom in a fraction of a second. The vessel had been delivered only seventeen days earlier, on 31 December 1942, after sea trials without incident. There was no storm, no cargo overstress, no collision; the failure was internal to the steel and the welds, which is precisely why it became the textbook emblem of wartime hull fracture.
The Schenectady was an all-welded ship, one of thousands of emergency-program merchant vessels the United States built at unprecedented speed by replacing slow riveted construction with continuous welding. Welding made the hull a single monolithic body of steel with no riveted seams to interrupt a running crack — and that continuity is what doomed it: when a brittle crack started, nothing in its path stopped it. The night air had fallen to about minus 3 °C and the river to about 4 °C, and at that temperature the ship-plate steel of the day — high in sulphur, low in manganese, with a ductile-to-brittle transition temperature often well above freezing — had no toughness. It behaved like glass.
The crack initiated at a weld near a stress concentration, propagated through the cold, notch-sensitive plate, and split the hull just aft of the superstructure. The deck and sides parted; the ship jack-knifed on the bottom plating that alone remained intact, the midbody rising clear of the water while bow and stern sagged toward the river bottom. The U.S. Coast Guard attributed the failure to faulty welding; a Board of Investigation weighed “locked-in” residual stresses, the sharp temperature drop, and design discontinuities. Later metallurgical work — most influentially that of Constance Tipper at Cambridge — settled the mechanism: the steel itself went brittle in the cold, and the welds and notch-bearing details merely gave the crack a place to start. Of 4,694 welded merchant ships in the emergency program, about 970 sustained hull fractures and nineteen broke completely in two; the Schenectady survives as the cleanest demonstration because it failed with zero external load, isolating the material and the weld from every other variable.
At about 12:30 in the afternoon of 15 January 1919, on Commercial Street in Boston’s North End, a 50-foot-tall riveted steel tank holding roughly 2.3 million US gallons of molasses split apart and discharged its entire contents in a wave that killed 21 people and injured around 150. The cause was not fermentation pressure, not sabotage, and not an explosion. It was a brittle fracture of the tank’s thin, low-manganese steel shell, initiated at the over-stressed rivet holes beside a manhole at the base, propagating at the speed of sound through plates that were never thick enough and never inspected.
The tank belonged to United States Industrial Alcohol (USIA), through its subsidiary Purity Distilling, and had been thrown up in 1915 to store the molasses that fed wartime demand for industrial alcohol and munitions. It was 50 feet high and 90 feet in diameter and held about 13,000 short tons of liquid when full. The man who ordered and oversaw its construction, USIA treasurer Arthur Jell, had no architectural or engineering training. He did not commission stress calculations, did not have the steelwork checked by an engineer, and skipped the standard practice of filling the completed tank with water to test it under load — running only a few inches of water into the bottom rather than the full hydrostatic test that would have revealed the leaks before they revealed themselves.
The shell leaked from the day it was filled. Molasses wept from the seams so persistently that USIA painted the tank brown to disguise the runs, and neighborhood children collected the drips. At least one employee warned management that the structure was unsound; the company’s response was to re-caulk. The forensic verdict, settled across a three-year court audit at the time and confirmed by modern fracture-mechanics analysis decades later, was unambiguous: the wall plates were roughly half as thick as the load required, the steel was deficient in manganese and therefore brittle, and the failure began as cracks at the rivet holes where stress concentrated. On a January day the metal was well below its ductile-to-brittle transition temperature, and a structure already loaded near its limit failed without warning and without yielding.
The disaster ended in one of the longest civil proceedings in Massachusetts history. A court-appointed auditor found USIA liable after years of hearings, and the company paid out about $628,000 in settlements — on the order of $11 million in present money. The litigation also did what the absent engineer never had: it forced the lesson into law. Massachusetts and then jurisdictions across the country began requiring that major structures be designed, signed, and sealed by a licensed engineer or architect, the regulatory ancestor of the professional stamp.
At about 11:00 on the cold winter morning of 10 July 1962, a suspended span of the fifteen-month-old King Street Bridge over the Yarra River in central Melbourne sagged and broke up beneath a 47-ton low-loader that was within the bridge’s posted load limit; by chance no one was killed or injured. The cause was not overload, not a design under-strength, and not corrosion. It was a brittle fracture — a fast, low-energy crack that ran through all four main girders of the span — initiated at the toes of transverse fillet welds where cover plates had been welded onto the tension flanges, in a notch-sensitive high-tensile steel embrittled by hydrogen, locked-in welding stress, and a near-freezing temperature.
The bridge had been designed in 1959 by the consulting engineers Hardcastle & Richards for the contractor Utah Australia, fabricated in welded BHP high-tensile low-alloy steel by the subcontractor Johns & Waygood, and inspected for the Country Roads Board. The four suspended girders of the failed span, designated W.14-1 to W.14-4 and roughly 100 feet long, each carried thickening cover plates welded to the bottom (tension) flange to add section where the bending moment was highest. The transverse welds that closed off the ends of those cover plates were the fatal detail. They left a sharp geometric notch exactly where the steel was most highly stressed in tension, and in the heat-affected zone beside each weld the metal had been hardened, hydrogen-charged, and cracked during fabrication.
Brittle fracture needs three things together: a flaw, a tensile stress, and a steel cold and notch-sensitive enough to run the crack instead of yielding around it. The King Street span supplied all three at once. Pre-existing cracks sat at the weld toes; the cover-plate ends concentrated the tension; the high-carbon, high-tensile plate had poor notch toughness and was sitting at a temperature near its transition into brittle behaviour. When the heavy low-loader drove onto the western carriageway and raised the live-load stress, the cracks ran. The span dropped about a foot. The forensic verdict, established by the 1963 Royal Commission chaired by Sir George Barber, was unambiguous: brittle fracture from defective welds at the cover-plate terminations, in a steel and a detail that should never have been welded that way without preheat and toughness control.
The failure was, in retrospect, built in. The cracks at the weld toes had existed since fabrication and were not found by the inspectors of either Johns & Waygood or the Country Roads Board. Neither the contractor nor the fabricator fully grasped that high-tensile steel demanded a different welding discipline from ordinary mild steel — controlled heat input, preheat, low-hydrogen practice, and crack inspection. The bridge did not so much fail under traffic as wait for the first cold morning and the first heavy load to arrive together.