A structural and metallurgical perspective of design practices at the time and fracture evidence based on current understanding.
By Ronald A. Mayville, Ph.D., P.E.
On Wednesday, Jan. 15, 1919, at about half past noon, a tank containing 2.3 million gallons of molasses burst, sending a wave of sticky liquid through the streets of the North End of Boston. The spill killed 21 people, drowned several horses, and caused substantial structural damage to nearby buildings and the adjacent elevated railroad. The incident resulted in a three-year trial in which the court auditor found the operator of the tank liable, but was frustrated by more than 20,000 pages of conflicting testimony from dozens of expert witnesses.
The tank, owned by U.S. Industrial Alcohol, was used to store molasses shipped from the Caribbean for the production of a munitions ingredient. It was situated close to the harbor near Copp’s Hill Burying Ground, Boston’s second oldest graveyard and a popular tourist attraction today. Figure 1 shows a photograph of the imposing tank a few months before the failure. The first fill occurred in February 1916, almost three years before the end of WWI. Abundant information is available on the accident, largely in the form of transcripts for the case of Dorr vs. U.S. Industrial Alcohol. (Dorr, of the law firm Hale and Dorr, was the named plaintiff who had his house damaged from the tank failure.)
The case was, without doubt, an expensive one. The many experts included several from MIT and Harvard, split about evenly between sides. Organizations carried out large-scale tank explosion tests and wide panel fracture tests that included a replica manhole from which nearly everyone agreed the fracture initiated. An excellent account of the general circumstances surrounding the accident, with particular attention to the historical and social aspects at the time — anarchists were blamed by some for setting off an explosion in the tank — is given by Stephen Puleo (2003).
Analysis of the accident scene began immediately after the failure. Figure 2 is a photograph taken at the disaster site. Figure 3 is a schematic of the structures that were adjacent to the tank, as well as the approximate initial locations of some of those killed when the tank burst.
The weather conditions at the time of the failure were partly cloudy, about 40 °F (4 °C) with a wind of at most 24 mph, absent of snow or rain. The tank had been filled 29 times, but only four times to near maximum capacity. The last fill to near capacity occurred two days prior to the failure.
A 6-ton piece of the tank that included part of the manhole was found the furthest from the tank, about 200 feet, in the playground in the right foreground of Figure 1. Its distant location was considered a key piece of evidence that an explosion had occurred.
The tank was designed to contain 2.5 million gallons of liquid and erected on a very short schedule during the winter of 1915. The riveted construction was 90 feet in diameter, 50 feet tall, made of open hearth steel varying in thickness from 0.67 inch for the first ring at its base to 0.31 inch for the seventh ring at the top. The vertical joints between the pieces in the first course were made with double butt joints with a triple row of rivets on each side of the joint in a diamond pattern (Figure 4). The vertical joints in the remaining rings were made by lap joints using triple rivets in rings 2 and 3 and double rivets in rings 4 through 7. All horizontal joints were connected by lap joints with a single row of rivets. The connection between the bottom of ring 1 and the bottom of the tank was made through a 4-inch by 4-inch by 0.75-inch angle. The rivet holes were evidently punched and not subsequently reamed, likely to save time.
Riveted construction was the primary method of making connections at that time and, although there were apparently no tank design standards in that era, engineering textbooks described methods to design such joints (c.f. Benjamin and Hoffman, 1913; and Merriman, 1914). These books gave guidance on design stresses for the rivets and plates they joined. A factor of safety of four on ultimate strength was recommended for relatively static load applications such as buildings, and Merriman suggested a factor of four for steel (see Table 1).
In testimony, reference is made to the Boston Building Law, which in the 1915 provisions included the following limits on stress for steel:
Tension: 16,000 psi
Shearing (for rivets): 10,000 psi
Direct bearing (for riveted joints): 18,000 psi
The law included a note that the ultimate strength of steel must be 55-65 ksi, that the yield strength (elastic limit) must be at least one-half the ultimate strength, and that the percent elongation in 8 inches must be at least 1.4 x 106, divided by the ultimate strength. (For an ultimate strength of 55 ksi, this equates to a minimum elongation of 25.5 percent.) All of the steel used to fabricate the tank satisfied these values.
The practice of the time for the design of riveted joints was to check for three failure modes: tensile failure through a row of holes; shear failure of rivets; and bearing failure between the rivets and their holes. Applying the design methodology from Benjamin and Hoffman (1913) to the joint configuration of Figure 4 gives the calculated stresses shown in Table 2. The allowable stresses were exceeded.
The manhole in ring 1 was key to the investigation, for the physical evidence (fracture surface markings) strongly indicated that the break initiated at either the very top (12 o’clock position) or the bottom. The 20.5-inch-diameter hole was reinforced only on the outside by the 0.375-inch-thick, 4-inch-wide flange of the manway collar. In addition, the lap joint made in the manhole neck flange at the tank plate was at the top of the manhole (Figure 5). Finite element analysis indicates that the rivet hole at the 12 o’clock position above the manhole was very highly stressed and could have experienced significant plastic deformation for the level of molasses at the time of failure.
Two metallurgical aspects of the molasses tank steel are of note: the likely brittle behavior at the tank failure temperature, 40 °F (4 °C), and the observation of a microstructural feature called Neumann bands. The latter was used by many defense witnesses as a key piece of evidence to support an explosion as the cause of failure.
Most structural steel production in 1915 was by the open hearth process. The material exhibited very good strength with high ductility. At that time, toughness was not specified in material procurements. The chemistry for the steel in the plate containing the manhole is shown in Table 3, compared with the chemistries of two modern steels.
Note the particularly low value of manganese in the steel used to construct the molasses tank. Chemistry, thermal processing, and quality of production all affect the fracture properties of steel. For the class of steels corresponding to the molasses tank, known as pearlitic steels, carbon has the greatest effect on the propensity for brittle behavior: the higher the carbon, the higher the temperature at which there is a transition (the transition temperature) from brittle to ductile behavior. Manganese does not have as strong an effect as carbon, but, generally, the lower the manganese content, the higher the transition temperature. Some data (Rineholt and Harris, 1951) suggest the transition temperature for the molasses tank steel could have been as high as 59 °F (15 °C), significantly above the operating temperature at the time of failure.
The amount of testimony in the molasses tank failure litigation related to Neumann bands is astounding. Neumann bands are a feature first discovered in meteoric iron (by Neumann). They are narrow bands, a few micrometers wide, usually within a grain, and are now known as deformation twins because they are created by mechanical deformation and have a mirror image crystal structure across twin boundaries. Substantial research, even before the molasses tank failure, indicated that this feature was only produced in low carbon steels by explosive loading or at extremely low temperatures, such as for liquid air (-320 °F). Expert after expert for the defense produced micrographs of Neumann bands in cross sections taken near the primary fracture. We know today that a rapidly propagating crack in brittle low-carbon steel can produce Neumann bands without explosive loading. Crack velocities in brittle steels can reach values as high as 3,000 feet per second.
In 1925, six years after the tank failed and three years after testimony began, the auditor awarded the plaintiffs $300,000 ($4 million in 2014 dollars) for a tank that cost U.S. Industrial Alcohol $15,000 to fabricate. The auditor was swayed by the fact that the tank had been built with very low safety factors relative to the practice of the day, even though the exact cause of the failure was not determined.
Examining the evidence with the structural and metallurgical knowledge we have today indicates he decided correctly. The evidence points to a scenario in which the fracture initiated above the manhole in a highly stressed rivet hole, possibly damaged from the rivet hole punching process. Perhaps the last few and greatest fillings of molasses were sufficient to then grow a crack to a critical size in a susceptible, brittle material, leading to the complete rupture of the tank, and the loss of 21 lives.
Puleo, Stephen, 2003, Dark Tide, Beacon Press, Boston.
Benjamin, C.H. and Hoffman, J.D., 1913, Machine Design, Henry Hold and Company, New York.
Merriman, Mansfield, 1914, Mechanics of Materials, John Wiley & Sons, Inc., New York.
Colby expert report files.
Rineholt, J.A. and Harris Jr., W.J., 1951, Effect of Alloying Elements on Notch Toughness of Pearlitic Steels, Trans. ASM, 43, pages 1175-1214.
Ronald A. Mayville, Ph.D., P.E., senior principal, Simpson Gumpertz & Heger Inc., has been active in the field of materials and structural integrity for more than 30 years.